Search is not available for this dataset
article
stringlengths 4.36k
149k
| summary
stringlengths 32
3.35k
| section_headings
listlengths 1
91
| keywords
listlengths 0
141
| year
stringclasses 13
values | title
stringlengths 20
281
|
|---|---|---|---|---|---|
Rapid atrial arrhythmias such as atrial fibrillation ( AF ) predispose to ventricular arrhythmias , sudden cardiac death and stroke . Identifying the origin of atrial ectopic activity from the electrocardiogram ( ECG ) can help to diagnose the early onset of AF in a cost-effective manner . The complex and rapid atrial electrical activity during AF makes it difficult to obtain detailed information on atrial activation using the standard 12-lead ECG alone . Compared to conventional 12-lead ECG , more detailed ECG lead configurations may provide further information about spatio-temporal dynamics of the body surface potential ( BSP ) during atrial excitation . We apply a recently developed 3D human atrial model to simulate electrical activity during normal sinus rhythm and ectopic pacing . The atrial model is placed into a newly developed torso model which considers the presence of the lungs , liver and spinal cord . A boundary element method is used to compute the BSP resulting from atrial excitation . Elements of the torso mesh corresponding to the locations of the placement of the electrodes in the standard 12-lead and a more detailed 64-lead ECG configuration were selected . The ectopic focal activity was simulated at various origins across all the different regions of the atria . Simulated BSP maps during normal atrial excitation ( i . e . sinoatrial node excitation ) were compared to those observed experimentally ( obtained from the 64-lead ECG system ) , showing a strong agreement between the evolution in time of the simulated and experimental data in the P-wave morphology of the ECG and dipole evolution . An algorithm to obtain the location of the stimulus from a 64-lead ECG system was developed . The algorithm presented had a success rate of 93% , meaning that it correctly identified the origin of atrial focus in 75/80 simulations , and involved a general approach relevant to any multi-lead ECG system . This represents a significant improvement over previously developed algorithms .
Rapid atrial arrhythmias such as atrial tachycardia ( AT ) and atrial fibrillation ( AF ) can reduce cardiac output and predispose to ventricular arrhythmias and further complications , such as stroke and even sudden cardiac death [1–3] . Both AT and AF are associated with ectopic activity—rapid and irregular spontaneous excitation originating from regions of the atria other than the cardiac pacemaker , the sinoatrial node [4] . Such activity can interrupt normal sinus rhythm and mediate the development of the self-perpetuating re-entrant excitation associated with AT/AF [5] , therefore implicating an important role for ectopic activity in the initiation and recurrence of both arrhythmias . The pulmonary vein ( PV ) sleeves in the left atrium ( LA ) are usually identified as a major source of rapid ectopic activity [6–8] , and catheter ablation therapy targeting the PV sleeves is commonly used as a treatment of AF [4 , 7] . However , success rates for catheter ablation therapy are not entirely satisfactory ( about 50% in single-procedure ablation [9] ) . Consequently , repeated operations may be required , resulting in significant scar tissue in the LA . Such scarring may induce further complications , such as contributing towards a reduction in cardiac output as well as providing conduction barriers which may promote the development of re-entry [10] . Furthermore , ectopic activity is not associated with the PVs alone; focal beats have been observed to originate from multiple regions of both the left and right atria [11 , 12] . Hence , identifying the presence and location of ectopic activity is important for guiding ablation therapy , which may increase success rates and reduce the need for repeated operations . Moreover , identifying atrial ectopic activity and its origins may help in the diagnosis of early onset AF [13] and lead to timely treatment , inhibiting the development of persistent or chronic AF before the occurrence of permanent electrical and structural remodelling [13] . The electrocardiogram ( ECG ) is the most common non-invasive method of monitoring cardiac activity . The P-wave of the ECG is associated with atrial activation; irregular ectopic atrial activity may therefore be reflected as an alteration to the P-wave morphology ( PWM ) . Multi-electrode ECG systems , such a 64-lead ECG vest [14] , provide spatially detailed mapping of the body-surface potential ( BSP ) . However , it is unclear if such further detail provides significant benefits over the standard 12-lead ECG in terms of resolving the location of ectopic atrial activity . In this study , we have used a biophysically detailed computational model of the human atria and torso to investigate the correlation between PWM of 64-lead ECGs and the location of atrial ectopic activity , in order to develop a focus-location algorithm .
Previously we have developed a biophysically detailed computational model of the three-dimensional ( 3D ) human atria and torso [15–17] . The model accounts for atrial anatomy [18] including segmented regions for the major anatomical structures [16] ( Fig . 1Ai ) and detailed atrial electrophysiology including regional differences in electrical properties [16] . The model reproduces sinus rhythm depolarisation and repolarisation patterns ( Fig . 1Aii ) and has been used to study the mechanisms underlying AF genesis [16 , 17] . Implementation of the torso model proved useful in correlating PWM with the origin of atrial ectopic activity in a previous study [15] . However , detailed correlation between the two has not yet been established , and the torso geometry used in the previous study was idealised [15] . For a more comprehensive analysis of the relation between PWM and ectopic activity , a more realistic torso model must be used . In this study , we use our 3D model of the human atria and update the torso model in order to develop an algorithm to identify the location of focal ectopic activity in the atria ( Fig . 1 ) . Details of atrial model development and simulation protocols can be found in Colman et al . [16] , and in S1 Text . Two torso reconstructions are used in the present study ( Fig . 2 ) , based on segmentation of magnetic resonance imaging ( MRI ) images taken from the female and male visible human dataset [19] , by using the software ITK-SNAP [20] . Note that the atrial model does not account for gender differences in either anatomy or electrophysiology [16] and investigation of gender differences is not the aim of this study; rather , use of multiple torso geometries ensures generality of the developed algorithm . The models account for the structure and different electrical conductivities in the lungs , liver , spinal cord and blood masses [21] . The female torso model was discretised at a spatial resolution of 0 . 33mm × 0 . 33mm × 0 . 33 mm[19] , corresponding to that of the female atrial model [16] . Meanwhile the male torso model was discretised at a spatial resolution of 0 . 33mm × 0 . 33mm × 1 mm[19] . The 3D atrial model ( Fig . 1A ) [16] was then integrated into the two torso geometries and the BSP distribution was calculated through the use of a boundary element method ( Fig . 1B ) [22] . Two different positions of the atria inside the torso were used to account for variability between patients; one is based on Ho and Sanchez-Quintana [23] ( Fig . 2 Ai , Bi ) , and the second one is the position of the atria obtained directly from the segmentation of the visible human female dataset ( from which the atrial anatomical model was extracted ) ( Fig . 2 Aii , Bii ) . From the BSP , ECG signals can be derived by selecting elements of the torso mesh which correspond to the location of electrodes used in ECG systems . Ectopic focal activity was simulated by applying stimuli to various locations across all regions of the atria ( Fig . 1C ) . In this study , we replicated a 64-lead ECG system which measures the BSP on the front and back of the torso ( Fig . 1D ) as well as the standard 12-lead ECG . All leads in the 64-lead system are unipolar: the potential at the electrode is the positive terminal and Wilson’s Central Terminal[24] is the negative terminal . For each lead , the P-wave was characterised by its morphology and polarity . It was indexed as positive if the amplitude of the positive peak was greater than double that of the negative peak ( if there was one ) , and vice-versa for a negative P-wave . A biphasic P-wave is defined as one in which the second peak ( positive or negative ) was at least half of the amplitude of the largest peak ( negative or positive ) . Such a definition resulted in the best performance of our focus location algorithm ( described in the next section ) , and is not intended as a general definition for other purposes . The P-wave dipole pattern was constructed based on the maximum positive potentials ( positive pole ) and the minimum negative potentials ( negative pole ) in the body surface at every time step [14] . As the atrial activation evolves , the amplitude and spatial distribution of the poles across the surface of the body change dynamically . Furthermore , we constructed spatial polarity maps based on the polarity ( positive/negative/biphasic ) of the P-wave at each electrode location . To simulate ectopic focal activity the model was stimulated by a sequence of external supra-threshold electrical pulses applied to various locations across all different regions of the atria ( Fig . 1C ) , representing the range of ectopic foci observed experimentally [11 , 12] . Stimuli were applied to each location at both slow ( cycle length = 700ms ) and fast ( cycle length = 300ms ) rates to ensure that rate dependent changes in PWM are accounted for . In each case , the P-wave resulting from the final of three stimuli was analysed . Simulated BSP maps and ECG P-waves varied significantly with the location of the ectopic focus ( Fig . 3 ) . The P-wave polarity map offered the most effective method of quantifying such differences , offering more information than the temporal evolution of the dipole peaks while being less affected by noise than the raw P-waves . P-wave polarity maps therefore form the basis of the development of an algorithm to determine the location of an atrial focus from 64-lead ECG measurements . To relate polarity patterns to atrial anatomical sites , both the atria and the torso were divided into two sets of quadrants , four in the anterior part and four in the posterior part of each anatomical model ( Fig . 3B , C ) . For the torso model , Qt1 to Qt8 were used to label the quadrants ( Fig . 3Bi , ii ) . In the atria Qa1 to Qa8 were used ( Fig . 3C ) where each quadrant contains corresponding anatomical regions ( Table 1 ) . Note that the position of the atria within the torso had a significant effect on PWM and the P-wave polarity map ( S1 Fig . ) , and that the atrial anatomical locations associated with each atrial quadrant differ for both orientations considered . As such , patient variability in the orientation of the heart within the torso must be considered , and can be accounted for in this table rather than the algorithm itself , which operates by relating atrial and torso quadrants . Schematic illustration of the algorithm is shown in Fig . 4 , and details of the algorithm are described below: Construct the spatial polarity map . Assign a numerical value to each electrode position based on the polarity of the P-wave at that position; 2 for a negative P-wave , 1 for a bi-phasic P-wave and 0 for a positive P-wave . Take the mean average of all the values in each torso quadrant , denoted Sp . Determine the largest value of Sp across all quadrants , denoted Sp_max . If there is a single quadrant which contains this value ( Qtx , x = 1–8 ) , then the location of the atrial focus is in the corresponding atrial quadrant ( Qax ) . If there are multiple quadrants which contain Sp_max , then further analysis is required: If the value of Sp in two quadrants is equal to Sp_max , then two adjacent quadrants must be compared . Then , the quadrant Sp_max , adjacent to the larger Sp from the second comparison , will be identified as the origin . Note: for example , if both superior-right and superior-left anterior quadrants have the same Sp_max value , then the Sp in the inferior-left and inferior-right anterior quadrants are compared , as long as they are different . If the inferior left has a greater Sp , then the atrial focal is in the superior-left region . If there are 3 quadrants with the same Sp_max value , then the corner quadrant will be identified as the atrial focal origin . Note: for example , if the anterior superior-right , the anterior superior-left and the anterior inferior-left quadrants have the same maximal value , then the anterior superior-left quadrant will be the origin . If 4 or more quadrants have the same Sp_max , the adjacent quadrants with different Sp will be compared , and the quadrant with a larger Sp will be identified as the origin . Note: for example , if the four anterior quadrants have the Sp_max , a subsequent maximal Sp in the posterior quadrants will be searched . If there is one , say the superior-right posterior one , then the superior-right anterior quadrant will be identified as the origin .
Validation of the atrial activation sequence during control conditions has been discussed in [16 , 17] . In order to validate the 3D atria-torso model , we first compared the simulated BSP pattern and 12- and 64- lead ECG P-waves for the control case to experimental data obtained from eight healthy subjects . It was demonstrated that the simulated data of the 64-lead ECG ( Fig . 5 ) as well as the 12-lead ECG and BSP pattern are in fair agreement to the experimental data . Then , we further compared the simulated P-wave polarity to the experimental data . In both simulations and experimental data , the polarity of P-waves was mainly positive in the left-superior part of the body , negative in the right , inferior part of the body , and biphasic or flat in the intermediary locations ( Fig . 6A , B ) . To assess quantitatively the agreement between the polarity patterns in simulation and experiment , the polarity of the simulated P-wave at every electrode was compared with each experimental dataset . Inter-patient variability was quantified by also comparing experimental datasets to each-other . The simulation data showed a range of agreement between 87 . 1% and 94 . 5% with experimental data , comparable to the range observed within the experimental data of 81 . 5% and 93 . 7% . Furthermore , the simulated temporal evolution of the dipole location ( Fig . 6Ci ) and amplitude ( Fig . 6Cii ) agreed with experimental data [14] . Hence , the model was validated for the control condition , and suitable for investigating the correlation between ectopic atrial activity and P-wave profiles . The algorithm was developed based on results from 30 simulations with different atrial foci . It was then tested with 50 further simulations to determine its success rate ( i . e . the proportion of cases in which the algorithm correctly identified the origin of atrial focus from the P-wave polarity pattern ) . In such blind tests , the success rate was 93% . Note that pacing rate affected PWM only to a small degree , and had no effect on the P-wave polarity map ( S2 Fig . ) , hence ensuring the algorithm is appropriate for both fast and slow pacing rates . There were five cases for which the origin identified by the algorithm did not match the actual excitation site . In those cases the mismatch was a result of the definition of a biphasic P-wave , when the PWM was highly irregular . These irregularities could impact the value of the average for each quadrant , leading to a mismatch in the location of the ectopic focus , mainly when the focal origin was close to the boundary between two or more quadrants . Further refinements to the spatial resolution of the quadrants could be performed with the aim to improve the specificity of the algorithm for locating the focal origin site , by dividing each quadrant into sub-quadrants . Accordingly , the algorithm was updated as follows: if a quadrant adjacent to the quadrant with Sp_max has an Sp value close to that of the maximum quadrant ( i . e . within 0 . 1 in this case ) , then the activation focus is determined to be in the sub-quadrant that is close to the boundary between the two quadrants ( i . e . within the quadrant of maximum Sp value in close proximity to the neighbouring quadrant considered ) . Conversely , if the difference in values between the two quadrants is very large ( i . e . greater than 0 . 1 ) then the focus of the activation is determined to be within the sub-quadrant that is far from the boundary of the two quadrants . Though such a spatial refinement improved the detection accuracy in terms of the spatial resolution , the success rate of detection showed a slight decrease , down to 89% . This could be due to the limitation of the 64-lead ECG to map the BSP .
The computational model implemented for this study was an update of our previous model of the human atria and torso [15–17] . The updated model has the following advantages compared to the previous model [15 , 17]: ( i ) realistic torso meshes were used for male and female , rather than an idealised one as used in the previous studies [15 , 17]; ( ii ) a greater level of detail was considered within the torso , including the spine and liver as well as blood masses and lungs; ( iii ) various , experimentally justified orientations of the atria [23] were considered . The developed atria-torso models were validated by their ability to simulate BSP patterns , 12- and 64-lead ECG PWM , 64-lead ECG polarity patterns and the spatio-temporal evolution of the dipole peaks , all of which matched to experimental data from eight healthy patients . Note that experimental P-waves were filtered and averaged over a time period of 1 minute—this has the effect of smoothing the signals compared to the simulated P-waves , for which averaging would have no smoothing effect due to the model being deterministic and subsequent P-waves being identical . Therefore , the presented models provide a useful platform for simulating atrial excitations and their BSP patterns in variant physiological conditions . Several human atria-torso models have been developed by other groups in previous studies [25–29] , including the one by Krueger et al . [25] , in which personalised atrial geometries were implemented for reproducing accurate patient specific P-waves . The model in that study considered fat and muscle tissue , which can affect the P-wave . However , due to the difficulty in segmenting both tissue types , few models include them [25 , 30] . That model also considered soft tissues of the bowels , kidneys and spleen , which were absent in the present model . However , the simulated PWM from the present models were similar to those from Kruger et al . [25] , suggesting that these soft tissues play only a small role in affecting the polarity of the P-waves , as also suggested in a previous study [30–32] . Furthermore , agreement of PWM between simulation and experiment were similar in both studies , despite Krueger et al . being patient-specific . Though other atria-torso models have been developed for simulating body surface potential maps and multi-lead ECGs , the focus of those studies were in finding the ideal number of electrodes to obtain more information of the atria as compared to the standard 12-lead ECG system [26 , 27] , or to create a database for detecting atrial fibrillation [33 , 34] . To our knowledge , the present study is the first attempt to establish a detailed correlation between the polarity map of body surface potentials and origins of atrial ectopic focus . Comparison to previous algorithms . Focus-location algorithms have been developed previously based on the standard 12-lead ECG system [35 , 36] , including the well-established Kistler et al . algorithm [11] . However , the 12-lead based algorithms have limited effectiveness due to the smaller number of electrodes that provided incomplete information on atrial excitations . In their study , Kistler et al . reported 93% focus detection accuracy . However , subsequent studies have found a lower accuracy [35 , 36] . When we applied the Kistler et al . algorithm to simulation data of P-waves , an accuracy of 73% was achieved , which is within the 55–78% range observed in other studies [35 , 36] . In this study , we presented an algorithm for identifying atrial focal origins based on simulated 64-lead ECG system . The developed algorithm showed a higher success rate on the same data than the Kistler et al . algorithm ( 93% vs 73% respectively ) . Our results suggest that the extra level of detail provided by 64-lead ECG compared to the 12-lead ECG system was useful in accurately locating atrial focal activity . The developed algorithm has two key strengths compared to previous algorithms: ( i ) splitting the torso into two sets of quadrants means that the algorithm is not specific to an electrode array set up – any array which covers the front and back of the torso ( symmetric or asymmetric ) may be used , and the algorithm need not be adjusted . Similarly , relation of atrial anatomy to torso quadrants via a correlation table intrinsically accounts for patient variability , also without the need to adjust the algorithm itself; ( ii ) the algorithm is based on polarity patterns of the P-waves , rather than the detailed PWM . Whereas this does not provide a full level of detail as with PWM , such an approach has the following advantages: ( a ) inter-patient variations manifest as alterations in PWM but have a much smaller effect on P-wave polarity; ( b ) similarly , noise will not affect the P-wave polarity pattern but may have a significant effect on PWM , especially regarding bifidity; ( c ) we did not consider bifidity in our definition of polarity , therefore avoiding the limitations of algorithms which use bifidity , such as ambiguity in the definition of the magnitude of bump necessary to be considered bifid and the effect of noise on accentuating or reducing bifidity . Note that this was one of the primary limitations of the Kistler et al . algorithm [11] , responsible for the majority of its errors . Another possible approach for locating atrial ectopic foci is to implement an inverse solution . However , inverse solutions are computationally intensive and have several limitations as discussed in other studies [28 , 37 , 38] . Potential application to the clinic . In the current study , torso quadrants are associated with atrial anatomical locations by Table 1 . For potential use of the algorithm in the clinic , a patient specific atria-torso correlation table could be constructed if necessary . Low resolution MRI image data can provide information of the orientation of the atria in the torso; typical MRI images would be sufficient to construct such patient specific table , and allow correlation between torso quadrants and atrial anatomical sites . As the algorithm itself is generic , it could be applicable to patients without a need for individual adjustment . Limitations . The torso model lacks considerations of some other tissue types or organs ( such as muscles , fat tissue , bowels , kidneys , spleen and skin ) that may affect body surface potentials . However , the absence of those tissues does not have a big effect on the polarity of the P-waves [30 , 31] , which is the characteristic used in the present algorithm . For example , test simulations in which the conductivity of the torso was replaced by an average tissue conductivity accounting for muscle , fat and skin in various configurations demonstrates significant changes to P-wave amplitude but not to the polarity patterns ( S3 Fig . ) . The developed algorithm was based on simulation data , lacking consideration of the measurement noise as seen in real data . However , the use of P-wave polarity in the detecting algorithm can minimise the influence of noise as this may affect the amplitude of P-wave signals , but have less impact on the P-wave polarity . Whereas polarity patterns may be affected by large degrees of noise , such signals would be unsuitable for use in any clinical diagnosis . In the algorithm , eight quadrants were defined to cover the torso . The spatial resolution of the quadrant may require further refinement . For example , each quadrant can be split into eight sub-quadrants . However , finer spatial resolution of the quadrants may not help to improve the detection success rate as it decreases to 89% when eight sub-quadrants were used for each quadrant . Another potential limitation of the algorithm arises from the definition of a “biphasic” P-wave as it may lead to a miscalculation of the atrial activation site . Although the use of P-wave polarity overcomes the problems arising from the “bifid” definition as implemented in the Kistler et al . algorithm [11] , the present algorithm requires a well-defined “biphasic wave” to optimise the performance of the algorithm . This “biphasic” definition leads to a better performance of the algorithm than the use of the Kistler’s “bifid” wave . In the present study , we only tested the effectiveness of the algorithm for detecting atrial focal activity . Its use for detecting the organisation centre of rotor activity has not been performed . For that purpose , consideration of combined use of the present algorithm with vectorcardiograms [39] , phase relationships [40] and correlation analysis [41] may be necessary , warranting further investigation . Finally , the algorithm was based on simulation data . Though it provides a theoretical basis for detecting atrial focus from multi-lead ECGs , it requires further tests on real ECG data from patients or animal models with known atrial foci . Nevertheless , a test of 50 simulated atrial focus activities , different to those used to develop the algorithm , was performed , which showed a similar success rate in both male and female torso models with varying atrial position . Using a biophysically detailed computer model of the human atria-torso , we have demonstrated a correlation between atrial focal origin and polarity pattern of the BSP . Based on such correlation , a new algorithm has been developed to identify the atrial origin from the BSP reconstructed from 64-lead ECG . This study provides a theoretical basis for non-invasively detecting atrial focal origins , which is important for designing AF ablation protocol , and demonstrates the advantages of multi-lead ECG systems over the standard 12-lead ECG in detecting the origin of focal activity .
|
Ectopic activity is associated with multiple cardiac disorders and has been implicated in the initiation of self-sustaining re-entrant excitation . Identifying the presence and origin of ectopic activity may be vital in improving diagnosis and treatment of disorders such as atrial fibrillation , and has been the subject of multiple studies . The electrical activity of the heart can be non-invasively monitored through the electrocardiogram . However , the standard 12-lead electrocardiogram may not provide sufficient information to resolve the focus of ectopic activity satisfactorily and accurately; more detailed multi-lead electrocardiograms may provide more information to be able to produce an algorithm to locate the origin of ectopic activity . Using a 3D computational atria-torso model developed in our laboratory , we simulated the electrical activity of the atria under normal and different ectopic conditions . The model was first validated by comparison to experimental data , and then used to develop an algorithm to identify the location of atrial ectopic focus using a 64-lead electrocardiogram . The algorithm developed was able to identify the origin of atrial ectopic activity in 75/80 simulations , which is a significant improvement compared to previously developed algorithms . Furthermore , the study suggests that multi-lead electrocardiograms provide significant benefits over the standard 12-lead configuration .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[] |
2015
|
A New Algorithm to Diagnose Atrial Ectopic Origin from Multi Lead ECG Systems - Insights from 3D Virtual Human Atria and Torso
|
Adaptive immunity relies on the generation and maintenance of memory T cells to provide protection against repeated antigen exposure . It has been hypothesised that a self-renewing population of T cells , named stem cell–like memory T ( TSCM ) cells , are responsible for maintaining memory . However , it is not clear if the dynamics of TSCM cells in vivo are compatible with this hypothesis . To address this issue , we investigated the dynamics of TSCM cells under physiological conditions in humans in vivo using a multidisciplinary approach that combines mathematical modelling , stable isotope labelling , telomere length analysis , and cross-sectional data from vaccine recipients . We show that , unexpectedly , the average longevity of a TSCM clone is very short ( half-life < 1 year , degree of self-renewal = 430 days ) : far too short to constitute a stem cell population . However , we also find that the TSCM population is comprised of at least 2 kinetically distinct subpopulations that turn over at different rates . Whilst one subpopulation is rapidly replaced ( half-life = 5 months ) and explains the rapid average turnover of the bulk TSCM population , the half-life of the other TSCM subpopulation is approximately 9 years , consistent with the longevity of the recall response . We also show that this latter population exhibited a high degree of self-renewal , with a cell residing without dying or differentiating for 15% of our lifetime . Finally , although small , the population was not subject to excessive stochasticity . We conclude that the majority of TSCM cells are not stem cell–like but that there is a subpopulation of TSCM cells whose dynamics are compatible with their putative role in the maintenance of T cell memory .
The maintenance of long-lived T cell memory is one of the hallmarks of adaptive immunity [1 , 2] . Multiple studies have shown that the recall response to a previously encountered antigen has a half-life of the order of decades [3 , 4] . It has been hypothesised that this T cell memory is dynamically maintained by differentiation of a precursor stem cell–like memory population [5] . Alternative , nonexclusive explanations include replacement by proliferation of differentiated memory T cells or the existence of a putative subpopulation of long-lived memory T cells that has not yet been identified , either because such cells are very rare or because they reside primarily outside of the peripheral blood [6–9] . Central memory T ( TCM ) cells ( CD45RA―CCR7+ in humans ) were previously thought to constitute the stem cell–like memory precursor population . Evidence supporting the ‘stemness’ of TCM cells includes their capacity to differentiate into effector memory T ( TEM ) cells and T effector ( TEFF ) cells [10 , 11] . This hypothesis was further strengthened by cell fate–tracking experiments in mice ( using genetic barcoding and single-cell transfer ) , showing that TCM cells had the capacity to self-renew and that a single TCM cell could reconstitute immune protection against an otherwise lethal pathogen [12 , 13] . However , the concept of TCM as the stem cell population has been challenged by the identification of ‘stem cell–like’ memory T ( TSCM ) cells—which have enhanced stem cell–like properties compared to TCM cells—in mice [14] , nonhuman primates [15] , and humans [16] . In humans , like naïve cells , TSCM cells are CD45RA+CD45RO― , and they express high levels of CD27 , CD28 , interleukin 7 receptor alpha ( IL-7Rα ) , CD62L , and C-C chemokine receptor 7 ( CCR7 ) . Unlike naïve cells , TSCM cells are clonally expanded and express the memory markers CD95 and CD122 [1 , 16] . TSCM cells exhibit enhanced proliferative capacity compared to TCM cells , the potential to differentiate into all other classically defined T cell memory subsets ( including TCM ) , and the ability to retain their phenotype following proliferation both in vitro and in mice in vivo [1 , 14–16] . In light of these observations , it has been suggested that TSCM cells are the main stem cell memory population and play a key role in maintaining long-term memory in vivo [15–18] . There are 3 basic prerequisites for T cell memory stemness: multipotency , self-renewal , and clonal longevity . In this study , we focus on the related dynamic properties of self-renewal and clonal longevity . Self-renewal of human TSCM cells has been demonstrated in vitro [19] , but it remains a concern that the local microenvironment , which may crucially affect the degree of self-renewal , will be different in vivo and in vitro . However , proving self-renewal of human TSCM cells in vivo has so far not been possible because of ethical and technical limitations . The second property we investigate is clonal longevity . Long-lived T cell memory requires that memory T cell clonotypes expressing the same T cell receptor ( TCR ) persist for several decades in vivo . For example , influenza immunity has been shown to last for several decades [20] , and small pox vaccine–induced T cell memory has a half-life of 8–15 years [3 , 4] . For TSCM cells to constitute a potential precursor population for T cell memory in vivo , the survival of TSCM clones needs to be consistent with those estimates . A number of studies suggest that TSCM clones can survive for several years . Biasco and colleagues [17] observed that genetically engineered TSCM cells could persist for many years in patients suffering from severe combined immunodeficiency disease . Fuertes Marraco and colleagues [21] identified a yellow fever virus ( YFV ) -specific TSCM population up to 25 years after vaccination . Finally , in leukaemia patients who had undergone haematopoietic stem cell ( HSC ) transplantation , Oliveira and colleagues [22] reported that gene-modified TSCM cells could be detected in the circulation up to 14 years after treatment . These studies support the concept of TSCM longevity , albeit in scenarios of lymphocyte depletion or profound CD8+ T cell expansion . However , it has been shown that the dynamics of posttransplant haematopoiesis in mice differs significantly from normal , unperturbed haematopoiesis [23–25] , and so it cannot be assumed that these transplantation studies in humans necessarily recapitulate the healthy human system . In short , TSCM longevity has not been quantified in normal , unperturbed homeostasis in humans , and the related question of the ability of TSCM to self-renew has not been addressed in any human in vivo system . In order to investigate human TSCM cells in homeostasis , we previously performed stable isotope labelling of healthy volunteers and analysed label uptake in CD4+ and CD8+ , naïve T and TSCM cells . We found that the TSCM population was rapidly turning over ( median 0 . 02 per day , interquantile range 0 . 016–0 . 037 per day , half-life < 1 year ) and concluded that the TSCM population is dynamically maintained [26] . However , in this previous study , only labelling data were modelled , and so it was not possible to address the central question of the ‘stemness’ of the TSCM pool . First , to constitute a stem cell population , it is not enough to have a stably maintained population of cells; stemness requires long-term clonal persistence [18] . That is , whilst the size of the TSCM population as a whole may be stably maintained , the lifespan of any given antigen-specific precursor population could be short; such limited lifespans would be difficult to reconcile with the hypothesis that TSCM cells are the repository of T cell memory . Second , the high turnover rates obtained in this labelling study [26] do not necessarily indicate that the majority of the TSCM population is replaced by the self-renewal of the TSCM pool; frequent naïve cell differentiation could also be responsible . Indeed , given the very large size of the naïve pool compared to the TSCM pool [19] , a relatively low proportion of proliferating naïve cells would be sufficient to replace lost TSCM cells . In this scenario , TSCM cells would simply represent transit cells on the differentiation pathway from naïve to effector rather than self-renewing stem cells . Here , we investigate whether the dynamics of TSCM cells in healthy humans are consistent with their putative role as memory stem cells . Specifically , we investigate both the capacity of TSCM cells to self-renew and the longevity of TSCM clones . It is challenging to address these questions in humans , and they cannot be answered using stable isotope labelling alone , since different scenarios ( e . g . , ‘all new TSCM cells come from naïve cell differentiation’ versus ‘all new TSCM cells come from TSCM proliferation’ ) can give rise to very similar levels of label in the TSCM population . To enable us to deconvolute these possibilities , we performed telomere length analysis and utilised cross-sectional TSCM cell data from YFV vaccine recipients . Deterministic and stochastic mechanistic mathematical modelling were then used to analyse all 3 datasets . This novel approach allows us to investigate human TSCM cell dynamics in vivo and to address questions previously only investigated in animal models .
Estimates of the rate of TSCM renewal and TSCM clonal lifespan will depend upon the kinetic structure of the TSCM pool . We therefore first asked whether there was evidence for kinetic heterogeneity ( i . e . , existence of subpopulations with differing kinetics ) within the TSCM pool by comparing the quality of fit of a homogenous and heterogeneous version of the mathematical model . In the homogeneous version of the model , we constrain the input rate ( proliferation rate + rate of new entrants due to differentiation from naïve cells ) of the whole TSCM population to be equal to the disappearance rate of labelled TSCM cells; this condition will be met for a kinetically homogenous population of constant size . In the heterogeneous version of the model , this constraint was relaxed to allow for the possibility of kinetic heterogeneity in the TSCM pool ( Methods , [28] ) . We used this implicit description of kinetic heterogeneity rather than an explicit description of the subpopulations because it requires fewer parameters ( 2 compared with 3 for the explicit model [28–30] ) and furthermore does not suffer from the parameter identifiability issue inherent in the explicit kinetic heterogeneity model , which arises due to the very strong correlation between the proliferation rate and size of a subpopulation [31] . A total of 9 datasets were included in this analysis , representing CD4+ and CD8+ T cells ( naïve and TSCM ) from 5 individuals ( 1 CD8+ TSCM cell dataset from 1 subject was not available ) . We found that , in 7 out of the 9 cases , constraining the TSCM population to be homogeneous resulted in a substantially worse description of the data ( Fig 3 ) ; and there was strong evidence to reject the assumption of homogeneity P = 5 . 8 × 10−7 , P = 4 . 1 × 10−6 ( median of p-values calculated using Fisher’s F-test for nested models between the homogeneous and heterogeneous models for CD4+ and CD8+ TSCM , respectively ) , indicating considerable support for the heterogeneous description of the TSCM pool in both CD4+ and CD8+ T cell populations ( S1 Table ) . In contrast , there was no evidence to reject the null hypothesis of homogeneity in the naïve cell pool ( P = 0 . 6 , P = 0 . 5 , for CD4+ and CD8+ TN , respectively; median of p-values calculated using Fisher’s F-test for nested models between the homogeneous and heterogeneous models; S1 Table ) . The size of a newly generated TSCM clone will be an important determinant of clonal longevity , as this determines not just the initial magnitude of a new clone but also the rate at which an existing clone is displaced by new entrants bearing different TCRs . Unfortunately , the size of the clonal expansion accompanying the differentiation of naïve to TSCM cells ( k in the model; Fig 2 , Methods ) was not identifiable . Different fitting runs to the same dataset ( with different initial conditions or different random seeds ) gave different estimates of the clonal expansion parameter k . Consistent with this , we found that if k was fixed to different constant values in the range 0–20 , then , with the exception of 1 individual for which the sum of squares increases dramatically for k above 15 , the sum of squares remained constant in every case for all values of k ( Fig 4A and Fig 4E ) . Henceforth , we systematically repeat all analyses for multiple values of k in the range 0–20 to ensure that results are robust despite uncertainty in the clonal expansion parameter . Values of k above 20 were not considered biologically plausible [13 , 32 , 33] . In summary , we were not able to estimate the size of the clonal expansion k upon differentiation of naïve cells to TSCM cells and instead utilised a strategy to investigate TSCM dynamics despite uncertainty in this parameter . Next , we quantified TSCM clonal longevity . We fitted the mathematical model ( implicit heterogeneous version ) simultaneously to the telomere length and isotope labelling data , with k fixed sequentially at different values in the range 0–20 . We found that , with the exception of 1 dataset ( CD8+ T cells in DW01 ) , the contribution of naïve cells to TSCM replacement was never less than 20% and could be as much as 90% ( Fig 4B and 4F ) . Correspondingly , the average half-life of a TSCM clone was short: the maximum ever observed ( across all values of k and across all individuals ) was 4 years , but typically , it was much shorter and in the range 0–500 days ( Fig 4C and 4G ) . These conclusions about short clonal longevity were robust to assumptions regarding the activity of telomerase . Specifically , for all values of telomerase compensation considered in the range 0–k , the estimated average clonal half-lives were never higher than those estimates reported above ( in which compensation was a free parameter ) . Importantly , this half-life represents the duration of memory to an antigen ( Methods ) , not the conventional population half-life . It is possible that a very small number of surviving TSCM cells is sufficient to generate a substantial recall response , and so a short clonal half-life is not necessarily incompatible with long-lived recall responses . Moreover , extinction of some clones specific for a given antigen is not necessarily problematic if other clones ( bearing different TCRs ) specific for the same antigen survive . To assess this possibility , we used the exact Gillespie algorithm ( Methods ) to quantify the time for the last cell of an antigen-specific precursor population to disappear . This is reported as the precursor lifespan in Fig 4D and 4H . Whilst this did lead to a considerable increase in longevity ( stochastic estimates of total antigen-specific precursor lifespan were typically 3 times longer than the deterministic clonal half-life ) , maximum estimates were still only of the order of 2 , 000 days ( about 5 years ) for most individuals . In summary , although individual parameters were poorly identifiable , all parameter combinations able to describe the experimental data were associated with average clonal half-life estimates , which were much lower than the 8–15 year half-life of the recall response [3 , 4] . Even total precursor lifespans ( times until the last cell of an antigen-specific precursor population disappears ) were lower than those values in most cases . We conclude that the average TSCM population is replaced too rapidly for it to be the stem cell population responsible for maintaining memory . The model used up to this point allows for heterogeneity but nevertheless reports population averages ( i . e . , the proliferation rate and clonal half-life averaged across the whole TSCM population ) . This averaging could be hiding a small , long-lived population within the bulk short-lived population . Expanding our model , which deals with heterogeneity implicitly ( and thus averages across the population ) , to one that deals with heterogeneity explicitly ( and thus provides estimates for the half-lives of all subpopulations ) is problematic , as even the simplest version of the explicit heterogeneity model suffers from severe identifiability issues [30] and fails to deliver the parameters of interest when fitted to labelling data . We confirmed that , for our more complex system with both naïve and TSCM cells , an explicit description of heterogeneity provided no information . To address this problem , we therefore sought an alternative class of data . We analysed published data of the vaccine-induced YFV-specific TSCM response in humans from Fuertes Marraco and colleagues [21] . In brief , the magnitude of the CD8+ TSCM cell response to the HLA-A*02-restricted YFV nonstructural protein 4b ( NS4b214−222 ) epitope was measured by HLA class I tetramer at different time points ( range 0 . 27–35 . 02 years ) postvaccination in a cross-sectional study of 37 recipients of the YF-17D YFV vaccine . We fitted the explicit heterogeneity version of the naïve T ( “TN” ) and TSCM model to all 3 types of CD8+ T cell data ( isotope labelling , telomere length , and YFV ) simultaneously ( Supplementary Methods in S1 Text ) . The fits are shown in Fig 5 . As for the implicit heterogeneity model , the fraction of new TSCM cells originating from naïve cells was high ( minimum 10% , median 44% ) . We found evidence for at least 2 subpopulations of CD8+ TSCM cells ( designated TSCM1 and TSCM2 for the purposes of this discussion ) . The majority of the TSCM cells generated upon clonal expansion of naïve cells differentiated into the TSCM1 subpopulation , characterised by a short half-life ( ≤1 year ) and a high replacement rate ( median 0 . 02 per day , interquantile range 0 . 024–0 . 045 per day ) , slightly higher than the average rates estimated by the previous model . The remaining fraction of the generated clone was observed to enter a long-lived subpopulation with a median half-life of 9 years ( Table 1 , S2 Table ) . Surprisingly , although the fraction of naïve cells entering the TSCM2 pool was low , because of its low death/differentiation rate , the number of long-lived TSCM2 cells in the circulation at any given time could be as high as , or even higher than , the number of rapidly proliferating TSCM1 cells . Results are summarised schematically in Fig 5C . Four other weighting strategies ( of the different types of data ) yielded the same conclusions in all cases ( S1 Fig , S3 Table ) . The long-lived TSCM subpopulation identified in the previous section ( Subpopulation kinetics ) is a potential candidate for the stem cell population responsible for the maintenance of immune memory . We therefore investigated the degree of self-renewal and clonal stability within this long-lived TSCM compartment . The degree of self-renewal of a population at steady state , 1 / ( death rate + differentiation rate–proliferation rate ) , quantifies the upstream input necessary to maintain a population . If there is a large upstream contribution , then the degree of self-renewal will be low [34] . A perfectly self-renewing population ( e . g . , HSCs ) will have an infinite degree of self-renewal . We quantified the degree of self-renewal for the long-lived population and found a median of 4 , 600 days , with the range 2 , 400–7 , 300 days ( S2 Table ) . This implies that , on average , a TSCM cell ( or its progeny ) from the long-lived subpopulation resides in the TSCM compartment without dying or differentiating for 4 , 600 days . The total CD8+ TSCM population is small ( 2%–3% of circulating CD8+ lymphocytes [19] , 1%–5% of lymph node–resident CD8+ T cells [35] ) . If only a proportion of this already small population is responsible for maintaining memory , then this raises the issue that , although the precursor population specific for a given antigen may have a long half-life , the small size of that population could mean that its dynamics are highly stochastic . That is , there may be wide ranges in the length of memory , and some antigen-specific precursor populations would be predicted to be lost by stochastic extinction soon after generation—i . e . , memory would be erratic and fallible . To investigate the stochasticity of the length of memory within the long-lived TSCM pool , we performed Gillespie simulations of the size of the antigen-specific precursor population based on the parameter estimates derived from model fitting for each of the 4 individuals with CD8+ T cell data . Although there was stochasticity in the half-lives of antigen-specific precursors across different runs , the variation was not large ( Fig 6A ) , and the different trajectories were tightly clustered ( Fig 6B ) . Concerned that the YFV vaccine , which is known to generate an exceptional CD8+ T cell response , may not be representative of a typical antigen , we also sought to study TSCM dynamics independent of the YFV dataset . Guided by the concept that there may be a long-lived TSCM subpopulation , we fitted the explicit kinetic heterogeneity model to the isotope labelling and telomere length data , ignoring the YFV data but imposing a half-life greater than 5 years on the long-lived TSCM subpopulation . As expected , parameters could no longer be reliably identified , but we were able to conclude that the dynamics of CD8+ TSCM cells were compatible with long subpopulation half-lives between 5 and 15 years ( S2 Fig ) . This approach also allowed us to study CD4+ TSCM cells ( which were not measured in the YFV study ) . Again , we found that the dynamics of CD4+ TSCM cells were compatible with the presence of a similarly long-lived subpopulation with a half-life between 5 and 15 years ( S3 Fig ) .
One leading explanation for the maintenance of long-term immunological memory is the existence of a stem cell–like population of memory T cells , able to both self-renew and to differentiate into all other subsets of the T cell memory pool [12 , 17 , 21] . It has been suggested that the recently discovered TSCM population is the main stem cell population responsible for maintaining T cell memory [18 , 36] . In a previous study , we used stable isotope labelling to investigate TSCM dynamics at equilibrium in healthy subjects [26] . This revealed unexpectedly high rates of turnover in the CD4+ and CD8+ TSCM compartments . Whilst supporting the concept that the TSCM population as a whole is stable , it did not establish whether new TSCM cells were generated from naïve cells or by TSCM proliferation . Moreover , the key parameters of clonal longevity and self-renewal , which are prerequisites for stemness , could not be deconvoluted . Given that TSCM cells were found to die and be replaced rapidly , the source of the replacing cell becomes critical . Even a small contribution from naïve cells can result in a weakly self-renewing population and loss of memory as existing clones are replaced by cells specific for a different antigen . In the present study , we overcame these limitations by simultaneously fitting telomere length and tetramer data to constrain the space of possible models and improve parameter identifiability . We report that firstly , the TSCM population is kinetically heterogeneous , with at least 2 kinetically distinct subpopulations turning over at different rates; and secondly , the dynamics of a subpopulation of TSCM cells are compatible with their hypothesised role as the main stem cell–like T cell precursor responsible for the maintenance of T cell memory . The best description of the data is one in which the kinetically heterogeneous TSCM population , despite its high average replenishment rate of approximately 2% per day , contained a fraction of long-lived TSCM cells . The half-life of this slower subpopulation was approximately 9 years , consistent with the 8–15 year half-life estimated for the recall response to a given antigen [3 , 4] . Furthermore , although this subpopulation was small , the dynamic behaviour of individual clones was not excessively stochastic , and the half-life of a given antigen-specific precursor population was tightly distributed . Finally , we estimated that the degree of self-renewal of the long-lived TSCM subpopulation was approximately 4 , 600 days . The quantification of the dynamics of TSCM cells is not easily addressed in humans , and different studies invariably involve different compromises . An advantage of our study is that it examines the natural dynamics of TSCM cells in healthy , lymphocyte-replete individuals . A disadvantage is that many of the model parameters were poorly identifiable; nevertheless , firm conclusions about clonal longevity and self-renewal can be drawn . A second disadvantage is that it utilises vaccination data generated using a vaccine ( the YFV-17D vaccine ) known for its ability to generate an exceptional CD8+ T cell response and which may not be representative of average immunity . To address this potential caveat , we repeated our analysis without reference to the YFV vaccination data . Whilst this reduced our ability to estimate individual parameters , it did confirm the conclusion that the TSCM population consists of subpopulations with different kinetics and that high turnover and short half-lives of the bulk TSCM population do not rule out the existence of a slowly turning over subpopulation with the dynamic properties required for TSCM cells to maintain both CD4+ and CD8+ T cell memory . Finally , it should be noted that our study was confined to circulating TSCM cells . In mice , it has been shown that memory T cell dynamics can vary across different anatomical compartments ( spleen versus bone marrow ) [37] . For human studies , ethical and technical considerations mean that repeated sampling of cells must be limited to the peripheral blood compartment . However , this does enable direct comparison with previously reported data , including the seminal description of human TSCM cells [16] , which was based on circulating cells . Our findings of kinetic heterogeneity in the human TSCM population are reminiscent of the proliferative heterogeneity described for transplanted HSCs in lethally irradiated mice [38] , in which levels of Kit receptor distinguished cell subpopulations with different expansion capacities . Similarly , studies of human CD4+ and CD8+ TSCM cells cultured in the presence of cytokines ( either interluekin 15 [IL-15] or interluekin 7 [IL-7] + IL-15 ) in vitro have also reported that a fraction of cells proliferated rapidly , while the majority remained quiescent [39] [19] . Finally , in rhesus macaques , circulating TSCM cells had a level of Ki-67 expression that the author remarked was ‘unexpectedly large’ ( mean > 10% ) [36]; this is consistent with our finding that , on average , the TSCM population in peripheral blood turns over rapidly . Our proliferation rate estimates for naïve and TSCM cells can be compared with proliferation rate estimates of other T cell subsets obtained using stable isotope labelling . We found that the proliferation of naïve T cells is slowest , p = 0 . 0005d−1 , and comparable with previous estimates [40] , though in this latter study , the gating strategy would have inadvertently included TSCM cells in the naïve cell gate . Next is the slow TSCM subpopulation with a proliferation rate an order of magnitude faster than naïve cells ( p = 0 . 002d−1 ) , then the fast TSCM subpopulation with a median proliferation rate of p = 0 . 015d−1 comparable with that of memory cells ( 0 . 006d−1–0 . 02d−1 [29 , 41] ) . Giving the following rank order of proliferation rates: naïve < slow TSCM < fast TSCM ≲ memory . Our estimate of the degree of self-renewal is more difficult to place in context . To the best of our knowledge , this parameter has only been quantified previously for murine HSCs . Our estimate of 4 , 600 days for the degree of self-renewal of the long-lived TSCM subpopulation is naturally less than the corresponding estimate for murine HSC—which is , by definition , infinite—but is greater than the degree of self-renewal of compartments immediately adjacent to HSCs in the differentiation pathway—namely , short-term HSCs ( degree of self-renewal of 90–150 days in mice ) and multipotent progenitors ( MPPs; degree of self-renewal of 7–28 days in mice ) [23 , 34] . If we convert to ‘animal lifespans’ ( 80 years for human , 2 years for mice; a scaling which appears valid for T cell kinetics [41] ) , then we see that the degree of self-renewal of TSCM cells is 0 . 15 lives; i . e . , a long-lived TSCM cell resides without dying or differentiating for approximately 15% of our lifespan . This is similar to the degree of self-renewal of short-term HSCs ( 0 . 12–0 . 2 lives ) and greater than the degree of self-renewal of MPPs ( 0 . 01–0 . 04 lives ) . It is remarkable that a peripheral cell population that is towards the end of the haematopoietic differentiation pathway should have a degree of self-renewal that is comparable with short-term HSCs . Unexpectedly , we found strong evidence for continual differentiation of naïve T cells into the TSCM cell pool despite the study volunteers being healthy with no symptomatic infection . For both the implicit and the explicit heterogeneity models , the contribution of naïve cells to TSCM replacement was typically about 50% and never less than 10% ( Fig 4B and 4F ) . This may represent differentiation of naïve cells in response to continual low-level exposure to novel environmental antigen and/or to persistent antigen . Considerable recruitment of naïve cells to the memory pool in the apparent absence of novel antigen has been previously described for mice persistently infected with Polyoma virus [42] or lymphocytic choriomeningitis virus [42] and for healthy mice [43] . The role of the short-lived TSCM subpopulation that we identify is unclear . Potentially , it is activated naïve cells rapidly transitioning to effectors whilst others are retained to form the long-lived TSCM pool that is the basis of memory . This is consistent with recent evidence that TSCM cells may pass through a phase in which they express effector molecules [44] . This work suggests a number of future directions . One important direction is to establish phenotypic markers to distinguish the ‘true’ TSCM subpopulation . A second is to develop a model to predict the TCR repertoire of the true TSCM subpopulation and whether this differs from bulk TSCM cells and the functional consequences of any such difference . Finally , it is important to know whether or not murine TSCM populations are similarly heterogeneous , since this would facilitate a whole range of experiments not possible in humans . Our results show that substantial kinetic heterogeneity exists within the TSCM pool , encompassing a long-lived subpopulation with the dynamic properties required to maintain both CD4+ and CD8+ T cell memory . Further characterisation of these bona fide TSCM cells may illuminate the mechanistic basis of durable immune protection and facilitate translational efforts to develop more effective vaccines and immunotherapies .
The fit of the homogenous and heterogeneous model was compared using Fisher’s F-test for nested models . This test compares the goodness of fit of nested models to data , taking into account the different number of parameters in the models [62] .
|
The human immune system remembers previously encountered pathogens so that , on meeting the same pathogen a second time , the response is quicker and more effective . This immune memory is the basis of all vaccinations . Immune memory persists for decades , but how memory is maintained is unclear . It has been hypothesised that there is a dedicated population of cells called stem cell–like memory T ( TSCM ) cells that have stem cell–like behaviour and are responsible for the persistence of T cell memory . Here , we show that a subset of TSCM cells , in healthy humans in vivo , have the dynamic properties of self-renewal and clonal longevity necessary to maintain long-lived immune memory .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"blood",
"cells",
"medicine",
"and",
"health",
"sciences",
"immune",
"cells",
"chromosome",
"structure",
"and",
"function",
"cell",
"cycle",
"and",
"cell",
"division",
"immunology",
"cell",
"processes",
"precursor",
"cells",
"cloning",
"cell",
"differentiation",
"telomeres",
"developmental",
"biology",
"cytotoxic",
"t",
"cells",
"molecular",
"biology",
"techniques",
"research",
"and",
"analysis",
"methods",
"telomere",
"length",
"white",
"blood",
"cells",
"memory",
"t",
"cells",
"animal",
"cells",
"chromosome",
"biology",
"t",
"cells",
"molecular",
"biology",
"cell",
"biology",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"chromosomes"
] |
2018
|
Human TSCM cell dynamics in vivo are compatible with long-lived immunological memory and stemness
|
Cleft lip with or without cleft palate ( CL/P ) is the most commonly occurring craniofacial birth defect . We provide insight into the genetic etiology of this birth defect by performing genome-wide association studies in two species: dogs and humans . In the dog , a genome-wide association study of 7 CL/P cases and 112 controls from the Nova Scotia Duck Tolling Retriever ( NSDTR ) breed identified a significantly associated region on canine chromosome 27 ( unadjusted p=1 . 1 x 10-13; adjusted p= 2 . 2 x 10-3 ) . Further analysis in NSDTR families and additional full sibling cases identified a 1 . 44 Mb homozygous haplotype ( chromosome 27: 9 . 29 – 10 . 73 Mb ) segregating with a more complex phenotype of cleft lip , cleft palate , and syndactyly ( CLPS ) in 13 cases . Whole-genome sequencing of 3 CLPS cases and 4 controls at 15X coverage led to the discovery of a frameshift mutation within ADAMTS20 ( c . 1360_1361delAA ( p . Lys453Ilefs*3 ) ) , which segregated concordant with the phenotype . In a parallel study in humans , a family-based association analysis ( DFAM ) of 125 CL/P cases , 420 unaffected relatives , and 392 controls from a Guatemalan cohort , identified a suggestive association ( rs10785430; p =2 . 67 x 10-6 ) with the same gene , ADAMTS20 . Sequencing of cases from the Guatemalan cohort was unable to identify a causative mutation within the coding region of ADAMTS20 , but four coding variants were found in additional cases of CL/P . In summary , this study provides genetic evidence for a role of ADAMTS20 in CL/P development in dogs and as a candidate gene for CL/P development in humans .
Nonsyndromic orofacial clefts , notably cleft lip ( CL ) with or without cleft palate ( CL/P ) and isolated cleft palate ( CP ) , are the most common craniofacial birth defects in humans and represent a substantial personal and societal burden . Clefts affect approximately 1 in 700 individuals[1] , with a lifetime cost of treatment in the U . S . A . estimated at $200 , 000[2–5] . Rates vary dramatically depending on population , with higher rates of CL/P found in Asians , South Americans , and American Indians compared with Caucasians [6] , while populations of African descent are the least often affected[7] . Interestingly , the CL/P birth prevalence differs between genders ( males more often affected ) , but the CP prevalence does not[6] . Although clefts can be surgically repaired , affected individuals often undergo multiple craniofacial and dental surgeries , as well as speech , hearing , and orthodontic therapies . Furthermore , individuals born with an orofacial cleft have increased incidence of mental health problems , higher mortality rates through all stages of life[2 , 8] , and a higher risk of various cancer types ( including breast , brain , and colon cancers ) that extends to their family members[9–12] . Orofacial clefts are complex birth defects resulting from genetic variations , environmental exposures , and their interactions[3] . Before the advent of genome-wide approaches , evaluation of candidate genes revealed at best modest associations with a number of genes[13 , 14]; further , despite exhaustive mutation screens , coding mutations were found in less than 10% of study participants , leaving a large portion of the genetic etiology unexplained[15–18] . Genome-wide studies , using both linkage and association methods[18] , have identified a number of different genes and genomic regions likely to contribute to the risk of orofacial clefts; both rare and common variants have been implicated in studies of Caucasian and Asian populations[16–18] with distinct SNPs associated in each ethnicity . In addition to human genetic studies , a variety of model organisms have been utilized to further understand the development of orofacial clefts . A number of mammalian species exhibit orofacial clefting , with mice being the most often studied; however , mouse models often exhibit CP alone and very rarely display CL/P . In contrast , the domestic dog has spontaneous and naturally occurring CL/P with a recessive mode of inheritance documented in several breeds[19–21] . Contributing further to their usefulness as a model organism , common breeding practices have created genetically isolated populations ( dog breeds ) that result in few haplotypes and extensive linkage disequilibrium[22 , 23] . Even within closed breeding populations , genetic heterogeneity of orofacial clefts is present within the Nova Scotia Duck Tolling Retriever ( NSDTR ) dog breed . We previously identified a DLX6 LINE-1 insertion as a cause of cleft palate and mandibular abnormalities in a subset of NSDTRs with orofacial clefts[24] . The DLX6 LINE-1 insertion failed to explain a series of cases with cleft lip , indicating molecular and phenotypic heterogeneity of orofacial clefting within the breed . Here , we present two parallel genome-wide association studies ( GWAS ) in the domestic dog and humans . Both studies provide evidence for a role of the same gene , ADAMTS20 , in CL/P development . A GWAS investigating the cause of cleft lip within a small cohort of NSDTRs identified an associated interval on canine chromosome 27 that segregates with a more complex phenotype of CL/P and syndactyly . Whole-genome sequencing of these dogs led to the identification of a frameshift mutation within ADAMTS20 . A GWAS within a cohort of native Guatemalans with CL/P identified a suggestive association with an interval encompassing ADAMTS20 . Sanger sequencing of ADAMTS20 within CL/P cases from the native Guatemalan cohort and additional cases , identified four novel risk variants for CL/P in humans .
To identify loci associated with CL/P in dogs , an allelic genome-wide association was performed using 7 CL/P case dogs and 112 control dogs from within the NSDTR breed . After quality control , association analysis of 110 , 021 remaining SNPs identified a highly associated region on canine chromosome 27 ( unadjusted p-value of 1 . 1 x 10-13 , CFA27: 11419150 ) ( Fig . 1A ) . A genomic inflation factor ( λ ) of 1 . 18 was observed , however , the significance of the positive association and lack of any other associated chromosomal regions led to the investigation of the associated region without further correction for population stratification . After 100 , 000 permutations to correct for multiple tests , the adjusted empirical p-value was 2 . 2 x 10-3 with 20 SNPs reaching genome-wide significance ( i . e . , adjusted p-value ≤ 0 . 05; Fig . 1B ) . Underlying this evidence of association was a 2 . 88 Mb homozygous haplotype spanning from CFA27: 9 . 29–12 . 16 Mb in 6 cases used in the GWAS when compared to the controls ( Fig . 1C ) . This homozygous haplotype was not identified in one of the cases with CL/P nor in any of the 112 controls . Homozygosity mapping of the associated interval was performed in the six full sibling cases not included in the GWAS . These six full sibling cases were excluded from the GWAS analysis in order to reduce possible population stratification . All six cases were found to be homozygous throughout the associated region . Previous studies investigating cleft palate within the NSDTR breed identified two unexplained cases of cleft palate[24] . Homozygosity mapping was also performed in these two unexplained cases and the homozygous haplotype was identified in one of the two CP NSDTRs . Closer inspection of the CP NSDTR with the homozygous haplotype revealed bilateral clefts of the lower alar nasal folds . In summary , a homozygous region concordant with the CFA27-associated interval was identified in 13 of 15 total NSDTRs with orofacial clefts ( Table 1 ) . Recombination breakpoints in the 13 dogs reduced the critical interval from 2 . 88 Mb to 1 . 44 Mb ( CFA27: 9 . 29–10 . 73 Mb; Fig . 1D ) . Segregation analysis of the associated haplotype was performed in unaffected family members with enough available DNA ( parents n = 6; littermates n = 6 ) . None of the 12 unaffected family members were homozygous throughout the associated interval and the parents were all heterozygous , suggesting a recessive mode of inheritance . The phenotypes observed in the 13 dogs with the homozygous haplotype on CFA27 include a range of clefting phenotypes including bilateral clefts of the lower alar nasal folds ( n = 1 ) , bilateral clefts of the lower alar nasal folds and CP ( n = 6 ) , a right unilateral complete CL and CP ( n = 1 ) , and bilateral complete CL and CP ( n = 5 ) ( Fig . 2A-C ) . To identify additional craniofacial defects , micro computed tomography ( microCT ) analysis was performed on two severely affected NSDTR cases with bilateral complete CL and CP and three NSDTR controls . MicroCT findings were consistent with the presence of bone and soft tissue clefts of the primary and the secondary palate ( Fig . 2D-G ) . In both affected individuals , bilateral dentoalveolar clefting was evident between the incisors and canine teeth . This clefting was associated with a complete absence of ossification of the dorsal and lateral aspects of the premaxilla , which would normally form a suture with the maxillae and rostral aspect of the nasal bones . The ventromedial component of the premaxilla , from which the incisors arise , appeared largely normal . However , in contrast , to the controls in which the six incisors are evident in the premaxilla , the affected individuals presented with only four incisors in the remnant premaxillary bone . The lateral-most incisors were not integrated in bone and were relatively poorly formed and abnormally oriented in the cleft individuals . Interestingly , the nasal bones were normal in appearance and length . Compared to controls , the rostral portion of each maxillae was hypoplastic ( S1 Fig and S2 Fig . ) . In one of the two individuals , a cleft of the secondary palate extending the full length of maxillary and palatine bones was observed , resulting in free communication between the nasal cavity and oral cavity . In the second individual , one palatal shelf had extended to the midline and had approximated with the anterior nasal septum while the other showed little lateral outgrowth , although the palatal rugae were still evident on both shelves . This asymmetry in shelf growth was mirrored by the asymmetric palatal bone growth . In contrast to the palate , the basisphenoid and basioccipital bones of the cranial base were similar in appearance to that of controls . Of note , however , both affected individuals had smaller ( average ~20% smaller relative to skull length ) and more laterally positioned tympanic rings compared to the unaffected individuals ( Fig . 2FG ) . In one individual , the tympanic rings were also slightly dimorphic . In one individual there was notable asymmetry in length of one of the component hyoid bones . The mandible , although largely normal , had a slightly more narrow appearance and slightly delayed ossification around the mandibular canines and incisors . In both cleft individuals the forehead was slightly more tapered . In addition to CL/P , simple complete and/or partial syndactyly of the third and fourth digits was observed in 10 of the 13 dogs ( Fig . 2H-J ) . It is unknown whether or not the 3 remaining dogs had syndactyly . Simple syndactyly demonstrating only soft tissue involvement was observed on radiographs of a paw with complete syndactyly ( Fig . 2K ) . We designate this phenotype cleft lip , palate , and syndactyly ( CLPS ) . The observed phenotypic spectrum of CLPS cases is summarized in Table 2 . To identify variants within the critical interval , whole-genome sequencing was performed on three CLPS NSDTR cases and four control NSDTRs that were homozygous wild type throughout the associated interval and unaffected with CLPS . Within the associated interval ( CFA27: 9 . 29–10 . 73 Mb ) there were 14 , 167 SNP and in/del variants when compared to the CanFam 3 . 1 boxer reference genome [22] . Based on the homozygous haplotype observed in the cases , we hypothesized a recessive mode of inheritance and excluded variants that did not segregate with the phenotype . From the set of homozygous variants in the CLPS cases , we excluded variants that were also homozygous in at least one control dog . This reduced the number of variants segregating with the phenotype to 142 ( Table 3 ) , when compared to a reference set of 26 control dogs . Breeds of all dogs with whole genome sequence are summarized in S1 Table . Of these 142 variants , only two were predicted to affect the coding region of genes within the critical interval ( S2 Table ) [25] . One synonymous coding variant ( PUS7L c . 278A>G ( p . ( = ) ) ) was identified as homozygous in cases and heterozygous in 10 control dogs ( 1 NSDTR , 5 Dalmatians , 1 Kelpie , 1 Bearded Collie , 2 Weimaraners ) . Due to the severe nature of the CLPS phenotype , we hypothesized that the causative allele would occur at a low frequency across all dog breeds . Because approximately one third of the randomly sampled control dogs were heterozygous for PUS7L c . 278A>G ( p . ( = ) ) , we concluded that this variant was unlikely to be causal for the CLPS phenotype . The second variant , ADAMTS20 c . 1360_1361delAA ( p . Lys453Ilefs *3 ) , was predicted to be a frameshift mutation in the metalloprotease domain resulting in premature truncation of 1461 amino acids from the 1916 amino acid protein ( Fig . 3AB ) [26] . It was not observed in any of the 30 control dogs . To confirm the ADAMTS20 c . 1360_1361delAA ( p . Lys453Ilefs*3 ) frameshift mutation , Sanger sequencing was performed in cDNA from a CLPS NSDTR case and embryo control ( Fig . 3A ) . In addition to the deletion , seven SNPs were identified in the CLPS NSDTR case and embryo control when compared to CanFam 3 . 1 Boxer reference sequence ( Table 4 ) [22] . Six of the 7 SNPs did not segregate with the phenotype . ADAMTS20 c . 682C >T ( p . Val228Leu ) segregated within the CLPS NSDTR case , embryo control , and reference sequence , but is a known SNP previously identified in wolves and 15 other dog breeds[27] . Further investigation of this SNP in the 8 NSDTRs with available whole-genome sequence also confirmed that this SNP did not segregate with the phenotype: 1 NSDTR was homozygous for the boxer reference allele , 3 NSDTRs were heterozygous , and 3 NSDTRs were homozygous for the alternate allele ( 3 cases , 1 control ) . We hypothesize that the predicted premature stop codon in ADAMTS20 c . 1360_1361delAA ( p . Lys453Ilefs*3 ) results in decreased expression levels when compared to wild type . Quantitative real time PCR analysis was performed on cDNA from heart tissue of CLPS NSDTRs and compared to cDNA from heart tissue of unaffected NSDTRs that were homozygous wild type for the deletion . REST analysis indicated that the ADAMTS20 transcript in CLPS NSDTRs was significantly down regulated by a mean factor of 0 . 549 ( p = 0 . 005 ) in CLPS cases when compared to controls[28] . Genotyping was performed in available parents ( n = 8 ) , littermates ( n = 13 ) , and all NSDTR cases ( n = 13 CLPS; 2 unexplained ) ( Fig . 4 ) . All 13 CLPS cases were homozygous for the deletion . The two unexplained NSDTRs cases without the associated haplotype were homozygous for the wild type allele . We genotyped 97 unrelated control NSDTRs and identified this deletion in the heterozygous form in 3 control dogs . To determine if this allele is found in other breeds , we genotyped 53 dogs with orofacial clefts from 25 breeds and 288 unaffected dogs from more than 70 breeds ( S3 Table ) . All genotyped individuals were homozygous for the wild type allele , suggesting this ADAMTS20 deletion is private to the NSDTR breed . Our previous work in dogs identified a DLX6 LINE-1 insertion responsible for cleft palate and mandibular abnormalities in a subset of NSDTRs [24] . The ADAMTS20 c . 1360_1361delAA ( p . Lys453Ilefs*3 ) deletion was also genotyped in these cases ( n = 19 ) . Two of the cases were heterozygous , while the remaining 17 dogs were homozygous wild type for the ADAMTS20 c . 1360_1361delAA ( p . Lys453Ilefs*3 ) mutation . Of all control NSDTRs , including relatives , genotyped for both mutations ( n = 99 ) , five were heterozygous for both mutations and were phenotypically normal . To investigate the cause of CL/P in humans , a DFAM allelic association[29] analysis was performed in a Guatemalan study population of 125 nonsyndromic CL/P cases , 420 unaffected relatives ( 545 total from case families ) , plus 392 controls with no family history of orofacial clefts ( Fig . 5A ) . No p-value reached a conservative Bonferroni corrected p-value of less than 1 . 1 x 10-7 ( alpha = 0 . 05 ) , but several SNPs had p-values suggestive of association ( p-values ~10-6 ) . Table 5 shows the top 5 CL/P associated SNPs according to the DFAM analysis within the Guatemalan cohort , as well as the corresponding TDT , sib-TDT , and logistic regression p-values . S4 Table lists all SNPs with DFAM p-value less than 0 . 0001 . The QQ plot of the allelic GWAS results shows no evidence of genomic inflation in this family based study ( S3A Fig . ) . The strongest association observed was for rs10785430 ( DFAM p-value = 2 . 69 x 10-6 ) on chromosome 12 , which mapped to an intron in ADAMTS20 . This SNP also showed significant associations ( p<0 . 05 ) using the TDT ( p = 1 . 8 x 10-4 ) , sib-TDT ( p = 3 . 8 x 10-4 ) , and case-control ( p = 0 . 007 ) analyses . To further explore this region , we imputed unobserved SNPs on chromosome 12 using the 1000 Genomes reference sample ( http://browser . 1000genomes . org ) . Based on SNPs in high LD with rs10785430 , association tests revealed a 157kb region of association ( from rs11182055 to rs9988939 ) yielding p-values less than 1x10-4 ( Fig . 5B ) . We next performed a gene-level analysis with DFAM on 17 , 578 genes . The QQ plot of gene p-values revealed no significant deviations from expected ( S3B Fig . ) indicating that the VEGAS method adequately controlled for LD ( shown for ADAMTS20 in Fig . 5C ) . The top 5 associated genes derived from the Guatemalan GWAS using the VEGAS method are summarized in Table 6 . ADAMTS20 had the lowest gene-wise p-value ( p = 5 . 3x 10-5 ) , in agreement with the single SNP results . Previous genome-wide studies were performed in Caucasian and Asian trios[17 , 18] , but did not identify ADAMTS20 as a top hit . However , several SNPs within ADAMTS20 did show nominal significance ( p<0 . 05 ) in Caucasians[17] , so we performed a meta-analysis of the Guatemalan and Caucasian results for the ADAMTS20 SNPs . Overall , smaller p-values were observed for some SNPs after the addition of the Caucasian results , but not for the most significant SNP in the Guatemalans ( S4 Fig . ) . This suggests a distinct genetic etiology for CL/P formation in Guatemalans . We sequenced all protein coding exons of ADAMTS20 in 20 Guatemalan CL/P cases to determine if a novel , population-specific common variant could explain the association with markers in ADAMTS20; however , no such variants were identified . We also sequenced 19 cases from the Philippines to explore the possibility that rare variants in ADAMTS20 could confer risk of CL/P in other populations . Similarly , we looked for rare coding variants in our Guatemalan cases that could contribute to CL/P risk independent of the association with the common SNP rs10785430 . No private variants were found in the Guatemalan cases , but three novel missense variants were found among Filipino cases ( S5AB Fig . ) . All three novel Filipino variants were inherited from unaffected parents . Notably , two of these variants occurred on a common haplotype and were found in both affected children in the family ( S5A Fig . ) . A summary of all of the variants found in the nonsyndromic CL/P cases is found in S5 Table . We also sequenced 44 individuals of diverse ethnicities with CL/P plus syndactyly or limb defects including amniotic bands , polydactyly , and ectrodactyly ( which was motivated by the CLPS phenotype observed in dogs which includes syndactyly ) . From these samples , only one missense variant ( chr12: g . 43824214C>T ( p . A1108T ) ) was found in an individual with CL/P , facial asymmetry , and a single transverse palmar crease of the left palm . However , this variant did not segregate with clefting in the family ( S5C Fig . ) .
This study presents independent genome-wide association studies that provide evidence of the involvement of ADAMTS20 in the development of orofacial clefts in dogs and humans . The canine study identified a 1 . 44 Mb region of homozygosity underlying an association on CFA27 where subsequent whole-genome sequencing identified a frameshift mutation in ADAMTS20 that segregated with a complex phenotype of syndromic cleft lip , cleft palate , and syndactyly . The parallel human studies applied combinational-based association statistics to identify suggestive allelic ( DFAM ) and gene-level ( VEGAS ) associations with SNPs in ADAMTS20 in a cohort of native Guatemalans with nonsyndromic CL/P . Both studies identify ADAMTS20 as a candidate gene for CL/P development in humans . This work describes a second causative mutation for an independently segregating locus of orofacial cleft formation within the NSDTR dog breed . CLPS is characterized by a syndromic form of cleft lip , cleft palate , and syndactyly that segregates with a recessive mode of inheritance . This is independent of the previously identified CP1 locus that is characterized by a cleft palate and shortened mandible[24] . A genetic cause has not been identified in two cases of orofacial clefts within NSDTRs , further exemplifying the genetic heterogeneity within the NSDTR that has been previously documented[24] . This heterogeneity mimics what is observed in human cleft cases and is likely indicative of what will be observed in other dog breeds . ADAMTS20 ( a disintegrin-like and metalloprotease with thrombospondin type-1 motifs ) is part of a large family of secreted zinc metalloproteases sharing a similar domain organization[26] that are involved in cleaving extracellular matrix ( ECM ) proteins and processing procollagen[30] . ADAMTS20 cleaves the ECM proteoglycan , versican[26 , 31] , and is involved in a variety of biological processes including promotion of melanoblast survival , palatogenesis , and interdigital web regression[31–33] . Expression studies in mouse embryos identify craniofacial expression of Adamts20 in the first pharyngeal arch , between the medial nasal processes[34] , and broad expression in the palatal mesenchyme , where it plays a role in the sculpting and extension of the palate[32] . Adamts20 is also expressed in the developing fore- and hind limbs , the interdigital tissue , and at the medial border of the developing autopod[33 , 34] . In mice , mutations in Adamts20 are best known to cause a fully penetrant recessive , ventral to dorsal white belted phenotype ( bt ) [34] . In addition , bt mice have a low penetrance of cleft palate ( 3% ) and soft tissue syndactyly ( 18% ) [32 , 33] . Full penetrance of cleft palate was observed in bt mice with additional mutations in Adamts9 ( Adamts9+/-;bt/bt ) [32] . Within the NSDTRs , the CLPS deletion results in 100% penetrance of primary palate clefts . There is variation in the primary palate phenotype that ranges from clefting of the lower alar nasal folds to bilateral cleft lip . Secondary palate clefts are 92% penetrant , but also exhibit some variability in presentation . Syndactyly is likely fully penetrant as it was observed in 10 of the 13 dogs and the status of three remaining full-sibling cases is unknown . NSDTRs have minimal white spotting segregating in the breed , but none of the dogs with the associated haplotype exhibited any obvious midline white markings similar to what is observed in bt mice . This work complements what has been identified within the mouse by providing further evidence for role of ADAMTS20 in cleft palate and syndactyly formation . This highlights that ADAMTS20 should further be investigated for its role in CL/P and craniofacial development . Previous work describing mutations in ADAMTS20 and other ADAMTS family members may provide insight into the observed phenotype of bt mice and CLPS NSDTRs . ADAMTS proteins share identical N-terminal domains ( e . g . metalloprotease ) , but the type and number of C-terminal ancillary domains vary . These ancillary domains are critical for activity , inhibition , tissue localization , and substrate specificity[35] . Work on ADAMTS13 demonstrated that point mutations do not consistently function as null alleles[36] and deletion of different ancillary domains in ADAMTS5 and ADAMTS13 resulted in mutant constructs that retained partial function depending on the specific domains that were deleted[37–40] . Point mutations have been described in four bt alleles ( bt—c . 1598C>T p . Pro533Leu; bt9J—c . 2451C>T p . Leu761Phe; btBei1—c . 2860C>T p . Arg954Ter; bt Mri1—c . 4073A>C p . His1357Pro ) , which are located downstream of the catalytic metalloprotease domain[31 , 34] . In comparison , the deletion identified in CLPS NSDTRs ( c . 1360_1361delAA ( p . Lys453Ilefs*3 ) ) is closer to the N terminus and within the catalytic metalloprotease domain ( Fig . 3B ) [31 , 34] . Furthermore , the allele commonly used to study the bt phenotype ( btBei1 ) is a nonsense mutation that truncates 471 amino acids or 33% of the full-length protein ( Fig . 3B ) [32 , 34] . In ADAMTS13 , truncation after the spacer domain ( a mutation similar to the nonsense btBei1 allele ) results in a metalloprotease that is still active[37 , 38] . The mutation identified in CLPS NSDTRs truncates 1461 amino acids or 76% of full-length protein and may explain the higher penetrance of craniofacial defects and syndactyly observed in CLPS NSDTRs . In summary , this mutation could be the result of a more severe hypomorph than bt , a null , or a species difference . It is also interesting to note that the CLPS NSDTRs do not appear to have the white spotting pattern that is characteristic of the bt mice . Further studies examining the activity and regulation of ADAMTS20 will be necessary to dissect the molecular impact of the CLPS mutation and to determine if it is a null allele or a hypomorph . ADAMTS20 transcript expression levels are observed at only 55% in CLPS NSDTRs when compared to wild type NSDTRs . Transcripts with premature stop codons often undergo degradation by nonsense mediated mRNA decay to prevent accumulation of truncated proteins[41 , 42] , but often much lower expression levels are observed[43] . Since expression levels were analyzed in neonatal heart tissue , it is possible that a further decrease in expression levels may be observed in the appropriate tissue during the correct developmental time point . Although no SNP reached genome-wide significance in the human GWAS , the data presents a suggestive association with a SNP within ADAMTS20 ( rs10785430 ) . A similar phenotype observed in dogs and mice with mutations in the same gene , combined with the biological relevance of ADAMTS20 to development of the phenotype observed in the Guatemalans , suggests that this gene should be further investigated for CL/P development in humans . These results also indicate that even suggestive loci identified in human GWAS warrant further investigation . Sequencing of individuals with nonsyndromic CL/P from Guatemala and the Philippines did not identify any coding variants with obvious functional impact . Since the GWAS result suggests the existence of one or more common etiologic variants , it is possible that rare coding variants may yet be found in a subset of individuals with CL/P . We hypothesize that the etiologic variant ( s ) will be located in regulatory elements of ADAMTS20 , perhaps located within introns . We also sequenced ADAMTS20 in an additional cohort of individuals with syndromic CL/P who also had syndactyly or other limb defects . Although we did not find any coding variants in these individuals , ADAMTS20 mutations causing a cleft and syndactyly syndrome may be extremely rare . Variants in and around ADAMTS20 may also act as modifiers of the phenotype in clefting syndromes that including syndactyly as part of the phenotypic spectrum , such as Van der Woude syndrome and the ectodermal dysplasias . In conclusion , we showed separate association studies in humans and dogs that provide evidence of association between CL/P and variants in ADAMTS20 . This complements what is known in the mouse and suggests that ADAMTS20 should be further investigated for its role in CL/P development . Notably , dogs have long been used as models for craniofacial surgical techniques , but this study also demonstrates that they have the potential to be relevant models for the genetics of CL/P . Here we highlighted how dogs are a genetically amenable model organism with naturally occurring cleft lip ( the most common orofacial cleft type in humans ) and provide a different genetic background to study mutations .
Subjects for this study were recruited from various sites in Guatemala as part of the Pittsburgh Oral-Facial Cleft study , a large research program investigating factors that contribute to the development of CL/P and CP . Recruitment was done in collaboration with the nonprofit organization Children of the Americas ( www . childrenoftheamericas . org/ ) . Individuals and their family members seeking cleft lip and palate repair between 2004 and 2010 at multiple sites in Guatemala ( San Juan Sacatepequez , Huehuetenango , Tiquisate , Quiche and Retalhuleu ) were invited to participate in our study . This project was approved by both the University of Pittsburgh Institutional Review Board and the Oversight Ethics Committees of each of the participating hospitals; all participants gave informed , written consent in their native languages . Age appropriate assent documents were used for children between 7 and 14 years of age and informed , written consent was obtained from the child , as well as from the parents . Collection of canine samples in this study was approved by the University of California , Davis Animal Care and Use Committee ( protocol #16892 ) . Buccal swabs , blood , or tissue samples were collected from privately owned NSDTR dogs from the following phenotypic groups: CL/P ( n = 13 ) ; CP with a previously identified DLX6 LINE-1insertion ( n = 19 ) [24]; CP without a known causative mutation ( n = 2; 1 later identified to have CL/P ) [24]; healthy littermates of CL/P NSDTRs ( n = 13 ) ; healthy parents of CL/P NSDTRs ( n = 8 ) ; control NSDTRs ( n = 205 ) . Samples from dogs with orofacial clefts across 25 other breeds ( n = 53 ) were also obtained from privately owned dogs . Blood samples from control dogs ( n = 288 ) across 70 other breeds were collected from the William R . Pritchard Veterinary Medical Teaching Hospital ( VMTH ) . Embryos were collected from pregnant bitches undergoing ovariohysterectomy at the VMTH . Embryos were staged based on measurements of crown-to-rump length and the observation of external features[44] . Gross phenotypic evaluation of the orofacial clefts were performed by a board certified veterinary dentist who is experienced in the evaluation of cleft palate . Further phenotyping was performed by high-resolution microCT imaging of two CLPS NSDTR cases and three NSDTRs controls as previously described[24] . Genomic DNA was extracted from whole blood , tissue samples , or buccal swabs using Qiagen kits ( Valencia , CA ) . Genome-wide SNP genotyping was performed in 13 cases and 112 controls using the Illumina CanineHD BeadChip with 173 , 662 markers . All samples had a genotyping call rate of ≥ 90% . 62 , 546 SNPs were excluded due to a minor allele frequency ≤ 5% and 3 , 199 SNPs were excluded for a high failure rate ( ≥ 10% ) . 110 , 021 SNPs were used in the final analysis . Chi-square analysis was performed in PLINK[29] on one case from each litter ( n = 7 ) and discordant sibling pairs ( n = 4 ) when available . 100 , 000 permutations were performed to correct for multiple tests and the genomic inflation factor was calculated in PLINK[29] . Segregation analysis of the associated haplotype was performed in parents ( n = 6 ) and littermates ( n = 6 ) with enough available DNA for genotyping on the Illumina CanineHD BeadChip . Homozygosity throughout the associated interval was analyzed by visual inspection facilitated by color-coding homozygous genotypes in excel . The genotypes are available from the Dryad Digital Repository ( doi:10 . 5061/dryad . j8r8q ) . Seven NSDTRs were selected for whole-genome sequencing . Three CLPS NSDTR cases that were representative of the phenotypic spectrum were selected for sequencing: bilateral CL and CP with complete syndactyly of all paws , bilateral cleft of the lower alar nasal folds and CP with complete rear and incomplete front syndactyly , and bilateral cleft of the lower alar nasal folds with CP and incomplete rear syndactyly . Four control NSDTRs were selected for sequencing that were homozygous wild type throughout the associated region identified on CFA27 . Two of the four NSDTR controls had normal craniofacial structures . One NSDTR control had a cleft palate and shortened mandible explained by a DLX6 LINE-1 insertion[24] and the remaining NSDTR had a bilateral complete cleft lip with an unexplained genetic cause . Library preparation and DNA sequencing was carried out by the Ramaciotti Centre at the University of New South Wales , Kensington . Genomic DNA was size selected for 500bp fragments and sequenced on the HiSeq 2000 sequencing platform ( Illumina , San Diego , CA ) according to vendor’s instructions . Paired-end reads of 101 bp were generated for each sample on a single lane of the sequencer’s flow cell , yielding between 179 and 233 million read pairs per individual . Assuming a genome size of 2 . 5 Gb , this data reflected raw coverage of 14 . 4–18 . 8 fold . The canine genome sequence ( Canfam3 . 1;[22] ) was retrieved from the University of California , Santa Cruz genome browser ( UCSC , http://genome . ucsc . edu ) and indexed with the Burrows-Wheeler Transform Smith Waterman tool in the Burrows-Wheeler Alignment ( BWA ) package version 0 . 6 . 2[45] . Reads were aligned as pairs to the indexed reference genome using BWA , applying the default parameters for paired-end read alignment using this package . Alignment statistics generated using the idxstats tool within the SAMtools package version 0 . 1 . 18[46] indicated average mapped coverage ranging between 14 . 1 and 16 . 8 fold per individual . Snpsift was used to sort variants within the 1 . 44Mb interval by the ‘hom’ ‘any’ case and control filter options[25] . Annotation of remaining variants was performed with SnpEff using the xenoRefGene genes and gene prediction tracks annotation downloaded from the Table Browser window on the UCSC Genome Browser[25] . VCF files of the critical interval are available from the Dryad Digital Repository ( doi:10 . 5061/dryad . j8r8q ) . All primers described were designed in Primer 3 ( see S6 Table ) . Expression of ADAMTS20 was evaluated as described previously[24] . Total RNA was isolated from tissue samples using Qiagen QIAamp Blood Mini Kit tissue protocols . RNA was synthesized into cDNA using Invitrogen Superscript III First Strand Synthesis System to RT PCR protocols . ADAMTS20 cDNA was PCR amplified in heart tissue from one NSDTR case and a whole embryo ( collected at day 30 ) control . Areas with high GC content were amplified using Invitrogen AccuPrime GC-Rich DNA Polymerase protocols . PCR products were sequenced on an ABI 3500 Genetic Analyzer and analyzed using Vector NTI ( Informax , Frederick , MD , USA ) . Sequences were aligned to each other and Boxer ( Can Fam 3 . 1 ) reference sequence to identify any polymorphisms[22] . Primer sequences were generated using Primer3Plus ( http://primer3plus . com/ ) ( S6 Table ) . Semi-quantitative PCR using AmpliTaq Gold DNA Polymerase was performed to test the quality of cDNA and primers , to confirm product size , and to check for the presence of genomic DNA contamination . Real-time PCR was performed using the Rotor-Gene SYBR Green PCR Kit ( QIAGEN , Valencia , CA ) using a 2-step cycle protocol ( 35 cycles; Initial denaturation-5 minutes at 95°C; Annealing- 5 seconds at 95°C; Extension- 10 seconds at 60°C; Final Melt curve ) on the Rotor Gene Q real-time PCR system . cDNA from heart tissue of 3 neonatal NSDTR controls and 3 neonatal CLPS NSDTR cases were run in triplicates with each replicate containing 4–5 ng template cDNA . All data was normalized to the housekeeping gene , B2M[47] . Amplification and takeoff values were analyzed and graphed by REST2009[28] to determine any significant expression differences in ADAMTS20 transcript levels between case and control cDNA samples . PCR genotyping was performed according to standard protocols[24] using a shared FAM labeled forward primer ( S6 Table ) . GeneScan 500 ROX size standards were used and the reaction was analyzed on an ABI 3500 Genetic Analyzer . 97 unrelated control NSDTRs , 288 control dogs across 70 breeds , and 53 dogs with orofacial clefts across 25 breeds were genotyped for the deletion . All genotypes were analyzed using ABI GeneMapper software . From the larger study population , 937 individuals were genotyped ( see methods below ) : 545 from case families ( 125 affected with nonsyndromic CL/P and 420 unaffected relatives ) , plus 392 controls with no family history of orofacial clefts . From the genotyping , the population structure of the study subjects was compared to HapMap controls ( see S6 Fig . ) . The results show some European Caucasian admixture based on HapMap controls , plus substantial overlap with the Mexican HapMap Controls . All participants self-identified as Mayan; many spoke Quichean as well as Spanish . Trained health care professionals evaluated cleft phenotypes and ruled out syndromes for each participant . Each participant also provided detailed demographics , medical history , and family history , and female participants provided a detailed pregnancy history . Blood or saliva samples were obtained for DNA extraction using Qiagen kits ( Valencia , CA ) . Study participants were genotyped for 620 , 901 markers on the Illumina Human610-Quad ( Illumina Inc , San Diego , CA , USA ) . All individuals had a genotyping rate greater than 90% and therefore were included in the analysis . Deviations from Hardy-Weinberg equilibrium assessed in PLINK[29] found 622 markers with significant deviation from expectation in founder controls ( p ≤ 1e-06 ) . An additional 47 , 687 SNPs were eliminated because of high genotype failure rate ( ≥10% ) . An exclusion criterion of a minor allele frequency <5% in founders removed 105 , 758 SNPs . After removing non-autosomal SNPs a total of 457 , 969 SNPs remained for the analyses reported here . To further explore a putative association found on chromosome 12 ( see results below ) , we imputed 270 , 467 SNPs on this chromosome using IMPUTE2 software[48] and the 1000 Genomes Project as the reference sample ( http://browser . 1000genomes . org ) . The genotypes and phenotypes for the Guatemalan study population are available in dbGaP ( http://www . ncbi . nih . gov/gap ) , accession number phs000440 . v1 . p1 . Due to the heterogeneous family structures in our Guatemalan cohort , we performed an association analysis using the DFAM test implemented in PLINK , which integrates a standard TDT , discordant sib-TDT , and Cochran-Mantel-Haenzel clustered-analysis for case-control testing[29] . Independence between each test is established by considering individuals only once in the above statistical tests . For example , participants with relevant standard TDT information were not considered in sib-TDT or Cochran-Mantel-Haenszel clustered analysis . The order of assignment begins with a standard TDT , followed by a sib-TDT , then all remaining unrelated members are considered for the case-control analysis . To verify the top hits from the DFAM analysis we separately ran a standard TDT ( -tdt option in PLINK[29 , 49] , n = 65 trios ) , a sib-TDT ( -dfam command in PLINK in non-founders from nuclear families with multiple siblings and at least one affected family member , n = 49 families ) , and a case-control test ( 113 randomly chosen cases and 241 controls ) . To maximize sample size within these tests , we investigated each using all potential participants with the understanding the derived p-values are not necessarily independent between tests . For case-control analyses , we used logistic regression under the additive genetic model where genotypes were coded by the number of minor alleles ( 0 , 1 , 2 ) in each case or control . In addition to the DFAM analysis , we used a multivariate analysis to combine multiple SNPs within a gene into a single statistic . Gene level analysis has the potential benefit of increased power to detect associated regions containing multiple moderate effects[50] . VEGAS is a versatile gene-based test designed to handle any type of data input as long as a p-value can be generated for each individual marker[51] . Within a gene , p-values are converted to an upper tail chi-squared statistic with one degree of freedom and then combined . An empirical null distribution is created from a Monte Carlo simulation on a multivariate normally distributed random vector with a correlation equal to those predicted from a reference population through a Cholesky decomposition matrix . The proportion of simulated test statistics exceeding the observed gene-based test statistic gives the empiric p-value . In this situation , founders from the Guatemala data set were used as a reference population since no publicly available genetic data set sufficiently matches the participants’ genetic background . Gene plots with LD diagrams were generated with Locus Zoom and KGG2 . 5[52 , 53] . Primers covering the protein coding exons of ADAMTS20 were designed with Primer3 ( http://frodo . wi . mit . edu/primer3/ ) . Primer sequences and annealing temperatures are available in S7 Table . PCR products were sequenced on an ABI 3730XL ( Functional Biosciences , Inc . , Madison , WI ) . Chromatograms were then transferred to a Unix workstation , base-called with PHRED ( v . 0 . 961028 ) , assembled with PHRAP ( v . 0 . 960731 ) , scanned by POLYPHRED ( v . 0 . 970312 ) , and visualized with the CONSED program ( v . 4 . 0 ) . The functional effects of variants were predicted using the Ensembl database’s Variant Effect Predictor tool[54] .
|
Cleft lip with or without cleft palate ( CL/P ) is a commonly occurring birth defect that can lead to a lifetime of complications in affected children . To better understand the genetic cause of these disorders , we investigated CL/P in both dogs and humans . Genome-wide association studies in both species independently identify ADAMTS20 as a candidate gene for CL/P development . In dogs , a deletion within a functional domain of ADAMTS20 is responsible for CL/P in the Nova Scotia Duck Tolling Retriever dog breed . In humans , an associated region containing the same gene , ADAMTS20 , was identified in a study population of native Guatemalans . Subsequent sequencing in humans was unable to identify a causative mutation within the coding region of ADAMTS20 in the Guatemalan cohort; however , sequencing of ADAMTS20 in additional cases with CL/P identified four novel coding variants . This work provides genetic evidence for a role for ADAMTS20 in CL/P development in both dogs and humans .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2015
|
Genome-Wide Association Studies in Dogs and Humans Identify ADAMTS20 as a Risk Variant for Cleft Lip and Palate
|
Oral miltefosine has been shown to be non-inferior to first-line , injectable meglumine antimoniate ( MA ) for the treatment of cutaneous leishmaniasis ( CL ) in children . Miltefosine may be administered via in-home caregiver Directly Observed Therapy ( cDOT ) , while patients must travel to clinics to receive MA . We performed a cost-effectiveness analysis comparing miltefosine by cDOT versus MA for pediatric CL in southwest Colombia . We developed a Monte Carlo model comparing the cost-per-cure of miltefosine by cDOT compared to MA from patient , government payer , and societal perspectives ( societal = sum of patient and government payer perspective costs ) . Drug effectiveness and adverse events were estimated from clinical trials . Healthcare utilization and costs of travel were obtained from surveys of providers and published sources . The primary outcome was cost-per-cure reported in 2015 USD . Treatment efficacy , costs , and adherence were varied in sensitivity analysis to assess robustness of results . Treatment with miltefosine resulted in substantially lower cost-per-cure from a societal and patient perspective , and slightly higher cost-per-cure from a government payer perspective compared to MA . Mean societal cost-per-cure were $531 ( SD±$239 ) for MA and $188 ( SD±$100 ) for miltefosine , a mean cost-per-cure difference of +$343 . Mean cost-per-cure from a patient perspective were $442 ( SD ±$233 ) for MA and $30 ( SD±$16 ) for miltefosine , a mean difference of +$412 . Mean cost-per-cure from a government perspective were $89 ( SD±$55 ) for MA and $158 ( SD±$98 ) for miltefosine , with a mean difference of -$69 . Results were robust across a variety of assumptions in univariate and multi-way analysis . Treatment of pediatric cutaneous leishmaniasis with miltefosine via cDOT is cost saving from patient and societal perspectives , and moderately more costly from the government payer perspective compared to treatment with MA . Results were robust over a range of sensitivity analyses . Lower drug price for miltefosine could result in cost saving from a government perspective .
Cutaneous leishmaniasis ( CL ) is a neglected tropical disease primarily affecting poor , marginalized populations . Worldwide incidence of CL has been estimated at 1–5 million cases per year [1 , 2] . In Latin America , over 57 , 000 annual cases were reported on average between 2001 and 2013 [3] . In Colombia , 7 , 000–18 , 000 cases of CL are reported per year [4] . Of the 7 , 777 cases reported in 2015 , 71 . 5% were from rural areas , and 17% were younger than 15 years of age [5] . A recent retrospective study of the clinical and epidemiological profile of pediatric CL patients in Colombia indicates that children are increasingly affected by the disease due to population movements and environmental factors bringing vectors into closer contact with domestic settings [6] . CL patients in rural areas face economic and geographic challenges to securing treatment , including travel through zones of armed conflict . Pediatric populations are expected to incur particularly high costs , as their caregivers must accompany them to clinic for treatment . Additionally , MA has been shown to have a higher rate of renal clearance in pediatric patients , contributing to lower systemic exposure and a higher failure rate than adults [7 , 8] . Finally , MA has been associated with infrequent but serious adverse reactions , as well as intolerance of the intramuscular route of administration [9] . Miltefosine , a well tolerated medication for CL , has been shown to be non-inferior to MA in a clinical trial in pediatric patients in Colombia [10] . Miltefosine is administered orally and could be given via caregiver directly observed therapy ( cDOT ) at the patient’s home . Directly observed therapy ( DOT ) describes any protocol in which a trained observer watches medication administration to ensure compliance in order to avoid treatment failure and microbe resistance . Traditionally , this observer has been a health care professional , but protocols in which the observation is done in the home and by lay observers have been shown to be non-inferior in the case of tuberculosis [11] , the disease currently most commonly treated by DOT protocols . In our study , cDOT implies education of a pediatric patient’s caregiver in a manner that ensures course completion and appropriate use , including safe medication usage , storage and disposal . While cDOT has not yet been implemented for CL treatment , the efficacy [12–17] and cost-effectiveness [12 , 16] of the cDOT model for the treatment of tuberculosis have been established in a variety of contexts , including in pediatric populations [15 , 17] . Caregiver administration could ease the economic burden of CL treatment on families of affected children , as well as improve access and adherence to treatment in remote areas . Despite evidence demonstrating efficacy , the pediatric formulation of miltefosine is not widely available in Colombia . This study describes the relative costs of pediatric CL treatments with MA and miltefosine treatment for patients in southwest Colombia . This information is intended to guide policy makers , health ministries , and healthcare providers in countries with endemic CL .
We developed a cost-effectiveness analysis study using a Monte Carlo simulation model of CL treatment to examine the potential clinical and cost impact of miltefosine cDOT versus MA for a pediatric population with CL . The model incorporated data from multiple sources including public health databases [18–20] , primary surveys , expert opinion , and published data in order to project cost to stakeholders over the course of CL treatment . Simulated strategies were based on the interventional arms of the RCTs undertaken in Colombia and Brazil between 2007–2010 comparing the efficacy of intramuscular MA and oral miltefosine . Independent parameters were subjected to random assignment along assigned probability distributions in order to represent uncertainty and heterogeneity in these parameters . The study was conducted with public health data from four municipalities in two recognized endemic areas in Colombia [21] which have active leishmaniasis treatment programs—the lowland Pacific coastal municipalities of Buenaventura and Tumaco , and the Andean Central Cordillera municipalities of Chaparral and Rovira . These sites are among the municipalities with greatest transmission of CL in Colombia [5] . The University of Chicago Institutional Review Board and the Ethics Committee of CIDEIM and Universidad Icesi approved and monitored the study . Cost-effectiveness analyses were performed from patient , government payer , and societal perspectives [22] . The patient perspective included out-of-pocket costs assumed by patient caregivers during the course of treatment , such as transportation to clinics , meals outside the home , lodging , childcare , and medical supplies . The patient perspective excluded drug costs as these are publically covered . The government payer perspective included costs assumed by the Colombian healthcare system in the course of treatment such as drug costs , clinical medical supplies , and treatment associated with adverse events . The societal perspective combined the patient and government payer perspectives to estimate total cost associated with treatment . Costs are reported in 2015 USD [23] , and no discounting was applied , as the time frame of treatment was under one month . Sensitivity analyses were performed to test the stability of model outputs with variation of parameters . Primary outcomes were societal , patient , and government payer cost-per-cure for each treatment strategy . Cost difference is the MA cost-per-cure minus the miltefosine cDOT cost-per-cure . Cost neutrality is the point at which MA and miltefosine cDOT treatments incur the same cost-per-cure . We employed a Monte Carlo simulation model ( SimVoi v3 . 02 plugin for Microsoft Excel ) . Cases of clinically confirmed CL were simulated in a patient level probabilistic model in which unique patients entered the model and accrued costs to themselves and the government payer based on their treatment assignment , and left the model in either a cured or uncured state . Cure rate and adverse events during the course of treatment were modeled . Individuals failing treatment were not re-treated . Simulations of 100 , 000 patients were run for each potential intervention—meglumine antimoniate , miltefosine ( availability of adult and pediatric formulations ) , and miltefosine ( pediatric formulation only ) —to ensure stability of results . Baseline characteristics of children ages 2–12 with diagnosed CL were obtained from the pooled National Public Health Surveillance System ( SIVIGILA ) public health records from the municipality of Tumaco , Nariño from January 2012-May 10 , 2014 [18] and Chaparral , Tolima from March 7 , 2003-December 7 , 2011 [19] . The average age was 7 . 12 years and 49 . 7% of patients were female ( Table 1 ) . Patient weights were based on means and confidence intervals presented in the National Survey of the Nutritional Situation in Colombia in 2010 [24] . Weights were plotted on a normal distribution for each year of age for each sex . One- and multi-way deterministic analyses of selected parameters were performed to assess impact on base case results . The impact of treatment adherence was estimated by variation in treatment efficacy . Drug costs were varied , as was lost-time cost up to 100% of the Colombian minimum daily wage [35] during the course of MA treatment . Model inputs obtained from the survey were varied , including costs for supplies , treatment cost , transportation , meals , lodging , and childcare . A multi-way sensitivity analysis explored variations in the efficacy ratio of miltefosine cDOT over MA . The base case was considered to be equivalent efficacy ( efficacy ratio = 1 . 00 ) , and lower and upper bounds were derived from the upper 95% CI of miltefosine efficacy divided by the lower 95% CI of MA efficacy ( efficacy ratio = 1 . 19 ) and a reciprocal lower bound was calculated by subtracting the reciprocal change from the baseline assumption ( efficacy ratio = 0 . 81 ) . Cost-per-cure ratios ( miltefosine cDOT/MA ) , in which 1 indicates equivalent cost for miltefosine cDOT and MA , <1 indicates cost savings with miltefosine cDOT , and >1 indicates MA cost saving , were calculated .
In one-way sensitivity analysis of baseline assumptions ( Fig 1 ) , miltefosine cDOT remained cost saving compared to MA across a wide variation of parameters , including drug adherence varies between 50–100% . Cost superiority of miltefosine cDOT was also maintained as miltefosine and MA drug prices were varied from 50–200% of WHO negotiated prices . Increased cost-effectiveness was seen with the inclusion of up to 100% of daily minimum wage loss for the caregiver during the course of MA treatment . One-way sensitivity analyses of cost parameters collected by survey instrument - medical supplies cost , treatment cost , food , lodging , childcare , and transportation—showed no change in cost-per-cure superiority when varied between 50–200% of mean collected data ( Fig 2 ) . In multi-way sensitivity analysis , cost-per-cure ratio remained below 1 over a wide range of miltefosine cDOT-associated government payer costs and MA-associated patient costs ( Fig 3 ) . Miltefosine cDOT remained a cost-saving option from a societal perspective when MA-related patient costs were above 18% of the base case , and miltefosine cDOT-associated government payer costs was less than 355% of the base case . Stability in cost-per-cure ratio over these ranges was also demonstrated with the availability of 50 and 10mg formulations , as well as 10mg formulation only .
Our analytic model of treatment of CL in pediatric patients with miltefosine by cDOT versus current first-line MA treatment indicates that the miltefosine regimen is cost saving from a societal perspective . This result reflects considerably lower travel-associated costs for patients treated with miltefosine cDOT versus MA , a savings that exceeded the increased drug cost of miltefosine versus MA to the government payer . These results were robust across wide variations in parameters including adherence , medication efficacy , MA patient costs , miltefosine government payer costs , lost-time cost , adverse events costs , and direct patients costs . The availability of 50mg and 10mg formulations was associated with lower costs than availability of 10mg alone , but did not affect conclusions regarding cost superiority . It should be noted that previous studies have estimated higher government payer cost-per-cure for MA , which may indicate further cost advantage of the miltefosine cDOT protocol . A study of an outbreak in Colombia estimated costs at $345 per cure; with MA drug costs of 300% of the cost assumed in our analysis [36] . Government payer cost-per-cure for MA in Guatemala and Peru has been estimated at $280 and $300 , respectively [37 , 38] . The MA cost from these studies would exceed that estimated for miltefosine cDOT , making miltefosine cDOT cost-per-cure superior from the government payer perspective . An analysis of government payer cost-effectiveness of miltefosine and MA for adult CL patients in Colombia showed that miltefosine costs were nearly equivalent to MA costs [39] . However , no other analysis has focused on pediatric populations nor included patient and societal viewpoints . Conversely , it should also be highlighted that our use of the WHO pricing guidelines may represent a low cost for miltefosine in Latin America . Despite these guidelines , procurement costs in practice have been observed to be considerably higher [39 , 40] . We emphasize that acquisition of competitive drug prices by government actors is a priority in providing miltefosine cDOT therapy in a cost-sensitive budgetary context . Acquisition of drugs for NTDs bought in the absence of a national public health program are likely to be higher than drug prices achievable though centralized high volume ordering [41–43] , and as such , coordinated purchasing represents an opportunity for improvement of cost to the government payer . Additionally , pricing guidelines are subject to eventual renegotiation , in which case it is imperative that national , international , and non-governmental actors push for advantageous pricing of drug that carry significant benefits for marginalized patients . Burden of disease studies have demonstrated that leishmaniasis and other NTDs cause significant detriment to the lives and livelihoods of patients and caregivers in endemic areas [44–48] . Our study highlights that decisions on public health matters by government payers should consider more than direct expense , and incorporate value added and costs avoided by different options , as well as the ethical mandate of protection of vulnerable populations . As in many low-resource settings , direct cost saving at the level of drug purchasing is attractive from a budgetary standpoint . However , a systemic perspective of costs of disease and treatment may reveal reversals of treatment cost-effectiveness superiority when patient and societal points of view are considered . The findings of this study should be interpreted in the context of certain limitations . Firstly , effectiveness data for the treatment of CL is unavailable in Colombia and scarce among all countries of the region [49] . Secondly , the strict compliance conditions under which clinical trials are conducted do not represent the typical clinical experience with unsupervised treatment [50] . A 2014 Pan American Health Organization epidemiological report on the state of leishmaniasis indicates that only 31 . 6% of cases entered in the trans-national SisLeish surveillance system included clinical course [49] . The baseline assumption that adherence to medication was as observed in RCTs and did not vary between treatment regimens is a conservative assumption that may underestimate the benefits of oral miltefosine cDOT . Oral treatment is intended to lower barriers to care versus intramuscular injection of MA . Given that literature has estimated adherence to unsupervised miltefosine treatment for visceral leishmaniasis in Asia at 83% [51] to 95% [52] , we consider a high degree of adherence under a cDOT program achievable . Nonetheless , adherence will be a crucial consideration during the design and implementation of a cDOT program . Thirdly , modeling of costs , rather than direct costing of study participants was necessary due to the inclusion of a to-date theoretical cDOT protocol for miltefosine administration . The establishment and testing of specific protocols for a cDOT protocol remains a crucial step for this use of miltefosine . Among other concerns , the protocol must address re-administration of medication in cases of vomiting , the provision of specific education materials , implementation of methods to ensure adherence and adverse event accounting , and prevention of the use of the medication by household members of childbearing potential , due to miltefosine’s known teratogenicity [26 , 53] . Fourthly , susceptibility of distinct Leishmania species to particular drugs was not taken into account; however , current public health protocols do not identify species before initiation of treatment . A recent in vitro study of prevalent Leishmania Viannia species indicates high levels of susceptibility to both MA and miltefosine [54] . L . panamensis is the predominant strain in the area of the study [55] and has been shown to have good in vitro susceptibility to miltefosine [54] . While early tests of miltefosine indicated poor susceptibility of L . braziliensis [56] , subsequent testing has found greater susceptibility in South American strains [57] . Local species and susceptibility patterns will be an important consideration in adapting miltefosine cDOT programs in other areas . Concern for the emergence of resistant strains as a result of poor adherence has been described in L . Viannia species [58] , and necessitates that any forthcoming cDOT protocol ensure close monitoring to ensure continued drug efficacy . Fifthly , variation in clinical course was simplified . Simulated patients did not experience spontaneous resolution of CL within the timeframe of the primary outcome and did not experience progression of their disease to disseminated or mucocutaneous leishmaniasis , since these variations in natural history would be expected to be comparable for equivalently efficacious drugs . Super-infection or other complications occurring during CL were not considered in the analysis . Rare but serious ( CTCAE grades 4 and 5 ) complications were not included in the model , as none were experienced in the course of the trial from which modeling parameters were derived . Sixthly , dosing parameters did not take into account re-dosing in the case of vomiting , or potential changes in pediatric dosing regimens given evidence from pharmacokinetic studies showing inadequate drug plasma levels under current dosing guidelines [59 , 60] , although the costs of such cases may be extrapolated from sensitivity analysis of drug costs . Finally , assessment of patient and clinic costs in remote , often conflict-stricken zones necessitated the use of surveys of healthcare provider to gain local perspectives of the costs to patient patients and caregivers . We believe that their assessment reasonably approximates the costs and logistics of treatment , including transportation costs , which were among the elements of highest impact in cost determination . In summary , CL is a NTD causing significant morbidity and social stigma among marginalized pediatric populations . As new drugs are proven efficacious in treating this disease [10 , 27 , 28] , opportunities for novel treatment protocols that reduce cost to both patients and national healthcare systems may be possible and merit further exploration . Our analysis shows that treatment of pediatric patients with a miltefosine cDOT protocol is cost saving from patient and societal perspectives across a range of assumptions , and efforts to reduce miltefosine pricing could ultimately lead to cost neutrality or cost savings from a government perspective . Development of such treatment programs represents a critical opportunity to improve treatment and outcomes for pediatric CL patients .
|
Cutaneous leishmaniasis ( CL ) is a tropical parasitic disease transmitted by sand flies that causes chronic skin and mucosal ulcers . Current standard of care therapy requires patients to travel to a clinic for twenty consecutive days for injections of meglumine antimoniate ( MA ) . This may represent an economic burden , particularly for patients living far from healthcare services , especially children and their caregivers . We performed mathematical modeling to compare costs of the standard of care treatment with costs of miltefosine , an equivalently efficacious oral medication that allows pediatric patients to be treated at home under trained supervision of a caregiver . In our model , miltefosine led to substantially lower costs for patients and only slightly higher costs to the healthcare system . Importantly , the cost to society ( combined patient and healthcare system costs ) was lower for miltefosine compared to MA . Treatment of pediatric CL with miltefosine in the patient’s home could decrease overall cost of treatment , while diminishing the barriers and cost burden on patients , their caregivers , and society .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"medicine",
"and",
"health",
"sciences",
"cost-effectiveness",
"analysis",
"clinical",
"research",
"design",
"engineering",
"and",
"technology",
"transportation",
"economic",
"analysis",
"geographical",
"locations",
"tropical",
"diseases",
"social",
"sciences",
"parasitic",
"diseases",
"pediatrics",
"research",
"design",
"pharmaceutics",
"neglected",
"tropical",
"diseases",
"research",
"and",
"analysis",
"methods",
"public",
"and",
"occupational",
"health",
"infectious",
"diseases",
"zoonoses",
"south",
"america",
"adverse",
"events",
"protozoan",
"infections",
"economics",
"people",
"and",
"places",
"colombia",
"leishmaniasis",
"drug",
"therapy"
] |
2017
|
Cost-effectiveness of meglumine antimoniate versus miltefosine caregiver DOT for the treatment of pediatric cutaneous leishmaniasis
|
The efficiency of translation termination depends on the nature of the stop codon and the surrounding nucleotides . Some molecules , such as aminoglycoside antibiotics ( gentamicin ) , decrease termination efficiency and are currently being evaluated for diseases caused by premature termination codons . However , the readthrough response to treatment is highly variable and little is known about the rules governing readthrough level and response to aminoglycosides . In this study , we carried out in-depth statistical analysis on a very large set of nonsense mutations to decipher the elements of nucleotide context responsible for modulating readthrough levels and gentamicin response . We quantified readthrough for 66 sequences containing a stop codon , in the presence and absence of gentamicin , in cultured mammalian cells . We demonstrated that the efficiency of readthrough after treatment is determined by the complex interplay between the stop codon and a larger sequence context . There was a strong positive correlation between basal and induced readthrough levels , and a weak negative correlation between basal readthrough level and gentamicin response ( i . e . the factor of increase from basal to induced readthrough levels ) . The identity of the stop codon did not affect the response to gentamicin treatment . In agreement with a previous report , we confirm that the presence of a cytosine in +4 position promotes higher basal and gentamicin-induced readthrough than other nucleotides . We highlight for the first time that the presence of a uracil residue immediately upstream from the stop codon is a major determinant of the response to gentamicin . Moreover , this effect was mediated by the nucleotide itself , rather than by the amino-acid or tRNA corresponding to the −1 codon . Finally , we point out that a uracil at this position associated with a cytosine at +4 results in an optimal gentamicin-induced readthrough , which is the therapeutically relevant variable .
Translation is terminated by a stop codon entering the A site of the ribosome , inducing the release of the polypeptide chain from the peptidyl-t-RNA [1] . Two polypeptide chain release factors have been identified in eukaryotes: eRF1 ( eukaryotic Release Factor one ) , which recognizes all three nonsense codons , and eRF3 ( eukaryotic Release Factor three ) which stimulates polypeptide release from the ribosome in a GTP and eRF1-dependent manner [2] . Under normal conditions , translation termination at natural termination codons is a very efficient process , with an estimated error rate of 0 . 01 to 0 . 1% in mammalian cells ( unpublished data ) . However , near-cognate aminoacyl-tRNAs ( with pairing of two of the three bases ) can compete with eRF1 for stop codon binding , resulting in translational readthrough . The rules governing readthrough efficiency are far from clear , but readthrough levels have been shown to depend on the type of stop codon , with UAA being a better terminator than UAG and UGA , and even more strongly on the surrounding nucleotide context [3]–[6] . The effect of nucleotide context on readthrough level has been extensively studied in yeast: Bonetti and coworkers demonstrated that upstream and downstream components act in synergy to determine readthrough efficiency [7] . Two studies based on the screening of a degenerate oligonucleotide library established a consensus sequence , NAA STOP CA ( A/G ) N ( U/C/G ) A , promoting high levels of readthrough [6] , [8] . Less is currently known about the effect of sequence context on readthrough levels in mammalians cells , because only a small number of contexts surrounding stop codons have been investigated . Readthrough can be stimulated by aminoglycoside antibiotics , such as gentamicin , making it possible to generate a full-length protein from genes carrying a nonsense mutation [9] , [10] . These antibiotics interact with the highly conserved decoding center of ribosomal RNA , promoting the recognition of the stop codon by a near-cognate tRNA [11] , [12] . A number of studies and clinical trials have investigated the possible use of this antibiotic for the treatment of human diseases resulting from the presence of a premature termination codon ( PTC ) in a particular gene ( for review , see [13] , [14] ) . However , basal and induced readthrough efficiencies differ considerably between nonsense mutations [3] , [15] so only a subset of patients would be likely to benefit from gentamicin treatment . Moreover , due to the complexity of the mechanisms involved , it is not possible to predict readthrough efficiency from the nucleotide context of the nonsense mutation . It is crucial to determine the patients most likely to benefit from treatment , and it is currently necessary to measure the readthrough level of each nonsense mutation in cell culture , as readthrough levels in culture are correlated with those in vivo [16] . Many studies have indicated that the nucleotide immediately downstream from the stop codon ( defined as +4 ) is a crucial determinant of termination efficiency in eukaryotes [17] . Moreover , this nucleotide has been shown to crosslink with release factor class I [18] . A cytosine ( C ) residue in the +4 position generally promotes higher levels of readthrough in the presence or absence of aminoglycosides . However , some nonsense mutations with a C residue in the +4 position may display moderate levels of readthrough [3] . Thus , the identity of the nucleotide immediately downstream from the stop codon is not sufficient to predict readthrough efficiency for a given nonsense mutation . No systematic study has been performed and our knowledge of the effects of nucleotide context on readthrough level and gentamicin response ( i . e . the factor of increase between basal readthrough and drug-induced readthrough ) is therefore incomplete . We used a set of 66 sequences , each containing a stop codon - mostly nonsense mutations implicated in various human diseases - inserted into the same reporter vector for an extensive statistical analysis of the determinants of readthrough levels and gentamicin response . We found a strong correlation between basal readthrough level and antibiotic-induced readthrough level and a very weak negative correlation between basal readthrough level and gentamicin response . The nature of the stop codon did not affect the sensitivity of the nonsense mutation to gentamicin treatment . A comprehensive analysis of the surrounding nucleotides identified positions playing an important role in determining readthrough levels and gentamicin response . In particular , we demonstrated that the nucleotide immediately upstream from the stop codon was a major determinant of gentamicin response and that this effect was mediated by the nucleotide itself , rather than by the nature of the last amino acid or the tRNA present in the ribosomal P-site . Based on these findings , we have developed the first rules for predicting the sensitivity of nonsense mutations to aminoglycoside treatments based on the surrounding nucleotide sequence .
We analyzed readthrough levels for 66 stop codons , including one natural termination codon and 65 nonsense mutations implicated in various diseases ( Figure 1 and Table S1 ) : The CFTR gene for cystic fibrosis [19] , the dystrophin gene for Duchenne muscular dystrophy [3] , the LAMA-2 gene for congenital muscular dystrophy [16] , the beta-globin gene for beta-thalassemia ( sequences provided by Jacques Rochette , INSERM U 925-UPJV , Amiens ) and the p53 and APC ( adenomatous polyposis coli ) genes for cancers [20] , [21] . The stop codon present in the mouse mdx gene is denoted “MDX” . “STOP LAM” is the natural termination codon of laminin and “STOP PLATI” is the mouse platinum coat color mutation . Nonsense mutations are named according to the position of the modified amino acid in the protein sequence . For each sequence , the stop codon and the surrounding nucleotide context , shown in Table S1 , were inserted into the dual reporter vector pAC99 [22] . Readthrough levels were quantified in NIH3T3 cells transiently transfected with the dual reporter vector , in the presence or absence of gentamicin . Some of these nonsense mutations have already been tested in previous studies in our laboratory . However , as variability is commonly observed between batches of gentamicin [23] , we test all 66 stop sequences with the same gentamicin preparation ( see Materials and Methods ) . Readthrough rates ranged from 0 . 01% ( DMD 2726 , beta 43 , APC 1131 ) to 0 . 52% ( CF 122 ) for basal readthrough ( B ) , and from 0 . 04% ( p53 327 ) to 2 . 79% ( p53 213 ) in the presence of 800 µg/ml gentamicin ( G ) ( Figure 1 , Table S1 ) . The gentamicin response is defined as the factor of increase ( I ) between basal and gentamicin-induced readthrough levels . This factor of increase varied from 1 . 6 ( DMD 2125 ) to 16 . 3 ( APC 1131 ) . Considerable variability for the three variables was observed , as previously described . We characterized readthrough levels and the gentamicin response in mammalian cells in more detail , by carrying out statistical analysis . We studied the distribution and characteristics of the variables B , G and I , by descriptive statistical analysis ( Table S2 ) . For B , the mean was 0 . 07% and the median was 0 . 04%; for G , the mean was 0 . 37% and the median was 0 . 23%; for I , the mean was 6 . 04 and the median was 5 . 46 . The difference between the mean and the median indicates asymmetry in the distribution . We also carried out a graphical analysis to visualize the distribution of each variable ( Figure 2 ) . The variables B and G had a very high kurtosis ( flattening coefficient; 11 . 11 and 12 . 71 , respectively ) , indicating a sharper peak than for a Gaussian distribution ( kurtosis = 0 ) . The asymmetry coefficient was 3 . 22 for variable B and 3 . 23 for G , respectively , indicating a strongly asymmetric and L-shaped distribution , with a high proportion of low values . The values were found to be homogeneously distributed , with most located in the first two intervals ( Figure 2 ) : For increase factor ( variable I ) , we observed a kurtosis slightly higher ( 1 . 63 ) than expected for a Gaussian distribution . Its asymmetry coefficient ( 1 . 17 ) was similar to that for a Gaussian distribution ( = 1 ) . The values were homogeneously distributed and 78% of the values had ranks between 4 and 8 . Values greater than 8 accounted for 19 . 7% of all values and were defined as a “high” factor of increase . We then investigated whether these three distributions could be converted to Gaussian distributions using the Box-Cox transformation ( λ = −0 . 217 ) ( see Materials and Methods ) , which would make it possible to use more powerful parametric statistics . After transformation , a Shapiro-Wilk test allowed us to conclude that B , G and I variables indeed followed a normal distribution ( Figure S1 , Table S3 ) . Previous observations have suggested that there is no correlation between the basal readthrough level for a nonsense mutation and its sensitivity to gentamicin treatment [3] . However , the sets of mutations analyzed to date have been too small to demonstrate this point statistically . We plotted the level of gentamicin-induced readthrough against basal readthrough and the increase factor against basal readthrough or against gentamicin-induced readthrough before ( Figure S2 ) and after Box-Cox transformation ( Figure 3 ) . We used the parametric Bravais-Pearson correlation test to establish the statistical significance of these correlations ( Table S4 ) . There was a strong positive correlation between basal readthrough level and gentamicin-induced readthrough level ( R = 0 . 845 ) and this correlation was significant ( p<0 . 0001 ) . There was a weak negative correlation between basal readthrough and the factor of increase ( R = −0 . 29 , p = 0 . 016 ) , indicative of a trend , with nonsense mutations with a “high” basal readthrough level tending to be less responsive to gentamicin treatment . There was a weak positive correlation between gentamicin-induced readthrough level and the factor of increase ( R = 0 . 25 p = 0 . 045 ) . Thus , nonsense mutations with ‘high” gentamicin-induced readthrough levels also tended to have the highest factor of increase . These results provide the first description of the relationship between basal readthrough , gentamicin-induced readthrough and gentamicin response . They indicate that nonsense mutations with a “high” basal readthrough level give “high” levels of gentamicin-induced readthrough . However , some nonsense mutations with a “low” basal readthrough level presented a “high” gentamicin-induced readthrough level , because they had high factors of increase . Thus , nonsense mutations were found to behave in different ways and could be classified into three distinct groups ( Table S1 ) : Response-type 1: High basal readthrough levels and high gentamicin-induced readthrough levels ( for example , p53 213 , CF 122 or DMD931 ) . These nonsense mutations did not have a particularly high factor of increase . Response-type 2: A low or medium basal readthrough level associated with high factor of increase , resulting in essentially high levels of gentamicin-induced readthrough ( for example , APC 1114 or p53 192 ) . Response-type 3: A low or medium basal readthrough level and a weak or moderate gentamicin response . Most of the nonsense mutations studied was of this type . The first two groups include mutations for which gentamicin treatment can promote high levels of readthrough . For these mutations , we would expect to observe clinical benefit for the treatment , with gentamicin , of diseases linked to the presence of a nonsense mutation . Indeed , similar levels of induced readthrough have already been shown to improve the clinical status of cystic fibrosis patients with CFTR mutations treated with gentamicin [19] . We hypothesized that the differences in the behavior of these nonsense mutations should depend on the nature of the stop codon and the nucleotide context . We therefore assessed the contribution of each of these factors . The statistical approach used is described in the Materials and Methods section and in Figure 4 . These 66 sequences were assigned to three different groups , according to the nature of the stop codon . There were 14 UAA , 25 UAG and 27 UGA stop codons . The medians of the three variables are shown , for each group , in Figure 5 . After Box-Cox transformation which allowed us to obtain a normal distribution for these 3 variables , an ANOVA test revealed that the three groups differed significantly in terms of their basal and induced readthrough levels . A LSD test yielded the following hierarchy: UGA>UAG>UAA ( the sign>represents a statistical difference ) for both basal and induced readthrough levels ( Table 1 and Table S5 ) . This hierarchy is consistent with previous reports but , to our knowledge , this study provides the first evidence of a statistically significant difference between stop codons . However , some UAA codons have higher readthrough levels than some UGA or UAG codons ( i . e . CF 122 ) , highlighting the crucial role of nucleotide context in determining readthrough level in the presence or absence of gentamicin . Conversely , ANOVA test revealed that the factor of increase did not differ significantly between the three groups ( Table S5 ) . We show here that the factor of increase , which reflects the capacity of a nonsense mutation to respond to treatment , was independent of the nature of the stop codon . The factor of increase therefore probably depends only on the nucleotide context of a given nonsense mutation . We then investigated the effects of nucleotide context , by the same statistical approach used for investigation of the effects of stop codon identity ( see Materials and Methods and Figure 4 ) . Six nucleotides upstream and downstream from the stop codon have already been shown to influence readthrough level in eukaryotes [6] . We therefore analyzed the effect of each nucleotide in this interval . Graphic representations of the medians of B , G and I ( before normalization of the data ) , for each nucleotide , at each position ( from −6 to +9 ) , were generated ( Figure 6 ) . After Box-Cox transformation we were able to use parametric statistical tests ( ANOVA and LSD ) to define a hierarchy for some positions ( Table 1 ) . Bartlett p value and ANOVA F and p-values are indicated in Table S6 , Table S7 and Table S8 . These analyses were conducted with the complete data set , but could not be applied to each class of stop codon separately , because the number of mutations in each class was too small for statistical analysis . Our analysis thus only identified determinants valid for all three types of stop codon . We first compared the effect of each nucleotide at a given position to the three others at the same position ( Table S6 ) . Our findings confirmed the involvement of distal 5′ and 3′ determinants of nucleotide sequence context in the control of readthrough level and gentamicin response . Indeed , for the nucleotides in positions −6 to +9 , we were able to establish correlations between particular classified bases and high levels of readthrough or strong gentamicin response . Nucleotides can be classified according to their effect , for at least one variable , for nine ( −6 , −5 , −3 , −2 , −1 , +4 , +5 , +8 , +9 ) of the twelve positions studied . For example , an adenine or a cytosine residue in position −6 was associated with higher basal readthrough levels than observed for a uracil residue . A guanine residue in position +8 was associated with a stronger gentamicin-induced readthrough than a cytosine residue in this position . The presence of a uracil residue in position +9 was associated with a stronger gentamicin response than a cytosine or a guanosine residue in this position ( Table 1 ) . These findings contrast with results previously obtained in yeast , which pointed out the role of the two adenine residues immediately upstream from the stop codon in the absence of treatment [8] . This discrepancy is possibly due to differences between mammals and yeast , or to the use of a sequence harboring a motif downstream from the stop codon responsible for promoting particularly high levels of readthrough in this previous study . Two major determinants were identified in this study ( Table 1 ) : However among mutation presenting a U in −1 position , 30% also present a C in +4 ( against 8% for mutation with an A , 12% for mutation with a G and 29% for mutation with a C in −1 ) . To assess that the effect of U in −1 is not biased by the presence of a C in +4 , we performed the same statistical analysis restricting the pool of mutations to those without a C in +4 . In this subset , the mean value of gentamicin-induced readthrough and increase factor is even better when there is a U in −1 position compared to the 3 other nucleotides ( Table S7 ) . This result confirms the effect of a uracil in −1 position independently of the presence of a C in +4 . Nevertheless , according to the statistical test performed , it can be noticed that for induced-readthrough level there is a combined effect between the U in −1 position with a C in +4 position . Indeed , all the nonsense mutations studied that carried a U Stop C sequence systematically displayed readthrough levels exceeding 0 . 5% in the presence of gentamicin . We checked that the determinants identified here were retrospectively consistent with published readthrough analyses . Keeling and Bedwell [5] measured the levels of readthrough induced by several aminoglycosides in a mammalian translation system . The mutation displaying the highest levels of gentamicin- and amikacin-induced readthrough was indeed the only one with a “U stop C” sequence . The combination of these two nucleotides on either side of the stop codon therefore constitutes the first rule ever elucidated for identifying patients with nonsense mutations likely to respond to aminoglycoside treatment . We also compared the effect of each nucleotide at a given position to the four nucleotides at all positions on B , G and I . This procedure reveal a clear effect for the increase factor p = 0 . 0002 ( Table S8 ) . For example a uracil residue in position −1 was associated with higher increase factor than observed for a guanosine residue in position −3 . We analyzed the effect of the nucleotide in position −1 independently of the influence of other nucleotides , by quantifying the readthrough levels of six nonsense mutations in which we changed only the nucleotide in position −1 , keeping the rest of the sequence constant: Readthrough levels were quantified in NIH3T3 cells in the presence or absence of gentamicin ( Figure 7 ) , statistical data and standard error of the mean are indicated in Table S9 . We found that the presence of a U residue in the −1 position was systematically associated with a higher factor of increase than the presence of any other nucleotide in this position . This result confirms the statistical analysis of the 66 nonsense mutations . However , for this narrow panel of stop contexts , gentamicin-induced readthrough levels were not necessarily higher in the presence of a U residue . These levels could be lower ( DMD 673 , p53 146 and CF 122 ) , equivalent ( DMD 931 ) or higher ( DMD 319 , beta 17 ) . The effect of the nucleotide in position −1 on gentamicin-induced readthrough level therefore depends strongly on the nature of the other nucleotides surrounding the stop codon . These results provide evidence that the nucleotide immediately upstream from the stop codon is a major determinant of gentamicin response , with an uracil residue in this position associated with stronger responses to gentamicin treatment . For therapeutic purpose , readthrough levels in presence of gentamicin are the relevant variable . However , the capacity of a nonsense mutation to increase its readthrough level after antibiotic treatment could be a crucial point in the future as new readthrough inducers will be available . Indeed several groups are currently developing news molecules derived from aminoglycosides and acting in a similar way but with a greater efficiency [25]–[27] . In this case , a nonsense mutation with a good increase factor could overtake the threshold of 0 . 5% of readthrough . We then examined how the nucleotide upstream from the stop codon exerted its effect on readthrough levels . In prokaryotes , the chemical properties of the ultimate amino acid in the nascent polypeptide chain have been reported to modulate translational readthrough [28] , [29] . The −1 nucleotide may also influence readthrough by interacting directly with the P site tRNA or indirectly with eRF1 . We therefore investigated whether the final tRNA or amino acid incorporated had an effect on readthrough levels . The nucleotide in the −1 position is the third base of the codon immediately upstream from the stop codon ( codon −1 ) . During translation termination , the stop codon is located in the ribosomal A-site and codon −1 is in the P-site . We therefore investigated whether having a hydrophilic or hydrophobic amino acid at the P site was correlated with higher levels of readthrough or stronger gentamicin responses . A two-tailed t-test comparing the two groups ( hydrophobic or hydrophilic amino acid ) for all the nonsense mutations studied showed that there is no relationship between the nature of the final amino acid and a high factor of increase ( t = −0 . 91; p = 0 . 36 ) or high readthrough rates ( t = 1 . 71; p = 0 . 09 for B and t = 1 . 28; p = 0 . 2 for G ) . Moreover , the amino acids encoded by codons ending in U do not belong to a particular chemical class . These results strongly suggest that the nature of the amino acid at the ribosomal P-site is not a major determinant of readthrough levels . We then investigated whether the effect of the nucleotide in the −1 position on the factor of increase was due to the nature of the tRNA at the P site . Nucleotides 1 , 2 and 3 of the mRNA codon are recognized by nucleotides 36 , 35 and 34 , respectively of the tRNA anticodon ( Figure S3A ) . Codons ending in a C or U residue may be recognized through wobble pairing at position 34 of the anticodon ( Figure S3B ) . In such situations , a single tRNA may recognize several codons . There are two possibilities in eukaryotes: nucleotides U3 and C3 of the codon may be recognized by nucleotides A34 or G34 on the anticodon . Thus , U3 may be recognized by wobble pairing ( G34 ) or Watson-Crick pairing ( A34 ) . We investigated the way in which recognition of the codon in position −1 affected readthrough levels or gentamicin response , by comparing nonsense mutations for which the −1 codon is recognized by wobble pairing with those recognized by Watson-Crick pairing , in two-tailed t-test . We found no significant difference between these two types of nonsense mutation , for B ( t = 0 . 61; p = 0 . 54 ) , G ( t = 0 . 63; p = 0 . 53 ) or I ( t = −0 . 099; p = 0 . 92 ) . The strength of base pairing between the −1 codon and the corresponding anticodon therefore seems to have no influence on readthrough levels or gentamicin response . We investigated whether the effect of the −1 nucleotide on the factor of increase was correlated with the identity of the tRNA , using four nonsense mutations: beta 17 , DMD 319 , ( Figure 7 ) and APC 1131 ( UAA , response-type 2 ) , APC 1114 ( UGA , response-type 2 ) ( Figure 8 ) . The third base of the −1 codon of these nonsense mutations was changed to create an alternative –1 codon recognized by the same tRNA ( Table 2 ) . For example , the AAU −1 codon of the APC 1114 nonsense mutation was replaced by an AAC codon , which is also recognized by the ( 3′→5′ ) UUG anticodon of the same tRNA ASN . Readthrough levels were quantified in NIH3T3 cells transiently transfected with the dual reporter vector containing the appropriate sequence , in the presence or absence of gentamicin . For these nonsense mutations , the factor of increase and the gentamicin-induced readthrough were higher when there was a U residue in position −1 than when there was another base in this position , while the modified codon was recognized by the same tRNA . These findings provide strong evidence for a lack of involvement of the tRNA at the ribosomal P-site in determining the gentamicin response . Thus , the nucleotide in the −1 position is itself a major determinant of the gentamicin response and of the gentamicin-induced readthrough level . We used the largest set of nonsense mutations ever analyzed for the first statistical analysis of the influence of nucleotide context on PTC readthrough and response to aminoglycoside treatment . We confirm the findings of previous studies concerning the importance of the nucleotide in the +4 position , at which the presence of a cytosine ( C ) residue is correlated with high basal and gentamicin-induced readthrough levels . We also show for the first time that the presence of a U residue in −1 is a key determinant of gentamicin-induced readthrough which is the relevant parameter for clinical applications . Moreover , we can notice that a U in −1 is also correlated with a higher increase factor between basal and gentamicin-induced readthrough . This finding may have important implications in fundamental aspects of structural interactions between readthrough inducers and the translational apparatus . We show that impact of the base in the −1 position is mediated neither by the last amino acid nor by the tRNA present at the ribosomal P site . These data are consistent with previous reports excluding a role for the last residue of the polypeptide chain or the last incorporated tRNA in readthrough efficiency in eukaryotes [8] , [30] . Different rules seem to apply in prokaryotes , as the two last amino acids and the tRNA present in the P site have been shown to influence termination efficiency in E . coli [31] . It remains unclear how this nucleotide modulates the factor of increase in mammals . One possible hypothesis is that the stacking properties of this base in the vicinity of the stop codon are involved in the balance between translation termination and suppression . More generally , this nucleotide , which is close to the decoding center targeted by aminoglycosides , may induce local structural rearrangements favoring the influence of aminoglycosides at the ribosomal A site . Finally , the consensus sequence U STOP C was systematically associated with induced readthrough levels greater than 0 . 5% . The combination of these two nucleotides before and after the stop codon may therefore provide an initial indicator of readthrough levels compatible with therapeutic benefit .
NIH3T3 cells ( embryonic mouse fibroblasts kindly provided by Marc Sitbon ) were cultured in DMEM plus GlutaMAX ( Invitrogen ) . The medium was supplemented with 10% foetal calf serum ( FCS , Invitrogen ) and 100 U/ml penicillin/streptomycin . Cells were kept in a humidified atmosphere containing 5 . 5% CO2 at 37°C . Complementary oligonucleotides corresponding to nonsense mutations embedded in their natural context ( sequences in Table S1 ) were annealed and ligated into the pAC99 dual reporter plasmid , as previously described [22] . This dual reporter can be used to quantify stop-codon readthrough , through the measurement of luciferase and beta-galactosidase ( internal calibration ) activities , as previously described [19] . Readthrough levels for nonsense mutations were analyzed in the presence or absence of gentamicin . All nonsense mutations were tested with batches of gentamicin with identical efficiency of readthrough levels . For p53 and APC , results were obtained from recent studies [20] , [21] . For CF , DMD and CMD [19] , [3] and [16] , readthrough levels had already been estimated but the tests were repeated in this study , to prevent discrepancies due to the use of different batches of gentamicin . NIH3T3 cells were electroporated with 20 µg of reporter plasmid . The following day , the cells were rinsed and fresh medium , with or without gentamicin ( 800 µg/ml ) , was added . No cell toxicity was observed with this dose of antibiotic . Cells were harvested 24 hours later , by trypsin-EDTA treatment ( Invitrogen ) and lysed . Beta-galactosidase and luciferase activities were assayed as previously described [22] . Readthrough efficiency was estimated by calculating the ratio of luciferase activity to beta-galactosidase activity obtained with the test construct , normalizing the value obtained with respect to that obtained with an in-frame control construct . For each construct , we performed at least three independent transfection experiments ( 3 to 10 experiments ) . Excel was used for statistical analysis: the Analysis Toolpack for descriptive statistics; the XL- stat for Bartlett correlation tests ( Bravais-Pearson ) and t-tests , Analyse-it module for ANOVA and LSD tests . Descriptive statistics ( Table S2 ) provided simple information ( mean , median etc . ) about three variables: basal readthrough level ( B ) , gentamicin-induced readthrough level ( G ) and the factor of increase between basal and induced readthrough levels ( I ) . The median , which is obtained by arranging the values in size order and selecting the middle value , is useful when the distribution does not follow a Gaussian distribution ( i . e . for variables that do not tend to cluster around a single mean value ) . Graphical analysis was performed and the values of each variable were ranked ( Figure 2 ) . The intervals between different ranks were identical and no minimal number of values per rank was required . The intervals and number of rows were defined according to a convention taking into account the total number of values and the minimal and maximal values for each of the variables studied . Values included in ranks corresponding to the best ∼20% of values were defined as “high” . Several parameters ( Kurtosis coefficient , asymmetry coefficient etc . ) and graphical analysis were used to determine whether the distribution of each variable followed a Gaussian distribution . A Gaussian distribution is characterized by a Kurtosis coefficient of 0 and an asymmetry coefficient of 1 . In order to perform a complete statistical analysis we chose to use parametric tests instead of low-power non-parametric analysis . To this aim , we performed a Box-Cox transformation for variables B , G and I with Ψλ ( x ) = ( xλ−1 ) /λ using the same lambda: −0 . 217 . After this transformation , a Shapiro-Wilk test allowed us to conclude that B , G and I follow a normal distribution . Correlations between variables were analyzed with the parametric Bravais-Pearson test . The null hypothesis ( H0 ) was “there is no correlation between the two variables studied ( R = 0 ) . The alternative hypothesis ( H1 ) was “there is a correlation between the two variables studied ( R≠0 ) ” . A perfect positive correlation gives an R value of +1 , whereas a perfect negative correlation gives an R value of −1 . The significance level was set at 0 . 05 . In order to analyze the effect of the nature of the stop codon or the nucleotide context on readthrough levels and gentamicin response the 66 stop codons were divided into three groups for stop codon studies ( UAA , UAG and UGA ) and four groups for nucleotide context studies ( U , C , A , G ) . For each group , we then used a Bartlett test to analyze heterogeneity of variance of each variable . If heterogeneity was not significant we performed one of the most commonly used multiple comparison procedure , the Fisher's Least Significant Difference ( LSD ) test . The LSD test is a two-step test . First an ANOVA ( Analysis Of Variance ) test is performed . The null hypothesis for ANOVA is that the mean ( average value of the dependent variable ) is the same for all groups . The alternative hypothesis is that the mean is not the same for all groups . When the null hypothesis is rejected , it means that at least 2 groups are different from each other . In a second step we determine which groups are different from which performing all pairwise t-tests . This last procedure allows to establish a hierarchy between stop or nucleotide at each position . In Table S6 each nucleotide at a given position is compared to the others at the same position and in Table S8 each nucleotide is compared to the four nucleotides at all position . Two-tailed Student's t-tests ( excel ) were used to study the influence of tRNA or the amino acid in the ribosomal P-site . For this test , we used a significance level α of 0 . 05 .
|
Nonsense mutations are single-nucleotide variations within the coding sequence of a gene that result in a premature termination codon . The presence of such mutations leads to the synthesis of a truncated protein unable to fulfill its normal function . Over the last ten years , treatment strategies have emerged based on the use of molecules , such as aminoglycoside antibiotics ( gentamicin ) that facilitate the readthrough of premature termination codons , thus restoring the synthesis of a full-length protein . Such strategies have been tested for various genetic diseases , including Duchenne muscular dystrophy and cystic fibrosis . The readthrough level depends on the nature of the stop codon and the surrounding nucleotide context , but little was known of the rules governing readthrough level and response to aminoglycosides . In this study , we use a large set of nonsense mutations for an in-depth statistical analysis designed to decipher the element of the nucleotide context responsible for modulating readthrough levels . We analyse the impact of the six nucleotides upstream and downstream from the stop codon . We demonstrate that the presence of a uracil residue immediately upstream the stop codon is associated with a stronger response to gentamicin treatment than the presence of any of the other three nucleotides .
|
[
"Abstract",
"Introduction",
"Results/Discussion",
"Materials",
"and",
"Methods"
] |
[
"mathematics",
"gene",
"expression",
"genetics",
"statistics",
"molecular",
"genetics",
"biology",
"human",
"genetics",
"genetics",
"and",
"genomics"
] |
2012
|
Statistical Analysis of Readthrough Levels for Nonsense Mutations in Mammalian Cells Reveals a Major Determinant of Response to Gentamicin
|
Chd proteins are ATP–dependent chromatin remodeling enzymes implicated in biological functions from transcriptional elongation to control of pluripotency . Previous studies of the Chd1 subclass of these proteins have implicated them in diverse roles in gene expression including functions during initiation , elongation , and termination . Furthermore , some evidence has suggested a role for Chd1 in replication-independent histone exchange or assembly . Here , we examine roles of Chd1 in replication-independent dynamics of histone H3 in both Drosophila and yeast . We find evidence of a role for Chd1 in H3 dynamics in both organisms . Using genome-wide ChIP-on-chip analysis , we find that Chd1 influences histone turnover at the 5′ and 3′ ends of genes , accelerating H3 replacement at the 5′ ends of genes while protecting the 3′ ends of genes from excessive H3 turnover . Although consistent with a direct role for Chd1 in exchange , these results may indicate that Chd1 stabilizes nucleosomes perturbed by transcription . Curiously , we observe a strong effect of gene length on Chd1's effects on H3 turnover . Finally , we show that Chd1 also affects histone modification patterns over genes , likely as a consequence of its effects on histone replacement . Taken together , our results emphasize a role for Chd1 in histone replacement in both budding yeast and Drosophila melanogaster , and surprisingly they show that the major effects of Chd1 on turnover occur at the 3′ ends of genes .
Eukaryotic genomes are packaged as chromatin , whose fundamental repeating subunit , the nucleosome , is composed of 147 bp of DNA wrapped 1 . 7 times around an octameric histone core . Nucleosomes may interact with each other to form higher-order levels of chromatin packaging necessary to compact an entire genome within a nucleus . This genome packaging strategy leads to a dominant theme in eukaryotic gene regulation: nucleosomes tend to repress gene expression , and a large array of gene regulatory mechanisms in eukaryotes operate by strengthening or weakening the repressive effects of nucleosomes on gene expression [1] . Genome-wide nucleosome mapping studies indicate that although the majority of a eukaryotic genome is typically covered with regularly spaced nucleosomes , nucleosome depleted or nucleosome free regions are frequently found over promoters and at the 3′ ends of genes ( reviewed in [2] ) . Although these studies give a fixed snapshot of chromatin organization , other analyses indicate that chromatin is dynamic . Studies in which histones were pulse-labeled with radioisotopes or tagged with GFP demonstrated that histones can be actively exchanged on chromatin , even in the absence of DNA replication [3] , [4] . More recent work has utilized induction of epitope-tagged alleles of histones in G1-arrested yeast cells followed by chromatin immunoprecipitation to examine histone H3 dynamics genome-wide [5] , [6] . These studies show that histone H3 exchanges at a high rate on promoters and in other intergenic regions such as downstream of the 3′ ends of genes . With the exception of highly-transcribed genes , the bodies of genes , even those that are transcribed at moderate rates , exhibit much lower H3 exchange rates . Although nucleosomes over transcribed genes appear to be relatively stable in vivo , nucleosomes form a strong barrier to elongating RNA polymerase II ( RNA Pol2 ) in vitro [7] . Thus , it is likely that accessory factors assist in transcription elongation to alleviate this barrier . These factors may promote the temporary disassembly or displacement of nucleosomes permitting the passage of elongating RNA Pol2 , and furthermore , they may assist in nucleosome ( re ) assembly after polymerases have passed . A wide variety of factors have been implicated in the dynamics and maintenance of chromatin structure over transcribed sequences . These include ATP-dependent chromatin remodeling enzymes , enzymes that post-translationally modify histones , histone chaperones and transcription elongation factors [8] . Interestingly , mutations affecting a number of these factors cause a cryptic transcription initiation phenotype , in which disruption of chromatin in the body of genes leads to activation of internal , normally quiescent promoters [9] . One factor implicated in the regulation of transcribed chromatin is the ATP-dependent chromatin remodeling enzyme Chd1 . Chd1 is the founding member of a family of highly conserved chromatin remodeling enzymes found throughout eukaryotes [10] . Although budding yeast only express a single Chd1 protein , at least 9 CHD family proteins are expressed in humans . Mammalian CHD family members have been implicated in diverse roles including promotion of normal organismal development , and the maintenance of pluripotency and prevention of heterochromatin formation in mouse embryonic stem cells [10] . In addition , mutations in CHD protein genes are implicated in several human cancers and CHARGE syndrome , which is characterized by a phenotypically heterogeneous set of developmental defects [10] , [11] . CHD proteins typically have a pair of N-terminal chromodomains , a central Snf2/Swi2 type helicase domain and a C-terminal domain that mediates DNA or nucleosome binding [10] . The chromodomains of human Chd1 bind histone H3 tails methylated at lysine 4 ( H3K4me ) suggesting a mechanism for recruitment [12] , [13] . However , yeast Chd1 does not bind H3K4-methylated tails [13] , and in Drosophila melanogaster , the chromodomains do not play an important role in its localization to chromatin [14] . Recent structural and biochemical studies suggest that rather than mediating chromatin localization , the chromodomains may regulate enzyme activity [15] . In vitro assays show that Chd1 has the ability to assemble , remodel , slide and promote regular spacing of nucleosomes [16]–[18] . Chromatin immunoprecipitation in budding and fission yeast , and immunostaining of Drosophila polytene chromosomes show that Chd1 associates with both promoters and transcribed regions of active genes [19]–[23] . Consistent with its localization on genes , genetic studies in yeast have implicated Chd1 in the regulation of transcription initiation , elongation and termination [22] , [24]–[28] . Although Chd1 can be purified as a monomer , its association with several complexes that regulate initiation and elongation , which include mediator , FACT , the Paf1 complex , SAGA and SLIK , provides further support to these conclusions [22] , [29]–[33] . Chd1 also associates with histone chaperones Nap1 in fission yeast , and HirA , a histone chaperone for histone H3 . 3 , in fruit flies [19] , [34] . Several studies suggest mechanisms for how Chd1's biochemical activity may relate to these biological functions . Chd1 can promote transcription and catalyze activator dependent , promoter specific nucleosome remodeling in vitro [35] , [36] . Furthermore , in Schizosaccharomyces pombe , Chd1 ( Hrp1 ) acts at a subset of promoters to disassemble nucleosomes close to the transcription initiation site [19] . In Drosophila , following fertilization of an egg , sperm chromatin is decondensed , protamines are removed and replaced with nucleosomes whose only form of histone H3 is the replication-independent variant H3 . 3 [37] . Interestingly , in chd1 mutants , H3 . 3 levels in decondensing sperm chromatin are greatly reduced and unevenly distributed , suggesting a role for Chd1 in the replication-independent assembly or distribution of H3 . 3 nucleosomes [34] , [38] . A recent high-resolution genome-wide nucleosome mapping study in budding yeast points to an in vivo role for Chd1's nucleosome remodeling activity . Nucleosomes are typically regularly positioned over genes in wild type yeast cells [39] . However , in a chd1Δ mutant , this positioning is largely lost over gene bodies [40] . Specifically , nucleosome free regions at the 5′ and 3′ ends of genes and the first ( +1 ) nucleosome over the transcribed region were minimally affected by loss of Chd1 , but downstream nucleosomes ( particularly those starting at the +3 position ) were dramatically delocalized in chd1Δ yeast cells . Curiously , micrococcal nuclease digestion patterns of bulk chromatin are not affected in a chd1 mutant , suggesting that Chd1 affects the positioning of nucleosome arrays primarily over the transcribed body of genes , rather that the precise spacing between any given pair of nucleosomes [40] , [41] . Although chd1 mutations have modest effects on gene expression in yeast , and are virtually indistinguishable from wild type strains in phenotypic assays , they do cause a cryptic initiation phenotype , consistent with the loss of nucleosome organization over the body of genes [9] , [28] , [42] , [43] . Although these data clearly demonstrate a role for Chd1 in nucleosome positioning in vivo , the mechanism underlying its in vivo function and its relationship to transcription remains unclear . In this study , we examine the role ( s ) of Chd1 in governing the replication-independent exchange of newly-expressed histone H3 onto chromatin in budding yeast and Drosophila using genome-wide methodologies . Chd1 mutants have dramatic defects in the localization of the replication-independent histone variant H3 . 3 in flies , while in Saccharomyces cerevisiae , chd1Δ mutants exhibit dramatic defects in H3 turnover in coding regions . Surprisingly , Chd1 predominantly affects histone H3 exchange at the 3′ ends of coding regions , and this effect on turnover depends on gene length – H3 turnover at 3′ ends is fairly concordant between wild type and chd1Δ strains for genes 1 kb and shorter , whereas Chd1 appears to specifically stabilize nucleosomes over the 3′ ends of longer genes . Finally , we show that loss of Chd1 globally alters histone modification patterns related to active transcription , with H3K36me3 in particular shifting in concert with the changed patterns of H3 replacement . Together , our results show that Chd1 plays a key role in histone H3 dynamics , and surprisingly , that yeast Chd1's influence on H3 dynamics is most apparent at the 3′ ends of genes .
Previously , Fyodorov and colleagues examined the distribution of epitope-tagged , full length H3 . 3 in the Drosophila syncytial blastoderm and only observed a modest defect in H3 . 3 distribution in chd1 null mutants [34] . Because the H3 . 3 N-terminal tail , which is required for replication-dependent assembly of H3 . 3 [44] , was intact in this experiment , we reasoned that any defect in replication-independent assembly of the tagged H3 . 3 might have been obscured . To reassess Chd1's role in replication-independent deposition of H3 . 3 , we imaged GFP-tagged histone H3 . 3core in live salivary glands from chd1 mutant larvae . We utilized a transgenic fly expressing an AB1-GAL4 driver and a Gal inducible histone H3 . 3core-GFP [44] . Because the H3 . 3core protein encoded by the transgene lacks the N-terminal tail , it is only incorporated into chromatin via the replication-independent pathway [44] . In an otherwise wild type background , H3 . 3core-GFP was deposited into the polytene chromosome arms of salivary glands ( Figure 1A ) . In some cases , we also observed a nucleoplasmic GFP signal in which the entire nucleus , including non-chromosomal territories , exhibited a strong GFP signal , although a chromosomal banding pattern was still evident . In flies that were heterozygous or homozygous for chd15 , a null allele of Chd1 [21] , we observed salivary gland nuclei with GFP signals similar to those of wild type , i . e . chromosomal or broad nucleoplasmic GFP fluorescence . However , we also observed nuclei with a novel , “non-chromosomal” phenotype where the polytene arms appear almost devoid of GFP signal and a substantial nuclear , non-chromosomal H3 . 3core-GFP signal was still apparent ( Figure 1B , 1C ) . We determined the relative frequencies of these phenotypes in wild type and mutant flies by blind scoring , and observed that the predominant chromosomal fluorescence pattern observed in wild type cells declined dramatically in chd1 mutants , whereas the nucleoplasmic and non-chromosomal patterns increased in frequency ( Figure 1D ) . We observed similar phenotypes when we repeated these experiments with independently derived chd15 flies using a different balancer chromosome ( data not shown ) . These results do not appear to be due to any peculiarity of the AB1-GAL4 driver as we observed similar fluorescence patterns when we used sgsGAL4 and eyelessGAL4 drivers ( data not shown ) . Furthermore , we did not observe obvious differences in the strength of H3 . 3core-GFP signals between flies with the three analyzed genotypes ( wild type , +/chd15 heterozygous and chd15/chd15 homozygous ) , nor between nuclei with the three observed staining patterns ( chromosomal , non-chromosomal and nucleoplasmic ) ( Figure S1 ) , suggesting that the observed localization patterns were not due to differences in H3 . 3core-GFP expression . Rather , we favor the idea that the variability observed here reflects perdurance of maternally contributed Chd1 , which has been observed previously [21] . Immunostaining of fixed polytene chromosomes similarly revealed a reduction of H3 . 3core-GFP on chromosomes derived from chd15 mutant larvae , while levels of full length H3 . 3-GFP were not affected by loss of Chd1 ( Figure S2 ) , consistent with the ability of full length H3 . 3 to incorporate through both replication-dependent and –independent pathways . Consistent with our observations in the chd15 mutants , we observed decreased association of H3 . 3core-GFP with polytene chromosomes when we knocked down Chd1 levels with either of two RNAi constructs ( Figure S2 and data not shown ) . Overall , these data are consistent with the possibility that Chd1 may contribute to replication-independent assembly of H3 . 3 containing nucleosomes . To further examine roles of Chd1 in nucleosome dynamics in vivo , we turned to budding yeast . To test the idea that Chd1 may modulate replication-independent nucleosome assembly or dynamics , we took advantage of the observation that the yeast H3 N-terminal tail is important for normal chromatin structure [45] . Reasoning that the H3 N-terminal tail deletion mutation likely interferes with replication-dependent assembly of H3 , as is the case in Drosophila and Physarum polycephalum , [44] , [46] , we predicted that loss of this function would sensitize cells to defects in other chromatin assembly or maintenance pathways , we used a plasmid shuffle strategy to create CHD1+ and chd1Δ yeast strains expressing either wild type histone H3 ( H3WT ) or a histone H3 N-terminal deletion mutation , H3Δ4-30 . Consistent with prior observations , the chd1Δ H3WT strain grew indistinguishably from wild type cells , and the CHD1 H3Δ4-30 strain exhibited a moderate growth defect ( Figure 2 ) . Interestingly , the chd1Δ H3Δ4-30 double mutant grew much more poorly than the CHD1 H3Δ4-30 single mutant , indicating that Chd1 and the N-terminal tail of H3 share a redundant function . In contrast to other model organisms , the budding yeast genome expresses only a single non-centromeric form of histone H3 . However , the major histone H3/H4 chaperones , including the H3 . 3 chaperone HirA , are conserved , suggesting that yeast retain distinctive replication dependent and independent chromatin assembly pathways [47] . We have obtained data consistent with this idea in a screen for genetic suppressors of a cold-sensitive allele of transcription elongation factor SPT5 . Among these suppressors were mutations in CHD1 , mutations in the H3K4 and H3K36 histone methyltransferases SET1 and SET2 , histone H3K4 and H3K36 substitutions , and mutations in members of the RPD3S histone deacetylase complex . Further characterization of these suppressors led us to propose that they act by lowering the chromatin barrier to efficient transcription elongation [28] . Given the observations described above , we recently screened a randomly mutagenized plasmid library for histone H3 mutations that suppress spt5Cs- ( to be described in detail elsewhere ) . Among the suppressor mutations obtained in that screen , we isolated a mutation , H3-S87P/G90S , which simultaneously alters two of the four residues that distinguish histone H3 . 1 from H3 . 3 in other eukaryotes . Yeast expressing the S87P/G90S form of histone H3 from the normal HHT2 locus are viable , indicating that this mutation is unlikely to strongly perturb replication coupled chromatin assembly . As with several other of the mutations that suppress spt5Cs- ( e . g . , H3K36R , set2 , mutations affecting Rpd3s ) , the H3-S87P/G90S mutant caused cryptic initiation of transcription ( Figure S3 ) . We therefore examined genetic interactions between the H3-S87P/G90P , chd1Δ and the H3Δ4-30 mutations using the plasmid shuffle assay described above ( Figure 2 ) . Interestingly , the chd1Δ H3-S87P/G90S double mutant exhibited no new mutant phenotypes , whereas combining H3-S87P/G90S with the H3Δ4-30 deletion resulted in a very poor growth phenotype and the chd1Δ H3-S87P/G90S H3Δ4-30 triple mutation showed an even more severe growth defect . Thus , like Chd1 , residues 87 and 90 of histone H3 function redundantly with the H3 N-terminal tail . It is tempting to argue that these data indicate that Chd1 interacts with histone H3 via a surface defined by residues 87 and 90 . However , the fact that the phenotype of the chd1Δ H3-S87P/G90S H3Δ4-30 triple mutant is more severe than that of the chd1Δ H3Δ4-30 double mutant suggests that H3 residues S87 and G90 may retain functions that are redundant with the H3 tail , even when Chd1 is absent . The data presented above suggest that Chd1 affects replication independent dynamics of histone H3 . To test this idea directly in budding yeast , we used a yeast strain carrying galactose-inducible Flag-tagged H3 , coupled with chromatin immunoprecipitation and tiling microarray ( ChIP on chip ) analysis , to follow the incorporation of newly-synthesized H3 genome-wide in cells arrested in the cell cycle [6] . Briefly , wild type or chd1Δ yeast strains are arrested in G1 phase using alpha factor , then Flag-H3 is induced with galactose , and after 60 minutes Flag-H3 and total H3-associated DNA are subject to ChIP enrichment and competitively hybridized on ∼250 bp resolution tiling microarrays . Resulting Flag/total H3 ratios provide locus-specific estimates of H3 turnover rates . Figure 3A shows a “metagene” analysis of H3 turnover in 3 biological replicate samples for wild type ( blue ) and chd1Δ ( red ) strains . The wild type profile recapitulates previous results from multiple labs [5] , [6] , [48] – H3 replacement is highest over promoters and at the 5′ ends of genes , with coding regions being remarkably protected from H3 replacement , and modest levels of turnover being seen at the 3′ ends of genes . Conversely , chd1Δ mutants exhibit H3 turnover patterns in which genes appear to effectively reverse polarity . Turnover is still lowest over coding regions , but the trough of minimal turnover has shifted 5′ along coding regions . Promoter and 5′ turnover are slower in chd1Δ cells , whereas maximal H3 replacement is instead observed at the 3′ ends of genes . This behavior is highly unusual , as several published [5] , [6] , [49] , [50] and a large number of unpublished ( OJR , unpublished data ) mutants exhibit quite distinct turnover defects . We confirmed the increased 3′ H3 replacement at two model genes ( Figure S4 ) using an entirely independent assay for histone replacement based on Cre-mediated recombination of C-terminal H3 epitope tags [51]–[53] . As a separate visualization , Figure 3B shows the average H3 turnover for various classes of genomic elements [6] , [54] . Even though a previous microarray analysis showed only a very modest effect of chd1Δ on transcription [18] , we considered the possibility that the altered H3 turnover in chd1 cells could be due to a large shift in cellular transcription . However , we observed strong concordance of ChIP on chip of RNA Pol2 signals for wild type and chd1Δ cells ( Figure S5 ) . Moreover , as noted below , Chd1's effects on H3 replacement are strongly gene length-dependent , but we find no correlation between mRNA abundance changes and gene length or transcription frequency ( Figure S6 ) . Thus , Chd1's effects on turnover are not secondary effects of altered transcription . We sought to understand what factors might contribute to Chd1 recruitment or function at gene ends . To this end , we first examined the genes with the greatest changes in H3 replacement at their 3′ ends in chd1Δ mutants . Notably , we observed that the genes with the greatest changes in 3′ end H3 turnover were among the longest ( >3 kb ) genes in budding yeast . We therefore systematically analyzed the effects of gene length on Chd1's role in H3 replacement . Figure 4 shows H3 turnover levels for wild type and chd1Δ yeast cells at gene ends ( the first and last 500 bp of coding regions ) as a function of gene length . At both gene ends there is strong length dependence for H3 turnover in wild type yeast cells , with turnover decreasing as a function of gene length . Notably , for both 5′ end and 3′ end H3 turnover , Chd1's effect on H3 turnover was greatest at unusually long genes . In addition , we found that Chd1's effect on 3′ turnover was greater at highly transcribed genes ( Figure S7 ) . The length dependence for 3′ end H3 replacement ( Figure 4B ) is particularly remarkable – H3 turnover is nearly identical in wild type and chd1Δ strains for genes of up to roughly 1 kb in length , at which point 3′ end turnover continues to decrease with gene length in wild type cells but stays essentially constant in chd1Δ cells . In other words , the role of Chd1 in wild type cells seems to be to help stabilize nucleosomes at the 3′ ends of genes over 1 kb in length . Chd1's effects on H3 turnover are greatest at genomic loci that are enriched in H3K36me3 or H3K4me3 modified nucleosomes [55] , and chd1 mutants exhibit synthetic genetic interactions with the H3K4 and H3K36 methyltransferases Set1 and Set2 [56] , [57] . We therefore determined if chd1Δ mutants affect histone modification patterns by genome-wide mapping of H3K4me3 and H3K36me3 in wild type and chd1Δ yeast cells . Crosslinked chromatin from these two strains was digested with micrococcal nuclease , immunoprecipitated with H3K4me3 or H3K36me3 antisera and competitively hybridized to microarrays with micrococcal nuclease digested input DNA . Figure 5 shows average H3K4me3 and H3K36me3 patterns in chd1Δ cells . On average , H3K4me3 patterns were minimally affected by loss of Chd1 , although we noticed a subtle increase in H3K4me3 at the 3′ ends of many genes . This may be a consequence of the fact that chd1Δ mutants show increased transcription from “cryptic” internal promoters [9] , [28] . Interestingly , the gain in H3K4me3 at the 3′ ends of genes was greatest at longer genes ( Figure S8 ) , which also exhibited the greatest defects in H3 turnover . More dramatically , H3K36me3 patterns were extensively altered in chd1Δ cells , with loss of H3K36me3 at the 3′ ends of genes and a shift in the H3K36me3 peak towards the 5′ ends of genes . Consistent with the loss of H3K36me3 at the 3′ ends of genes , we previously observed increased H3K9/K14 acetylation at the 3′ ends of several genes in a chd1Δ mutant [28] , as would be expected since reduced H3K36me3 results in reduced recruitment or activity of the Rpd3S deacetylase complex [27] , [58] , [59] . In our prior study , we did not observe any significant change in total levels of H3K4me3 or H3K36me3 in a chd1 mutant [28] . As H3K36me3 typically anticorrelates with H3 turnover [6] , we hypothesize that the altered H3K36me3 profile observed here is a consequence of Chd1's effects on H3 turnover – increased H3 turnover at 3′ ends of genes likely results in loss of H3K36me3 at these regions . Consistent with this hypothesis , we found that loss of 3′ H3K36me3 was greatest at longer genes ( Figure S8 ) .
We present evidence that Chd1 modulates replication-independent turnover of histone H3 in both Drosophila and budding yeast . Chd1's effects on H3 turnover are greatest at genomic loci that normally coincide with peaks of H3K4me3 and H3K36me3 modified nucleosomes . This observation is consistent with prior reports that Chd1 reduces nucleosome density at promoters , can catalyze activator-dependent nucleosome removal and promote transcription in vitro , and that it modulates the efficiency of transcription termination [19] , [24] , [36] . Chd1's effects on H3 turnover may reflect a direct role in histone eviction or deposition during replication-independent histone exchange , consistent with its ability to catalyze ATP-dependent assembly of nucleosomes in vitro , or it could reflect a role for Chd1 in stabilization of pre-existing nucleosomes . Here , we observed that the predominant effect of Chd1 on H3 turnover in budding yeast was to repress turnover over the 3′ ends of genes . While we do not yet understand the mechanism underlying this observation , we favor the idea that Chd1 acts upon nucleosomes that have been perturbed by elongating RNA polymerase II , restoring them to their normal structures or positions and thereby stabilizing them . Importantly , we do not favor the alternative model , that Chd1's effects on chromatin are secondary to perturbation of transcription; we and others do not observe significant alterations of gene expression in chd1 mutants in yeast ( Figure S6 and [21] ) and Pol II phospho-Ser2 staining of Drosophila polytene chromosomes is normal in chd1 mutants [21] . Chd1's effects on H3 turnover at the 3′ end of genes depended strongly upon gene length ( Figure 4 ) , and was also correlated with transcription rate ( Figure S7 ) . Given the model above , it is possible that in the absence of Chd1 , perturbation of nucleosome positioning by transcription complexes increases with gene length and nucleosome number . Alternatively , Chd1's function may relate to supercoiling changes driven by transcription . To test this we have preliminarily investigated whether additional deletion of the major topoisomerase Top1 affects the chromatin changes observed in chd1Δ yeast mutants . However , we have not observed any suppression of the chd1Δ turnover phenotype in chd1Δtop1Δ double mutants ( not shown ) . Thus , at present we have no additional evidence that supercoiling per se mediates the length dependence of Chd1 on H3 turnover , although given the ability of other topoisomerases to compensate for loss of Top1 we still consider this an appealing hypothesis . Previous results show that Chd1 has dramatic effects on nucleosome positioning over coding regions [40] . Our results extend this characterization by showing that Chd1 also has dramatic effects on H3 turnover over coding regions , raising the question of whether these two roles for Chd1 in chromatin structure are related . In other words , does Chd1's effect on H3 replacement follow from its role in establishing wild type nucleosome positions , or vice versa ? We have no evidence for either possibility , but note that our prior genetic analyses suggest that chd1Δ mutations lower the nucleosomal barrier to RNA Pol2 elongation [22] , [28] . Thus , we speculate that disorganized nucleosomes in chd1Δ mutants could be unusually susceptible to eviction by RNA Pol2 . This model is consistent with a recent suggestion that elongating polymerases could cause collisions and eviction of adjacent nucleosomes if they are spaced inappropriately [60] . However , arguing against this are observations that nucleosome ladders are little affected in chd1 mutants [40] , [41] . Future studies will be needed to address these mechanistic questions . Taken together , our results identify an evolutionarily conserved role for Chd1 in histone turnover in yeast and flies . Most surprising is our finding that the major site of Chd1 function appears to be at the 3′ ends of genes , suggesting that this enzyme may be recruited or regulated by 3′ histone marks such as H3K36me3 . Finally , we find that Chd1 largely affects H3 turnover over longer coding regions , raising the question of whether resolving superhelical tension could be a key role for Chd1 in maintaining wild type chromatin architecture .
Flies were raised on cornmeal , agar , yeast , and molasses medium , supplemented with methyl paraben and propionic acid . To drive the P[UHS-H3 . 3core-GFP] transgene [44] , [61] in the salivary gland , flies were crossed to P{GawB} AB1-Gal4 flies ( Bloomington Stock Center ) . Mutant chd15 flies were described previously [21] . All crosses were carried out at 18°C . Live analysis of polytene chromosome phenotypes was performed as described previously [62] . To analyze the effect of chd15 on H3 . 3core-GFP incorporation , chd15 b c sp/BcGla; P[UHS-H3 . 3core-GFP]/TM6B Tb Hu flies were crossed to chd15 b c sp/BcGla; P{GawB}AB1/TM6B Tb Hu flies at 18°C . Flies with chd15 balanced by CyO Kr-GFP instead of BcGla were also analyzed and yielded similar results . Salivary glands were dissected and imaged from heterozygous and homozygous chd15 third instar larvae . For control nuclei , P{GawB} AB1-Gal4 flies were crossed to P[UHS-H3 . 3core-GFP]/TM6B Tb Hu flies . H3 . 3core-GFP expression was quantitated by calculating sum pixel intensity in polytene nuclei using the Volocity software package as described previously [62] . Polytene chromosomes were prepared and fixed as described [63] and immunostained using primary antibodies directed against CHD1 ( [21] , 1∶300 dilution ) , H5 anti-RNA polymerase II ( specific for the Ser 2-phosphorylated form of Pol II CTD , Covance; 1∶50 dilution ) , and the JL-8 anti-GFP ( Clontech , 1∶300 dilution ) . Secondary antibodies donkey anti-rabbit IgG-Cy3 , donkey anti-mouse IgM-Cy2 , and donkey anti-mouse IgG Fc2a-DyLight 649 ( Jackson ImmunoResearch Laboratories , 1∶200 dilutions ) were tested with each individual primary antibody to ensure specificity . Images were examined on an Olympus 1X81 inverted fluorescence microscope and acquired using Image-Pro6 . 3 . Control and mutant chromosomes were photographed using identical exposure times , and images were processed identically in Adobe Photoshop CS3 . All S . cerevisiae strains used in this study ( see Table S1 ) were constructed by standard procedures , are isogenic to S288c and are GAL2+ [64] . Yeast media was made as described previously [65] . Plasmids used in this study are described in Table S2 . Plasmid pJH18-A06 was obtained by random PCR mutagenesis ( GAH , TKQ and Araceli Ortiz unpublished ) . pJH18-Δ4-30 , S87P/G90S was created by site-directed mutagenesis of pJH18-A06 . PGAL-H4-FlagH3 contains a KpnI-NotI fragment carrying pGAL-driven Flag-H3 from plasmid MDB61 [50] , in pRS416 . Strains transformed with pGAL-H4-FlagH3 were grown to ∼1 . 2×107 cells/ml in SC-Ura media with raffinose as the carbon source . Cells were G1 arrested with alpha factor and Flag-H3/H4WT expression was induced by addition of galactose ( 2% final concentration ) . ChIP assays were preformed as described previously [66] . 60 minutes after addition of galactose , cells were crosslinked with 1% formaldehyde for 15 min , disrupted by bead beating and chromatin was sonicated using a Diagenode Bioruptor to obtain an average size of 500 bp . Chromatin was immunoprecipitated using 40 µl ( 1∶2 slurry ) Anti-Flag M2 Affinity gel ( A2220; Sigma ) or 1 µg of a rabbit polyclonal antibody against the C-terminus of H3 ( ab1791; Abcam ) . Chelex 100 resin ( BioRad ) was added to the immunoprecipitated material and Input-DNA samples , and the suspensions were placed at 100°C for 10 min to reverse crosslinks . Samples were treated with proteinase K and DNA was recovered . Initial characterization and confirmatory analyses of ChIP samples were performed by qPCR in a Corbett Life Science Rotor Gene 6000 machine using SYBR Green as the detection dye ( qPCR MasterMix Plus for SYBR Green , Eurogentec ) . The fold difference between immunoprecipitated material ( IP ) and total Input sample for each qPCR amplified region was calculated as described in [67] , following the formula IP/Input = ( 2InputCt - IPCt ) . H3 turnover rates were measured as the final ratio between Flag-tagged H3 and total H3 ( Flag-H3/Input vs total H3/Input ) . The sequences of oligonucleotides used in these PCR reactions are listed in Table S3 . The immunoprecipitated DNA was initially PCR amplified using random hexamer primers as described in [68] . The number of cycles used to amplify the samples was adjusted to between 28 and 37 so that there was equal amplification of DNA in the IP vs . Flag-tagged H3 and the IP vs . total H3 samples . Amplified DNA was visualized on a 1% agarose gel and checked for a visible smear of DNA between 500 and 1 . 2 kB . Amplified DNA from Flag-tagged H3 and total H3 ChIPs samples were labeled and competitively hybridized to tiling microarrays as described below . Wt and chd1Δ cells were grown to log phase and fixed with 1% formaldehyde . Cell pellets ( from 100 mL cells ) were resuspended in 8 . 8 ml Buffer Z ( 1 M sorbitol , 50 mM Tris-Cl pH 7 . 4 ) , with addition 6 . 5 µl of ß-ME ( 14 . 3 M , final conc . 10 mM ) and 350 µL of zymolyase solution ( 10 mg/ml in Buffer Z; Seikagaku America ) , and the cells were incubated at 30°C shaking at 220 rpm . After spinning at 4000× g , 10 min , 4°C , spheroplast pellets were resuspended in 600 µl NP-S buffer ( 0 . 5 mM spermidine , 1 mM ß-ME , 0 . 075% NP-40 , 50 mM NaCl , 10 mM Tris pH 7 . 4 , 5 mM MgCl2 , 1 mM CaCl2 ) per 100 ml cell culture equivalent . 25–40 units ( depending on yeast strain and cell density ) of micrococcal nuclease ( Worthington Biochemical ) were added and spheroplasts were incubated at 37°C for 20 minutes . The digestion was halted by shifting the reactions to 4°C and adding 0 . 5 M EDTA to a final concentration of 10 mM . All steps were done at 4°C unless otherwise indicated . For each aliquot , Buffer L ( 50 mM Hepes-KOH pH 7 . 5 , 140 mM NaCl , 1 mM EDTA , 1% Triton X-100 , 0 . 1% sodium deoxycholate ) components were added from concentrated stocks ( 10–20× ) for a total volume of 0 . 8 ml per aliquot . Each aliquot was rotated for 1 hour with 100 µl 50% Sepharose Protein A Fast-Flow bead slurry ( Sigma ) previously equilibrated in Buffer L . The beads were pelleted at 3000× g for 30 sec , and approximately 100 µl of the supernatant was set aside for the input sample . With the remainder , antibodies were added to each aliquot ( equivalent to 100 ml of cell culture ) in the following volumes: 10 µl anti-H3K36me3 ( Abcam polyclonal ) , or 7 µl anti-H3K4me3 ( Millipore monoclonal ) . Immunoprecipitation , washing , protein degradation , and DNA isolation were performed as previously described [69] . The samples were amplified , with a starting amount of up to 75 ng for ChIP samples , using the DNA linear amplification method described previously [54] . 2 . 5 µg of aRNA produced from the linear amplification were labeled via the amino-allyl method as described on www . microarrays . org . Labeled probes ( a mixture of Cy5 labeled input and Cy3 labeled ChIP'ed material ) were hybridized onto an Agilent yeast 4×44 whole genome array . The arrays were scanned at 5 micron resolution with the Agilent array scanner . Image analysis and data normalization were performed as previously described [54] . Microarray data have been deposited in GEO ( Accession #GSE38540 ) .
|
Nucleosomes prevent transcription by interfering with transcription factor binding at the beginning of genes and blocking elongating RNA polymerase II across the bodies of genes . To overcome this repression , regulatory proteins move , remove , or structurally alter nucleosomes , allowing the transcription machinery access to gene sequences . Over the body of a gene , it is important that nucleosome structure be restored after a polymerase has passed by; failure to do so may lead to activation of transcription from internal gene sequences . Interestingly , although nucleosomes constantly move on and off of promoters , they are relatively stable over the bodies of genes . Thus , the same nucleosomes that are removed to allow a polymerase to pass by must be reassembled in its wake . Here , we examine the role of an ATP–dependent chromatin remodeling protein , Chd1 , in regulating nucleosome dynamics . We find that Chd1 is important for exchange of the histone H3 in both yeast and Drosophila and that , surprisingly , while it promotes exchange of histones at the beginning of genes , it prevents exchange at the ends of genes . Finally , we show that Chd1 helps determine the characteristic pattern of chemical modifications of histone H3 found over actively transcribed gene sequences .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"chromosome",
"biology",
"gene",
"expression",
"genetics",
"epigenetics",
"biology",
"genomics",
"chromatin",
"genetics",
"and",
"genomics",
"histone",
"modification"
] |
2012
|
A Key Role for Chd1 in Histone H3 Dynamics at the 3′ Ends of Long Genes in Yeast
|
MHV68 is a murine gammaherpesvirus that infects laboratory mice and thus provides a tractable small animal model for characterizing critical aspects of gammaherpesvirus pathogenesis . Having evolved with their natural host , herpesviruses encode numerous gene products that are involved in modulating host immune responses to facilitate the establishment and maintenance of lifelong chronic infection . One such protein , MHV68 M1 , is a secreted protein that has no known homologs , but has been shown to play a critical role in controlling virus reactivation from latently infected macrophages . We have previous demonstrated that M1 drives the activation and expansion of Vβ4+ CD8+ T cells , which are thought to be involved in controlling MHV68 reactivation through the secretion of interferon gamma . The mechanism of action and regulation of M1 expression are poorly understood . To gain insights into the function of M1 , we set out to evaluate the site of expression and transcriptional regulation of the M1 gene . Here , using a recombinant virus expressing a fluorescent protein driven by the M1 gene promoter , we identify plasma cells as the major cell type expressing M1 at the peak of infection in the spleen . In addition , we show that M1 gene transcription is regulated by both the essential viral immediate-early transcriptional activator Rta and cellular interferon regulatory factor 4 ( IRF4 ) , which together potently synergize to drive M1 gene expression . Finally , we show that IRF4 , a cellular transcription factor essential for plasma cell differentiation , can directly interact with Rta . The latter observation raises the possibility that the interaction of Rta and IRF4 may be involved in regulating a number of viral and cellular genes during MHV68 reactivation linked to plasma cell differentiation .
MHV68 is a naturally occurring murid gammaherpesvirus that has significant genetic and functional homology to the human gammaherpesviruses Epstein-Barr virus ( EBV ) and Kaposi's sarcoma-associated herpesvirus ( KSHV ) . Among herpesviruses , there are a large number of genes involved in virus replication that are conserved – both in sequence and spatial arrangement in the viral genome . However , every herpesvirus , having co-evolved with its host during speciation , has acquired unique genes - many of which function to modulate and/or evade the host immune response . Coevolution of with their hosts has led to some divergence of host-pathogen interactions; however , unique genes may reveal homologous functions required for chronic infection of the host . One such gene is the MHV68 M1 , which is found in a cluster of unique genes at the left end of the MHV68 genome . Initial functional studies of M1 , utilizing an M1-null virus revealed a hyper-reactivation phenotype from latently infected peritoneal exudate cells ( PEC ) [1] . Subsequent studies found that this hyper-reactivation phenotype was strain specific – occurring in C57Bl/6 mice , but not Balb/c mice [2] . In addition to the strain specific reactivation phenotype , a strain specific expansion of Vβ4+CD8+ T cells had previously been observed in response to MHV68 infection [3] . This pronounced T cell expansion and activation is a hallmark of MHV68 infection in many inbred mouse strains and is observed in peripheral lymphoid organs , as well as the blood , reaching peak levels after the virus has established latency [3] , [4] . Notably , the Vβ4+CD8+ T cells remain elevated during the course of chronic MHV68 infection , and do not adopt an exhausted phenotype [3] . Analysis of M1-null mutants revealed that a functional M1 gene is required for the Vβ4+CD8+ T cell expansion [2] . Furthermore , M1 was shown to be a secreted protein capable of stimulating Vβ4+CD8+ T cells to produce IFNγ and TNFα [2] . These analyses suggested that M1 may exert control over MHV68 reactivation from peritoneal macrophages through the induction of IFNγ from Vβ4+CD8+ T cells [2] , this is supported by the observations that: ( i ) IFNγ−/− mice exhibit hyper-reactivation from PECS [5]; and ( ii ) the demonstration that IFNγ can suppress MHV68 replication in macrophages [2] , [6] , [7] . Early experiments to evaluate the expansion in thymectomized mice suggested that Vβ4+CD8+ T cells are maintained through continued stimulation by a stimulatory ligand , which is now known to be M1 [8] . Interestingly , B cells appear to play a critical role in the expansion of Vβ4+CD8+ T cells , as no expansion is observed upon MHV68 infection of mice lacking B cells [9] , [10] . Other studies provide some clues to the timing and site of M1 expression during MHV68 infection , where B220+ splenocytes at 14 days post-infection were found to be capable of stimulating Vβ4+CD8+ T cell hybridomas [11] . Though no homolog to M1 has been found in other gammaherpseviruses , HVS has been shown to encode a viral superantigen , immediate early gene ie14/vsag [12] . Like M1 , ie14/vsag , is not essential for viral replication; and interestingly , ie14/vsag expression is elevated in phorbol ester treated cells , indicating a link with viral reactivation . In EBV , structural protein gp350 , as well as latent membrane proteins LMP-1 and LMP-2A have been shown to activate expression of an endogenous human retroviral superantigen , HERV-K18 , which results in a Vβ13+ T cell expansion [13]–[15] . Due to limitations in study of non-human primate and human patients it has been difficult to assess the role of these superantigens and the consequence of their resulting T cell expansion . We are therefore left to speculate what benefit they provide to their host . Do they aid in infection or the establishment of latency ? Do they divert the immune response ? Are they involved in control of infection ? We hope that a better understanding of the expression and role of M1 in MHV68 infection may shed light into the conserved use of viral superantigens by gammaherpseviruses . Though numerous studies to define the transcriptional program of MHV68 in vitro have identified M1 as an early through late gene [16]–[18] relatively little is known about when and where M1 is expressed during infection . Furthermore , while a number of transcriptome based analyses have detected transcripts extending through the M1 locus during in vivo infection , much of this data relies on methods that are not strand specific and therefore not definitive [19]–[22] . Due to the dearth of information about M1expression in vivo , we set out to characterize M1 expression using a novel approach wherein a fluorescent reporter virus would allow detection of M1 promoter activity during infection . This approach led to the identification of splenic plasma cells as the primary cell type expressing M1 in vivo . Furthermore , factors regulating M1 transcription were previously uncharacterized . The current studies have elucidated key cis-elements and transcription factors controlling the expression of M1 in plasma cells . Overall , these findings provide insights into the role of M1-mediated regulation of MHV68 pathogenesis . Moreover , we reveal a novel and potentially conserved mechanism which controls the timing and site of viral gene expression in response to reactivation in the B cell .
To identify cellular reservoirs in which the M1 gene is expressed in vivo , we generated a series of recombinant viruses that express yellow fluorescent protein ( YFP ) to mark infected cells . For detection of M1 promoter activity , the M1 coding sequence was replaced with that encoding YFP , creating a M1 promoter-driven YFP mutant ( Figure 1A ) . This strategy allows detection of the cellular reservoirs in which M1 is expressed during infection . Additionally , two important controls were used: MHV68-YFP , in which the YFP transgene under the control of the HCMV IE promoter was cloned into a neutral locus in the viral genome ( efficiently marking MHV68 infected B cells and plasma cells ) [23]; and ( ii ) MHV68-M1st . YFP , which contains the M1 translational stop mutation ( M1-null virus ) in the context of the YFP transgene cloned into the neutral locus ( Figure 1B ) . As M1 has previously been identified as a non-essential for both virus replication and for the establishment of latency in vivo [24] , we did not anticipate that the M1pYFP recombinant would change the cellular reservoirs infected by MHV68 . However , to formally address this issue , we have included analyses of the MHV68-M1st . YFP virus – which like the M1pYFP lacks a functional M1 gene . Analysis of MHV68 infection of splenocytes at day 14 post-infection revealed robust marking of splenocytes by both the MHV68-YFP and MHV68-M1stYFP viruses ( Figure 2 ) . We have previously noted that there is significant mouse to mouse variation in the frequency of infected splenocytes for a given virus [25] , and have recently determined that this directly correlates with the frequency of the CD4+ T follicular helper ( TFH ) response [26] . For these analyses we observed on average ca . 0 . 5% and 1 . 0% of splenocytes were YFP+ for the MHV68-YFP and MHV68-M1stYFP viruses , respectively ( Figure 2 ) . The latter result confirms that M1 function is dispensable for the establishment of latency in splenocytes . In contrast , only ca . 0 . 04% of splenocytes were YFP+ with the M1pYFP virus , indicating that the M1 promoter is active in only ca . 5–10% of infected splenocytes . We have previously shown that the majority ( ca . 70–90% ) of virally infected B cells , as indicated by YFP expression , exhibit a germinal center phenotype [23] , [27] . Individual mice were assessed for YFP marking and , consistent with previous observations , we found a similar frequency of virus infected ( YFP+ ) B cells with a germinal center phenotype for mice infected with either the MHV68 . YFP or MHV68-M1st . YFP viruses , both showing an average of ca . 70% ( Figure 3 ) . These results further substantiate that a functional M1 gene is dispensable for establishment of MHV68 latency in B cells . In contrast , few infected germinal center B cells were marked by the M1pYFP virus ( an average of ca . 20% of YFP+ cells ) – indicating that the majority of M1 expressing cells do not have a germinal center B cell phenotype . Based on the ca . 10-fold lower frequency of splenocytes marked by the M1pYFP virus ( see Figure 2 ) , we estimate that M1 promoter activity is only detectable in ca . 5% of infected germinal center B cells . Based on these results it is clear that M1 is predominantly expressed in some other MHV68 infected cellular reservoir . The other major cell population in the spleen that is infected by MHV68 are plasma cells ( CD138hi , B220low ) [23] , [27] . During infection , virus infection ( YFP marking ) of splenic plasma cells reaches peak levels at day 14 post-infection ( ca . 10–20% of virus infected splenocytes ) and begins to wane by day 18 post-infection ( ca . 5–10% of virus infected splenocytes ) [23] . We observed marking of splenic plasma cells for both MHV68-YFP and MHV68-M1st . YFP infected mice at day 14 post-infection consistent with previous observations , with ca . 10% YFP+ cells exhibiting a plasma cell phenotype ( no significant difference between these 2 groups ) ( Figure 4 ) . Strikingly , when assessing YFP marking of the splenic plasma cell population by the M1pYFP virus , the vast majority of YFP+ cells exhibited a plasma cell phenotype ( on average >75% of YFP+ cells ) ( Figure 4 ) . Thus , this strongly argues that M1 gene expression is largely limited to the infected plasma cell population . Notably , MHV68 reactivation from latently infected splenocytes is tightly linked to plasma cell differentiation [27] , which suggests that M1 expression is coupled to virus reactivation from B cells . Finally , when considering the frequency of M1pYFP marked cells with the frequency of MHV68-YFP and MHV68-M1st . YFP marked splenic plasma cells , it appears that the majority of virus infected plasma cells express M1 . Having identified the reservoir where M1 is expressed in vivo , we sought to characterize the structure of the M1 transcript and to identify the M1 promoter . Rapid amplification of cDNA ends ( RACE ) was done to identify the transcript initiation and termination sites in two cell lines: ( i ) infected NIH3T12 fibroblasts; and ( ii ) reactivated A20-HE2 cells . A20-HE2 cells are a stable lymphoblast B cell line which carry the MHV68 genome where viral reactivation can be induced by tetradecanoylphorbol acetate ( TPA ) [28] . RNA and protein were collected from both cell lines and lytic gene expression was confirmed prior to analysis ( data not shown ) . Transcript analysis revealed four initiation sites and a single termination site from an unspliced transcript ( Figure 5A , 5B ) . Though all transcript initiation sites were found in infected 3T12 cells , only transcripts starting at bp 2003 and bp 2013 were detected from reactivated A20-HE2 cells . The sizes of the predicted unspliced M1 transcripts were confirmed by northern analyses of RNA prepared from: ( i ) TPA stimulated A20-HE2 cells ( a MHV68 latently infected B cells ) ; and ( ii ) MHV68 infected NIH 3T12 fibroblasts ( data not shown ) . To identify the regulatory elements controlling M1 gene expression we next set out to characterize the M1 promoter . Serial truncations of the putative M1 promoter region were cloned into a luciferase reporter vector and tested for promoter activity in a variety of cell lines . Notably , minimal activity was detected in the murine B cell lines A20 , WEHI , NSO , and BCL1-3B3 ( data not shown ) – perhaps consistent with the failure to observe significant M1 promoter-driven YFP activity in most splenic B cell populations with the MHV68-M1pYFP virus in mice . In addition , we failed to detect significant activity from these reporter constructs in the murine macrophage cell line RAW264 . 7 ( data not shown ) . However , when these reporter constructs were transfected into the P3X68Ag8 murine plasmacytoma cell line significant basal promoter activity was observed ( Figure 6 ) . Similar levels of M1 promoter-driven luciferase activity were observed for the longer M1 promoter constructs ( M1p/−1025 bp , M1p/−525 bp , and M1p/−245 bp ) , while truncation of sequences upstream of −100 bp significantly decreased activity ( Figure 6 ) . Activity was further decreased to near background levels when sequences upstream of −50 bp were deleted ( Figure 6 ) . The region upstream of the M1 transcription initiation sites was screened for the presence of candidate transcription factor binding sites [University of Pennsylvania Transcription Element Search System ( TESS ) ] . TESS and manual sequence analyses identified a number of candidate transcription factor binding sites for NFκB , GATA3 , IRF8/IRF4 , and RBPJκ . Because M1 promoter activity was detected in plasma cells in vivo , interferon regulatory factor 4 ( IRF4 ) , a transcription factor upregulated in plasma cells which plays a critical role in plasma cell differentiation as well as immunoglobulin class switch recombination ( [29]–[31] and reviewed in [32] ) , was of particular interest . To characterize IRF4 binding to the candidate IRF site in the M1 promoter , an electrophoretic mobility shift assay ( EMSA ) was carried out ( Figure 7A ) . EMSA was performed using nuclear extracts from P3X63Ag8 cells grown under normal conditions , along with a [32P]-labeled oligonucleotide probe containing the candidate M1 promoter IRF4 binding site . As expected we observed shifted complexes , which could be competed away using unlabeled double stranded DNA probes containing the M1p IRF4 binding site , but not with a competitor containing an IRF binding site mutation which has previously been shown to disrupt IRF8 binding with DNA [33] ( Figure 7A ) . Furthermore , binding of IRF4 was confirmed by supershift analysis using an antibody against IRF4 ( Figure 7A ) . This analysis was extended by generating M1 promoter-driven luciferase reporter constructs in which mutations were introduced into the IRF binding site . Two mutations in the core interferon response sequence , which have previously been shown to ablate IRF8 DNA:protein interaction [33] , were introduced into the M1 promoter . Notably , either mutation led to a significant loss in basal M1 promoter activity ( ca . 8-fold decrease in promoter activity ) ( Figure 7B ) . Several studies have established a link between gammaherpesvirus reactivation from latency and plasma cell differentiation [27] , [34]–[39] . Given that our data shows: ( i ) M1 promoter expression is detected from plasma cells during in vivo infection; ( ii ) basal M1 promoter activity requires a functional IRF4 site; and ( iii ) viral reactivation is linked with plasma cell differentiation , we set out to evaluate whether the M1 promoter is responsive to the MHV68 viral lytic transactivator Rta . Expressing increasing amounts of Rta with an M1 promoter-driven reporter construct in the P3X63Ag8 plasmacytoma cell line resulted in a dosage dependent increase in M1 promoter activity ( Figure 8A ) . Moreover , the ability of Rta to efficiently transactivate the M1 promoter in the P3X63Ag8 cell line was dependent on the presence of an intact IRF4 binding site ( Figure 8B ) . To further assess whether Rta functionally synergizes with IRF4 to activate the M1 promoter , we chose a cell line ( 293T cells ) which lacks expression of Rta and IRF4 . In 293T cells we observed that either factor alone led to very modest increase in M1 promoter activity ( ca . 5–10 fold ) ( Figure 8C ) . However , when the two factors were co-expressed there was a significant increase in promoter activity ( ca . 250-fold ) ( Figure 8C ) . Importantly , disruption of the IRF4 binding site dramatically impaired the ability of IRF4 and Rta to synergistically activate the M1 promoter ( Figure 8C ) . Based on the synergy between Rta and IRF4 in activating the M1 promoter , we assessed whether these factors can physically interact with each other . A co-immunoprecipitation was performed with cell lysates from transfected 293T cells . Immunoprecipitation with anti-IRF4 antibody , followed by anti-Flag detection of Rta , resulted in detection of a 90 kD band corresponding to Rta that was present only when Rta and IRF4 were co-expressed in 293T cells ( Figure 8D ) . Following detection of Rta the blot was stripped and probed for IRF4 to confirm expression of the 52 kD band corresponding to IRF4 . IRF4 was detected in whole cell lysates and immunoprecipitated samples containing IRF4 . The reciprocal blot using anti-flag for immunoprecipitation and anti-IRF4 for detection showed a 52 kD band corresponding to IRF4 . Additionally , Rta was detected from whole cell lysates and immunoprecipitated samples containing Rta . These results are consistent with a physical interaction between Rta and IRF4 that likely facilitates that observed synergy of these factors in activating M1 gene expression . Several investigators have identified Rta responsive elements in viral promoters for both KSHV and MHV68 [40]–[44] . To date , the known Rta responsive genes are either regulated through: ( i ) direct interaction of Rta and DNA through a core Rta binding sequence; or ( ii ) Rta DNA binding is facilitated through protein-protein interactions – in the case of KSHV Rta , through interaction with the cellular transcription factor RBPJκ ( reviewed in [45] ) . In MHV68 gene 57 promoter , it appears that both types of Rta response elements may be utilized – although a role for RBPJκ in MHV68 Rta activation has not been formally demonstrated [40] , [41] . Interestingly , neither of the binding sites identified in the MHV68 gene 57 promoter are present in the M1 promoter , suggesting a novel Rta interaction motif . To identify the Rta response element ( s ) in the M1 promoter , a series of promoter truncations were generated and tested in the P3X63Ag8 plasmacytoma cell line . A candidate Rta response element was identified by evaluating promoter constructs which lost the ability to be transactivated by Rta . Using this approach we identified a putative Rta response element between −82 and −72 bp in the M1 promoter ( Figure 9A ) . This 12 bp sequence ( 5′-GGTCAGAAGGCT-3′ ) failed to show homology to any known Rta response element identified in the gammaherpesvirus family . However , a screen of the MHV68 genome identified a number of candidate sites upstream of other MHV68 replication-associated genes which share significant homology with the core 5′-TCAGAAG-3′ sequence in the putative M1 promoter Rta response element ( Figure 9B ) . Mutations of the three most central residues of the predicted Rta response element ( see M1pRREm in Figure 9B ) resulted in an ca . 10-fold reduction in transactivation in the plasma cell line ( Figure 9C ) , as well as an ca . 6-fold reduction in Rta and IRF4 synergistic transactivation of the M1 promoter in 293T cells ( Figure 9D ) . With the identification of a novel Rta response element , we next wanted to evaluate whether this element was functional in other viral promoters that appear to contain this RRE ( see Fig . 9C ) . Reporter constructs for the putative promoter regions of the M2 gene ( encoding an adaptor protein involved in B cell signaling ) , ORF8 ( encoding glycoprotein B ) , ORF22 ( encoding glycoprotein H ) , ORF63 ( encoding a tegument protein ) , and ORF73 ( encoding the MHV68 Latency Associated Nuclear Antigen ( LANA ) homolog ) were generated . In addition , the gene 50 proximal , distal , and N4/N5 promoter constructs previously described in Wakeman et al . [46] were evaluated for response to Rta expression . We observed varying levels of promoter response , with the strongest responses from ORF50pp , ORF8p , ORF22p , ORF63p , intermediate responses from the M1p , ORF50dp and ORF50 N4/N5p , and weak responses from M2p and ORF73p ( Figure 10A ) . To further investigate the role of the Rta response element in the observed transactivation , we engineered the same three nucleotide mutation used in the M1p ( Figure 9B ) into the proximal ORF50 promoter ( Figure 10B ) . Notably , mutation of this sequence resulted in a 38-fold reduction in Rta transactivation ( Figure 10C ) . Notably , with the exception of the M1 promoter , for all the other reporter construct we failed to observe any synergistic activation by the co-expression of Rta and IRF4 ( data not shown ) .
Here we described the characterization of a recombinant MHV68 in which a gene encoding a fluorescent protein ( YFP ) has been introduced into the viral genome in place of a non-essential viral gene . This approach allows identification of the site and timing of viral gene expression in vivo for viral genes that are dispensable for replication and/or dissemination of virus . For viral genes that play an important role in either replication or dissemination , other approaches - such as the generation of fusion gene products - may be required . Information obtained from such studies can provide significant insights into viral gene function and their mode of action . In the case of M1 , these analyses led to identification of the predominant cellular reservoir in which M1 is expressed , and subsequent identification of transcription factors involved in regulating M1 gene transcription . Coppola et al . have previously demonstrated the ability of either B220+ cells , or T cell depleted splenocytes , isolated from MHV68 infected mouse spleen to stimulate Vβ4+ CD8+ T cell hybridomas [11] . However , they also found that B cell depleted splenocytes from MHV68 infected mice retained Vβ4+ CD8+ T cell stimulatory activity – which they interpreted as the presence of other non-B cells populations in the spleen that are infected by MHV68 . However , based on our findings that M1 expression is largely restricted to plasma cells , it seems unlikely that either the isolation or depletion of B220+ cells would efficiently capture or eliminate , respectively , all MHV68 infected plasma cells . As such , one would anticipate Vβ4+ CD8+ T cell stimulatory activity in both the enriched and depleted fractions . This interpretation is consistent with the complete failure to observe any expansion of Vβ4+ CD8+ T cells in MHV68 infected B cell-deficient mice [9] , [10] – even though we have previously shown robust MHV68 infection in the spleens of B cell-deficient mice under some experimental conditions ( intraperitoneal inoculation of virus ) in the absence of any detectable Vβ4+ CD8+ T cell expansion [10] , [47] . As we have previously shown , MHV68 reactivation in the spleen is tightly linked to plasma cell differentiation [27] . The observation that M1 is predominantly expressed in plasma cells thus suggested that M1 expression is linked to virus reactivation/replication . This was substantiated by demonstration that Rta can strongly transactivate the M1 promoter in a plasma cell line ( see Figure 8A ) . We propose that during infection , in response to viral reactivation and the transition from germinal center or memory B cell to plasma cell , M1 expression is activated by the synergistic effects of viral Rta and cellular IRF4 . M1 protein is secreted from infected plasma cells and , by an undefined mechanism , stimulates Vβ4+ CD8+ T cell activation and expansion . It is likely that M1 activates Vβ4+ CD8+ T cells via a mechanism similar to classic viral super-antigens [2] . Activation does not require classical MHC class I molecules [11] , [48] , but does require an intact M1 protein - we have previously shown that proteolytic digestion , or denaturation of recombinant M1 renders it unable to activate Vβ4+ CD8+ T cell hybridomas [2] . Vβ4+ CD8+ T cells have been shown to traffic throughout the body , and can be detected in the blood , spleen , liver , lung , and peritoneal cavity [2] , [3] , [8] . These cells show cytolytic activity [8] and adopt an effector memory phenotype where upon re-stimulation with recombinant M1 protein ex vivo they degranulate and produce INFγ and TNF α ( [2] , unpublished observations ) . As IFNγ has been shown to regulate MHV68 reactivation from macrophages in the peritoneum , but not reactivation from splenic B cells [6] , we would predict that the Vβ4+ CD8+ T cells traffic to sites in which infection is less tightly controlled , to suppress MHV68 reactivation through the secretion of IFNγ in a paracrine fashion . We find it noteworthy that MHV68 M1-null infected mice exhibit hyper-reactivation in the peritoneal cavity and persistent viral replication in the lung [1] , [2] , [49] , further underscoring the importance of M1 expression in controlling viral infection . Our findings demonstrate that the M1 promoter is regulated by MHV68 Rta , a viral transcription factor that is essential for induction of viral reactivation . Rta activation of the M1 promoter synergizes with IRF4 , a transcription factor that plays a critical role in both plasma cell differentiation and immunoglobulin class switch recombination . Furthermore , we show that this interaction is likely mediated through both DNA-protein interactions with the M1 promoter sequence , as well as protein-protein interactions between Rta and IRF4 . We propose that during MHV68 latency , the viral latency-associated gene product M2 is expressed in a sub-population of latently infected germinal center and memory B cells [21] leading to expression of high levels of IRF4 [50] . M2 appears to play an important role in virus reactivation from latency: ( i ) MHV68 M2 null mutants exhibit a profound reactivation defect from B cells , but not latently infected macrophages [51] , [52]; ( ii ) exogenous expression of M2 in primary B cells results in acquisition of a pre-plasma memory phenotype [53]; ( iii ) M2 can drive B cell differentiation of a B lymphoma cell line in vitro [27]; ( iv ) M2 is required for efficient immunoglobulin class switch in infected B cells in vivo [27]; and ( v ) plasma cells are the primary source of viral reactivation from the spleen [27] . Taken together these data suggest that MHV68 is capable of driving plasma cell differentiation , and concurrent with this differentiation , viral reactivation . As a result of this transition , the increased expression of the transcription factors Rta and IRF4 lead to induction of M1 expression in plasma cells ( Figure 11 ) . That M1 is responsive to viral Rta and cellular IRF4 highlights the importance of tightly regulated gene expression in response to host and viral cues . This promotes cell type specific expression coordinated with viral reactivation . Furthermore , the interaction with Rta and IRF4 suggests a conserved strategy for gene regulation in MHV68 , allowing for better control of Rta responsive gene expression . In fact numerous lytic genes in MHV68 appear to share the Rta response element identified in the M1 promoter ( Figure 9B ) . Though our efforts to find other viral genes that are similarly responsive to the concerted effects of Rta and IRF4 were unsuccessful , we find it attractive to speculate that the partnership of Rta and IRF4 or other cellular transcription factors may mediate their gene expression in a cell type specific manner . However , many of the genes we evaluated showed response to the novel Rta response element . Our analysis was limited to the putative promoter regions of ORF50 , ORF8 , ORF22 , ORF63 , ORF73 , and M2 . However , many of these genes play critical role in the biology of the virus , either as structural genes- ORF8 and ORF 22 are both surface glycoproteins , or as genes involved initiating infection- ORF63 is a tegument protein; so it is not surprising to find a significant response to Rta but lack of synergy with IRF4 . Furthermore , some of the candidate genes are known to be involved in viral latency , ORF73- or murine latency associated nuclear antigen ( mLANA ) a homolog of EBV and KSHV LANA , has many functions including: viral replication , episomal maintenance , transcriptional regulation , and dysregulation of cell cycle and cell division ( reviewed in [54] ) . M2 , a latency associate protein appears to play roles in both maintenance and establishment of latency , as well as in viral reactivation [50] , [53] , [55] . We therefore find it plausible that these genes would have less stringent requirements for cell specific expression , and that other unidentified genes , might be regulated by Rta and IRF4 . Additionally , due to the differing functions of these genes in MHV68 biology , we were not surprised that in our studies we found differing levels of Rta responsiveness . Future studies using genome wide analyses will be necessary to identify genes which are temporally regulated by viral and host factors including Rta and IRF4 . Our identification of a partnered interaction between Rta and IRF4 suggests a conserved method for regulating MHV68 viral gene expression . Moreover , this mechanism appears throughout the gammaherpesviruses family as several studies have shown that Rta is capable of binding DNA through orchestration of complex protein-protein interactions . In KSHV , kRta has been shown to directly interact with cellular Oct1 and RBPJκ to regulate the KSHV bZip promoter [43] . This interaction with RBPJκ is maintained through a tetrameric protein complex of kRta flanking RBPJκ , and is mediated through a core “CANT” DNA repeat element found in the Mta promoter sequence [42] . Notably , kRTA has also been found to interact with viral IRF4 ( vIRF4 ) , one of several viral IRF homologs encoded by KSHV which in the case of vIRF4 is involved in counteracting innate antiviral defenses mediated by interferons to regulate vIRF1 , vIRF4 , PAN , and ORF57 gene expression [44] . In summary , the data reported here defines the timing and location of M1 expression during in vivo infection using a recombinant reporter virus – demonstrating that M1 is predominantly expressed from plasma cells . Furthermore , M1 gene transcription in plasma cells is driven by the viral immediate-early Rta in conjunction with cellular IRF4 – which potently synergize with each other to activate M1 gene transcription . Whether other viral ( and perhaps cellular genes ) are co-regulated by Rta and IRF4 remains to be determined , and will be the topic of future work . However , we find it interesting to speculate that this might be an effective strategy to target viral replication-associated gene expression in plasma cells .
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health . The protocol was approved by the Emory University Institutional Animal Care and Use Committee and in accordance with established guidelines and policies at Emory University School of Medicine ( Protocol Number: YER-2002245-031416GN ) . Six to eight week old female C57Bl/6 mice were obtained through Jackson Laboratory ( Bar Harbor , ME ) and housed at Emory University in accordance with university guidelines . Prior to infection mice were sedated with isofluorane and intranasally infected with 5×105 pfu in 20 ul of DMEM . Cells were grown under normal conditions at 37°C with 5% CO2 . A20-HE2 cell were grown in complete RPMI-1640 ( supplemented with 10% FCS , 100 U/mL penicillin , 100 mg/mL streptomycin , 2 mM L-glutamine , and 50 mM β-mercaptoethanol ) ; P3X63Ag8 ( ATCC TIB-9 ) were grown in compete RPMI-1640 with the addition of 10 mM non-essential amino acids , 1 mM sodium pyruvate , and 10 mM HEPES; and 293T cells ( a generous gift from Dr . Edward Mocarski ) were grown in complete DMEM ( supplemented with 10% FCS , 100 U/mL penicillin , 100 mg/mL streptomycin , and 2 mM L-glutamine ) . To generate the M1 promoter driven YFP virus a 500 bp homology arm immediately upstream of M1 ORF was amplified with LFA_MluI_1521-1573 ( 5′-TCCCCAATGACGCCAAAGTCTAAGTCCCTGTACAGGCTTAACTTTTTTAGAAT-3′ ) and LFA_SpeI_2005-2022 ( 5′-GGTCGCCGCTGCTCAATG-3′ ) and cloned into pCR Blunt-eYFP vector ( a kind gift from Dr . Chris Collins ) using MluI and SpeI to generate pCR Blunt-eYFP M1 LFA Flank . Next a 495 bp homology arm immediately downstream of the M1 ORF was PCR amplified using RFA_NotI_3286-3307 ( 5′- GCCTGAATACATGTTTACTGGG-3′ ) and RFA_NsiI_3758-3780 ( 5′- AACCTACGCGGCCACTCAACAGA -3′ ) was cloned into pCR Blunt-eYFP using NotI and NsiI to create pCR Blunt-eYFP M1 LFA RFA flank . The eYFP flanked by left and right homology arms for the M1 locus was then PCR amplified to include BglII and NsiI restriction sites and was cloned into pGS284 . The resulting plasmid pGS284-eYFP M1 LFA RFA flank was then electroporated into λPir electro-competent bacteria for allelic exchange with WT MHV68 BAC in GS500 RecA+ Escherichia coli . To generate the M1st-eYFP virus , GS500 containing M1st . BAC [2] were crossed with λPir containing pGS284-XL9CD-CMV-YFP-F for allelic exchange . Following allelic exchange virus preparation was performed as previously described [23] . Single cell suspensions of splenocytes were prepared and resuspended in PBS supplemented with 2% fetal bovine serum . Samples were stained using standard procedures . Following initial FC receptor block ( CD16/32 ) , samples were stained with a master mix containing: CD138-PE , CD3e-PerCP , CD95-PE , GL7-APC , B220-APC-Cy7 , CD19-Pacific Blue . 1–2×106 events were recorded on BD LSRII flow cytometer and results were analyzed using FloJo software ( Tree Star Inc ) . A20-HE2 cells were stimulated with 20 ng/mL tetradecanoylphorbol acetate ( TPA ) for 48 hours prior to RNA isolation . NIH3T12 were infected with an MOI of 1 . 5 for 18 hours prior to RNA isolation . RNA was extracted using Trizol Reagent ( Invitrogen Life Technologies ) according to manufacturer's instructions and RNA concentration was determined . Prior to RACE analysis RT-PCR was performed to detect M1 and pol transcripts using primers described previously [56] . 5′ and 3′ RACE analysis was performed using GeneRacer Kit L1502-02 ( Invitrogen Life Technologies ) according to manufacturer's specifications . Gene specific primers were generated for detection of M1 transcript . For the first round of PCR M1ORF_Rd1_Fwd ( 5′-GGCCATTATGTGGACGTGAAGAGAATTGTAGGTAT-3′ ) was used to amplify the 3′ region and M1ORF_Rd1_Rvs ( 5′-CCTTGGTATCATCCTCAGGAAATGGGTAGGTTTCA-3′ ) was used to amplify the 5′ region . For the second round of PCR M1ORF_Rd2_Fwd ( 5′-GGAAAACTCTCCAGAGCTGCTGTCGTG GGGGATGAT-3′ ) was used to amplify the 3′ region and M1ORF_Rd2_Rvs ( 5′-GCCAGTGAGCTATGCTTTGGCCCAGTATGCAGGAA-3′ ) was used to amplify the 5′ region . M1 promoter luciferase constructs were cloned into pGL4 . 10 ( Promega ) using BglII and KpnI restriction sites . With the exception of the M1pIRF4mut1 , M1pIRF4mut2 , and G50ppRREm binding site mutants , inserts were generated by PCR amplification of regions upstream of the M1 ORF , using WT BAC DNA as template , with primers listed in table 1 . The M2 promoter construct was the generous gift from Shariya Terrell . To generate M1pIRF4mut1 and 2 Overlapping PCR was used to introduce IRF4/IRF8 binding site mutations into the 197 bp M1 promoter region corresponding to nt . 1960–1961 and nt . 1961–1963 in the viral for mutants 1 and 2 respectively . Amplification of a 118 bp left flaking arm was done using 197 bp forward primer ( table 1 ) and reverse primers: ( 5′-TCTTTCTTGGTGTGTTCACTTCTAAACATG-3′ ) and ( 5′-TCTTTCTTGGTGGGACCACTTCTAAACATG -3′ ) for mutants 1 and 2 respectively . Amplification of a 70 bp right flanking arm was done using 197 bp reverse primer ( table 1 ) and forward primers: ( 5′-CATGTTTAGAAGTGAACACACCAAGAAAGA-3′ ) and ( 5′-CATGTTTAGAAGTGGTCCCACCAAGAAAGA-3′ ) for mutants 1 and 2 respectively . The left and right flanking arms were used as template and allowed to anneal for 6 rounds of the PCR cycle prior to the addition of the 197 bp forward and reverse primers . The resulting amplicon was then cloned into pGL4 . 10 . To generate the ORF50ppRREm MHV68 WT BAC DNA was used as a template for overlapping PCR . In the first PCR round left and right flanking arms were generated using ProxPromF ( 5′-GATCGCTAGCTCTTTATAGGTACCAGGGAA-3′ ) with ProxRREmR ( 5′-tcactctgttcaagaagttgcctgaggttcataaa-3′ ) , and ProxPromR ( 5′-TAGCAGATCTGGTCACATCTGACAGAGAAA-3′ ) with ProxRREmF ( 5′-ttcattttcaggccatttatgaacctcaggcaact-3′ ) respectively . These products were used as a template for a second round PCR amplification with primers ProxPromF and ProxPromR , and amplicons were cloned into pGL4 . 10 . Expression constructs were cloned into pCDNA 3 . 1 ( + ) ( Invitrogen ) using NotI and XhoI restriction sites using primers listed in table 1 . Both flag-tagged and non-tagged unspliced Rta were amplified from WT BAC DNA corresponding to viral genomic coordinates ( 66 , 761–69 , 374 ) . Murine IRF4 was amplified from pMSCV-IRF4-IRES-GFP ( a kind gift from Dr . Xiaozhen Liang ) . All PCR amplification was carried out using high fidelity Phusion DNA polymerase ( New England Biolabs ) , and sequence analysis confirmed completed plasmid constructs ( Macrogen ) . 5×105 293T cells were seeded into 6 well plates , the following day cells were transfected with 2 . 5 ug firefly luciferase and protein expression plasmids and 10 ng of pRL-TK ( Promega ) using TransIT 293T ( Mirus ) . 1×106 P3X63Ag8 cells were nucleofected with 5 ug firefly luciferase plasmids using Ingenio Electroporation Solution ( Mirus ) using setting X-01 on the Amaxa nucleofector . Reactions were done in triplicate for each condition , and 2–4 independent experiments were conducted . 48 hours later cells were lysed using passive lysis buffer ( 25 mM Tris-phosphate pH 7 . 8 , 2 mM DTT , 2 mM DCTA , 10% glycerol , 1% Triton X-100 ) . P3X63Ag8 cells were assessed for firefly luciferase activity using 10 µl lysate and 50 µl luciferase assay reagent ( LAR ) ( 75 mM HEPES pH 8 , 4 mM MgSO4 , 20 mM DTT , 100 µM EDTA , 53 . 0 µM ATP , 270 µM Coenzyme A , and 470 µM beetle Luciferin ) . A dual luciferase assay for firefly and renilla luciferase activity was performed on 293T cells using 10 µl cell lysate and 50 µl LAR , followed by the addition of 50 µl Stop & Glo reagent ( Promega ) . Light units were read on a TD-20/20 luminometer . Nuclear extracts of P3X63Ag8 cells grown under normal conditions were made as previously described [57] . Briefly , cells were washed with PBS , pelleted cells , resuspended in ice cold hypotonic lysis buffer and incubated on ice for 15 minutes . 10% Nonidet P-40 was added at 1/20 final volume and nuclei were spun down . Nuclei were then washed in hypotonic lysis buffer , resuspended in high salt buffer , and incubated with vigorous shaking for 2–3 hours at 4°C . Supernatants were collected following centrifugation and aliquoted on dry ice and stored at −80°C . Following isolation protein content in nuclear extract was quantified using DC Protein Assay ( BioRad ) , and western blot was performed to confirm presence of IRF4 . Electrophoretic mobility shift assay was performed using nuclear extracts as previously described [58] . Briefly , a binding reaction containing 10 µg of nuclear extract , 0 . 2 ng 32P-labeled double stranded oligonucleotide probe containing IRF4 consensus binding sequence ( underlined ) ( sense-5′-TTGGTGGTTTCACTTCTAAACA-3′ ) , and 2 ug poly ( di-dC ) was made up in binding buffer ( 10 mM Tris-HCl ( pH 7 . 5 ) , 10 mM HEPES , 50 mM KCl , 1 . 1 mM EDTA , and 15% glycerol , with 1 . 25 mM DTT ) and incubated on ice for 30 minutes . Supershift assays included 1 ug of IRF4 antibody ( clone M17 , Santa Cruz Biotech . ) or isotype control pSTAT1 antibody ( clone Tyr 701 , Santa Cruz Biotech . ) incubated with nuclear extracts slow shaking at 4°C for 1 hour . Competition experiments were performed with 2X and 20X unlabeled oligonucleotides containing WT or mutated ( underlined ) IRF4 consensus binding sequence ( sense 5′-TTGGTGGGACCACTTCTAAACA-3′ ) . Nucleoprotein complexes were run on 5% native polyacrylamide gel in 0 . 5X Tris Buffered EDTA at 180 V for 1 hour . Gel was dried under vacuum and analyzed with PhosphorImager analysis ( Typhoon 9410; Amerisham Bioscience ) . 10 cm dishes were seeded with 4×106 293T and were transfected the next day using TransIT-293T ( Mirus ) . 48 hours later cells were washed 2 times with ice cold PBS , and lysed while rocking at 4°C , in 1 mL Triton X Lysis Buffer ( 50 mM Tris HCl pH 7 . 4–7 . 5 , 150 mM NaCl , 1 mM EDTA , 0 . 1% Triton; supplemented with 1 mM NaF , 1 mM activated Na3V04 , and Roche EDTA free protease inhibitor cocktail tablet for 50 mL volume ) . Following lysis , membranes were pelleted and lysates were transferred to pre-chilled tubes . Protein concentration was determined using DC Protein Assay ( BioRad ) . 1 mg of cell lysate was precleared with prepared protein G beads ( Pierce ) , then 8 ug of IRF4 antibody ( clone M17 , Santa Cruz Biotech ) or flag antibody ( clone M2 , Sigma ) was added and lysates were incubated overnight at 4°C . Lysates were transferred to freshly prepared protein G beads for binding and were incubated at 4°C for 2 hours . Following wash , protein was eluted and samples were electrophoresed on 10% polyacrylamide gels , and transferred onto nitrocellulose membranes for western blot . The following detection antibodies were used: IRF4 ( clone H140 , Santa Cruz Biotech . ) and Flag ( clone M2 , Sigma ) .
|
Through coevolution with their hosts , gammaherpesviruses have acquired unique genes that aid in infection of a particular host . Here we study the regulation of the MHV68 M1 gene , which encodes a protein that modulates the host immune response . Using a strategy that allowed us to identify MHV68 infected cells in mice , we have determined that M1 expression is largely limited to the antibody producing plasma cells . In addition , we show that M1 gene expression is regulated by both cellular and viral factors , which allow the virus to fine-tune gene expression in response to environmental signals . These findings provide insights into M1 function through a better understanding of how M1 expression is regulated .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"animal",
"models",
"of",
"infection",
"viral",
"persistence",
"and",
"latency",
"virology",
"biology",
"and",
"life",
"sciences",
"microbiology",
"viral",
"replication"
] |
2014
|
The Murine Gammaherpesvirus Immediate-Early Rta Synergizes with IRF4, Targeting Expression of the Viral M1 Superantigen to Plasma Cells
|
Candida albicans is both a major fungal pathogen and a member of the commensal human microflora . The morphological switch from yeast to hyphal growth is associated with disease and many environmental factors are known to influence the yeast-to-hyphae switch . The Ras1-Cyr1-PKA pathway is a major regulator of C . albicans morphogenesis as well as biofilm formation and white-opaque switching . Previous studies have shown that hyphal growth is strongly repressed by mitochondrial inhibitors . Here , we show that mitochondrial inhibitors strongly decreased Ras1 GTP-binding and activity in C . albicans and similar effects were observed in other Candida species . Consistent with there being a connection between respiratory activity and GTP-Ras1 binding , mutants lacking complex I or complex IV grew as yeast in hypha-inducing conditions , had lower levels of GTP-Ras1 , and Ras1 GTP-binding was unaffected by respiratory inhibitors . Mitochondria-perturbing agents decreased intracellular ATP concentrations and metabolomics analyses of cells grown with different respiratory inhibitors found consistent perturbation of pyruvate metabolism and the TCA cycle , changes in redox state , increased catabolism of lipids , and decreased sterol content which suggested increased AMP kinase activity . Biochemical and genetic experiments provide strong evidence for a model in which the activation of Ras1 is controlled by ATP levels in an AMP kinase independent manner . The Ras1 GTPase activating protein , Ira2 , but not the Ras1 guanine nucleotide exchange factor , Cdc25 , was required for the reduction of Ras1-GTP in response to inhibitor-mediated reduction of ATP levels . Furthermore , Cyr1 , a well-characterized Ras1 effector , participated in the control of Ras1-GTP binding in response to decreased mitochondrial activity suggesting a revised model for Ras1 and Cyr1 signaling in which Cyr1 and Ras1 influence each other and , together with Ira2 , seem to form a master-regulatory complex necessary to integrate different environmental and intracellular signals , including metabolic status , to decide the fate of cellular morphology .
Candida albicans , one of the most common human fungal pathogens , is an important cause of morbidity and mortality in immunocompromised individuals , particularly in patients with AIDS or those undergoing cancer chemotherapy or transplantation procedures [1] . The prolonged use of antifungal agents in such compromised populations can lead to an increase in C . albicans resistance to many currently used therapies [2] . For this reason , there is an immediate need for new treatment options that can prevent or control diseases caused by C . albicans . In addition to being a pathogen , C . albicans is also a member of the commensal microflora of most individuals and the transition from commensal to pathogen is associated with the morphological switch from yeast to hyphal growth [3–5] . Environmental factors like 37°C , 5% CO2 , N-acetylglucosamine , pH , and serum , induce the yeast-to-hyphae switch [6] . However , most of these signals are always present in vivo , thus we still do not understand what governs the switch from benign colonization to symptomatic infection . During host colonization C . albicans lives amidst other microbes and , both , clinical data , that suggest a link between antibiotic usage and increased risk of fungal infections , and laboratory studies indicate that C . albicans interacts with bacteria in biologically important ways [7–14] . Further studies on bacterial-fungal interaction have led to the identification of new ways by which microbes modulate C . albicans growth . For example , 3-oxo-C12-homoserine lactone , produced by the Gram-negative bacterium Pseudomonas aeruginosa , inhibits hyphal growth by directly inhibiting the fungal Ras1-cAMP-protein kinase A ( PKA ) signaling pathway , a key regulator of the yeast-to-hyphae switch in C . albicans , by blocking cAMP synthesis [15 , 16] . The Ras1-cAMP-PKA signaling pathway is critical for C . albicans virulence in animal models [17–19] . Ras1 is a small GTPase that exists in the cell in an inactive ( GDP-bound ) form and an active ( GTP-bound ) form whose switch is regulated by the guanine nucleotide exchange factor ( GEF ) Cdc25 and GTPase-activating protein ( GAP ) Ira2 [20] . In its GTP-bound form , Ras1 directly interacts with the adenylate cyclase Cyr1 and stimulates cAMP production [18 , 21 , 22] . The cAMP signal subsequently derepresses two PKA isoforms which promote several cellular processes [23 , 24] . In current models of virulence , activation of the Ras1-cAMP-PKA pathway by host-associated stimuli induces the transition from yeast-to-hyphae growth and the expression of hypha-specific virulence factors . Hyphal growth increases tissue adherence and penetration , as well as the formation of adherent biofilms on medical devices [6 , 25 , 26] . This pathway also controls genes involved in glycolysis , stress resistance , cell wall composition , and mating [6 , 27] . More recently , an additional class of molecules that repress hyphal growth of C . albicans , phenazines , have been identified [28 , 29] . While phenazines are best known as small-molecule toxins with antibiotic properties toward bacterial and eukaryotic species at high concentrations [30] , recent studies have found that some phenazines ( phenazine-1-carboxylic acid ( PCA ) , phenazine methosulfate ( PMS ) , and pyocyanin ( PYO ) ) inhibit hyphal growth , intercellular adherence and biofilm development of C . albicans at low sub-lethal ( micromolar ) concentrations that are more than 100-fold below the concentrations which affect fungal survival [29] . Phenazines were found to inhibit C . albicans respiration [29] , which is consistent with other published data that phenazines can impact mitochondrial activity [31–33] . Subsequent analysis indicated that the decreased ability of C . albicans to develop wrinkled colonies ( consisting of hyphal and yeast cells ) or robust biofilms on plastic was due to inhibition in electron transport chain activity [29] . Indeed it was shown that during the filamentation process C . albicans activated the TCA cycle , inhibited the pentose phosphate pathway , and increased mitochondrial respiration [34] . This suggests that hyphal growth in C . albicans depends on functional respiration to cover the metabolic needs of the cell which is inhibited by phenazines . Early studies with mammalian mitochondria showed that phenazines uncouple oxidative phosphorylation by shunting electrons from endogenous pathways [35–37] , and this is most likely how respiration is inhibited in C . albicans . The inhibition of C . albicans filamentation by phenazines occurs despite the presence of robust fermentation pathways capable of supporting rapid growth in the absence of mitochondrial activity , and suggests communication between filamentation inducing pathways and metabolic state [29] . Indeed , in eukaryotes , it is becoming increasingly apparent that signaling pathways that sense and respond to extracellular cues often also incorporate input from the mitochondria themselves or from mitochondrially-derived molecules ( like ATP and reactive oxygen species ) [38 , 39] . In this report we show , that respiratory inhibition via genetic or biochemical manipulation decreases Ras1 activity and inhibits Ras1-dependent filamentation in C . albicans . Ras1 activation is also decreased by mitochondrial inhibition in the pathogenic Candida species , Candida parapsilosis and Candida tropicalis . Furthermore , utilizing a NRG1 overexpression strain and an efg1/efg1 null mutant we show that decreased Ras1 signaling in the presence of respiratory inhibitors is independent of morphological change . Filamentation was not repressed by MB in strains lacking Tup1 , a hyphal growth repressor , or in a strain overexpressing Ume6 , a transcription factor involved in the induction of hyphal growth indicating that the effects of MB on GTP-Ras1 can be circumvented with activation of downstream parts of the pathway . Analysis of overall metabolic changes due to respiratory inhibition shows perturbation of carbon metabolism , evidence for changes in redox state and increased AMP kinase activity ( increased β-oxidation , decreased sterol levels ) . Subsequent analysis showed that intracellular ATP modulates Ras1 activity independent of AMP kinase . Furthermore , while the GEF Cdc25 is dispensable for decreased Ras1 signaling due to respiratory inhibition , the GAP Ira2 is necessary . In addition , the adenylate cyclase Cyr1 is essential for this signaling cascade , showing for the first time that Cyr1 affects Ras1 activation state and with that it is not just a downstream effector of Ras1 . Rather Cyr1 , Ira2 , and Ras1 seem to form a regulatory complex that combines a multitude of signals to decide if the yeast-to-hyphae switch should take place .
Low micromolar concentrations of the bacterially-produced toxin , pyocyanin ( PYO ) , or its thioanalogue , methylene blue ( MB ) , perturb mitochondrial activity [40 , 41] and repress C . albicans filamentation ( Fig 1A ) [29] . Exposure to 1 . 5 μM MB , a compound used therapeutically in humans [41] , caused growth solely in the yeast form as indicated by a smooth colony morphology and cellular yeast morphology as determined by microscopy . While under control conditions , C . albicans grew as a mix of yeast and hyphae in wrinkled colonies . ( Fig 1A and S1A Fig panels 1 to 4 ) . Furthermore , MB led to decreased expression of hypha specific genes and increased levels of yeast-specific transcripts ( Fig 1B ) . Because the colony phenotype , cellular morphology , and expression profile of cells grown with MB were similar to those of the afilamentous ras1/ras1 mutant ( Fig 1A and 1B and S1A Fig panels 5 to 8 ) we sought to test the hypothesis that MB inhibits the yeast-to-hyphae switch by inhibition of Ras1 signaling , a pathway critical for C . albicans filamentation and virulence in animal models ( Fig 1C ) [17 , 18] . Examination of Ras1 protein levels in cells grown on solid medium with and without MB found that MB led to reductions in levels of active GTP-bound Ras1 ( GTP-Ras1 ) without affecting total Ras1 levels ( Fig 1D ) . When MB was added to C . albicans cultures in liquid YNBAGNP medium , we observed that MB decreased clumping and increased the percentage of cells in the yeast morphology at concentrations of 3 and 6 μM , but not at 1 . 5 μM , a concentration that completely inhibited filamentation on solid medium ( S2A Fig and Fig 1A ) . Analysis of the fraction of Ras1 in the GTP-bound state found that 3 and 6 μM MB also decreased the fraction of Ras1 in its active form ( S2A Fig ) . To test if these concentrations of MB have an impact on C . albicans growth we used a strain in which the hyphal gene repressor Nrg1 is overexpressed ( NRG1-OE ) . Nrg1 acts downstream of Ras1 and overexpression of this repressor prevents filamentation in the presence of hypha-inducing signals; the use of this strain makes it possible to measure growth via OD600 measurements in the presence of filamentation inducing signals [3 , 42] . The different concentrations of MB had no or only minimal impact on C . albicans growth excluding this as the reason for a decrease of GTP-Ras1 levels ( S2B Fig ) . Because filamentation is completely inhibited by 1 . 5 μM MB on solid media , but filamentation is only partially suppressed even at 6 μM in liquid medium , we used colony-grown cells in subsequent assays . We tested if MB affected GTP-binding of Ras1 in yeast growth conditions , and found that MB did not impact GTP-Ras1 levels ( Fig 1D and S2C Fig ) . This suggests that MB selectively inhibits the increase in GTP-Ras1 that occurs in the presence of filamentation-inducing signals which include 37°C , buffering at pH 7 , and the amino acids and N-acetylglucosamine in YNBAGNP medium . To determine whether the decrease of GTP-Ras1 by MB is specific to YNBAGNP , we tested another common filament-inducing condition ( YPD + 5% serum at 37°C ) . Consistent with our findings on medium with N-acetylglucosamine , amino acids , 37°C , and neutral pH , as the hyphal growth inducers , MB led to lower GTP-Ras1 levels when grown on medium with serum and repressed filamentation ( S2D Fig and S1B Fig ) . In liquid YPD + 5% serum , the effects of MB on morphology and GTP-Ras1 levels were modest suggesting that different media create different physiological states in C . albicans ( S2E Fig ) . All further experiments were conducted using YNBAGNP medium as it is a defined stimulus that mimics a number of aspects of the host ( pH 7 , amino acids , 0 . 2% glucose ) . To test whether a link between MB and Ras1 activation state can also be observed in other Candida species , we examined GTP-Ras1 levels of two other pathogenic Candida species , Candida parapsilosis and Candida tropicalis . These two fungal pathogens also had lower GTP-Ras1 levels on YNBAGNP with MB , indicating that Ras1 activation is also impacted by MB in other Candida species ( Fig 2 and S1A Fig panels 9 to 12 ) . Under these conditions , these fungi grow as yeast in the absence and presence of MB . In mammalian cells , MB decreases oxidative phosphorylation potential by oxidizing NAD ( P ) H-dependent dehydrogenase ( complex I ) and directly reducing cytochrome C thereby bypassing proton transfer by complex I and complex III ( Fig 3A ) [43] . Inhibition of mitochondrial activity with PYO and/or the complex III inhibitor Antimycin A ( AA ) reduced GTP-Ras1 levels; both PYO and AA also repressed hyphal growth as previously reported ( Fig 3B ) [29 , 44] . Mutants lacking complex I ( ndh51/ndh51 ) or complex IV ( cox4/cox4 ) did not filament and had low levels of GTP-Ras1 under control conditions ( Fig 3C and S1A Fig panels 13 to 16 for cellular morphology ) [45] . GTP-Ras1 levels were not further reduced by MB , and in fact levels increased in cells exposed to MB ( Fig 3C ) . Null mutants lacking complex II ( sdh1/sdh1 ) or both alternative oxidases ( aox1-A/aox1-A aox1-B/aox1-B ) , which do not participate in the formation of the proton gradient , still filamented and had high GTP-Ras1 levels ( S3A and S3B Fig ) . MB also caused a decrease in GTP-Ras1 in the complex II and alternative oxidase mutants comparable to wild type ( S3A and S3B Fig ) . All three complexes important for high GTP-Ras1 levels under hypha-inducing conditions ( complex I , III , and IV ) pump protons into the intermembrane space for use in ATP synthesis ( Fig 3A ) . MB reduces ATP synthesis in C . albicans as relative ATP levels were 2 . 3-fold lower upon growth of the wild type ( WT ) with MB ( Fig 3D ) . Furthermore , while intracellular ATP in ndh51/ndh51 and cox4/cox4 mutants in control conditions were significantly lower than in the WT ( CAF2 ) ( Fig 3D ) , ATP levels were not reduced by MB ( Fig 3C ) . To determine if filamentation was required for elevated GTP-Ras1 and higher ATP , we examined the effects of MB on mutant strains that are unable to undergo the yeast-to-hyphae switch . It is well known that the transcription factor Efg1 is an essential regulator for morphogenesis in C . albicans [46] . In hyphae inducing conditions Efg1 is activated through the Ras1–cAMP–PKA pathway and induces the expression of many hyphae specific genes that are essential for the yeast-to-hyphae transition [47 , 48] . In addition , we tested the NRG1 overexpression strain ( NRG1-OE ) . Because both Efg1 and Nrg1 act downstream of Ras1 we hypothesized that MB effects on Ras1 should be unaffected in these strains . As expected the efg1/efg1 mutant formed a smooth colony consisting of yeast cells in the presence and absence of MB ( S4 Fig and S1A Fig panels 17 and 18 ) . The NRG1-OE strain showed a weakly wrinkled colony morphology under control conditions consisting of mainly yeast cells with some elongated yeast cells and short pseudohyphae , while with MB a completely smooth colony consisting of only yeast cells was observed ( Fig 4A ) . Subsequent western blot analysis and intracellular ATP measurements showed that both strains had less GTP-Ras1 and less ATP with MB , as the WT ( Fig 4A and 4B and S4 Fig ) . The reduction of GTP-Ras1 in WT ranged from 26 . 5% to 94 . 5% with an average reduction of 63 . 6% over all experiments done; even in assays with only a 26 . 5% reduction in the ratio of GTP-Ras1/ total Ras1 relative to control , filamentation was repressed . Furthermore , we tested the effects of MB on mutant strains that are constitutively filamentous due to the alteration of downstream transcriptional regulators of hyphal growth to determine if the effects of MB were upstream in the hyphal growth pathway and if hyphal growth could be reactivated in the presence of MB . Loss of the hyphal gene repressor Tup1 or overexpression of the transcription factor Ume6 have been previously shown to result in constitutive filamentation [3 , 49 , 50] . Both strains are able to filament in the presence of MB , while GTP-Ras1 levels are decreased ( Fig 4C ) . Filamentation and wrinkled colony formation of these strains is not as strong as under control conditions . However , overall this shows that the effects of MB on ATP and GTP-Ras1 are upstream events in the control of C . albicans morphology and that low Ras1 signaling inhibits filamentation in the WT . While MB , PYO , and AA all modulate mitochondrial activity and reduced relative GTP-Ras1 levels ( Fig 3B ) , these compounds are not equivalent . For example , MB did not impact growth rates , while PYO and AA were inhibitory perhaps in part due to increased ROS formation . In addition , while AA and PYO led to acidification of the medium due to increased fermentation to acetate , MB did not [29 , 44] . Thus , we sought to determine the strongest common signals in order to gain insight into factors that control Ras1 GTP-binding . Metabolomics analysis of the WT SC5314 grown under control conditions or with MB , PYO , or AA revealed that some compounds were only differentially regulated in one condition ( Fig 5A , S5 Fig , and S1 Table ) . For instance , only MB-grown cells had significantly high relative levels of glycerol , an alternative fermentation product , and this is consistent with the observation that the medium pH was not altered even in the presence of this fermentation-inducing mitochondrial inhibitor ( S5 Fig ) . A large group of metabolites showed a similar pattern in the presence of all three compounds ( Fig 5A ) ; these signatures included increased lipid catabolism and decreased lipid biosynthesis ( higher levels of acetyl-CoA , increased lysophospholipids , and decreased fatty acids ( palmitate , oleate , and stearate ) ) as well as low ergosterol and related compounds ( Fig 5B and S5 Fig ) . The decrease of the fatty acid palmitate is particularly interesting because it is needed for one of two lipid modifications that tether Ras1 in C . albicans to the plasma membrane [51] . Loss of palmitoylation re-localizes Ras1 largely to endomembranes and changes in Ras1 localization negatively affect Ras1 activation [51] . To test if the changes in GTP-Ras1 levels seen with MB or the other inhibitors were due to re-localization of Ras1 away from the membrane , we determined GTP-Ras1 levels in two strains carrying truncated Ras1 proteins that are no longer associated with the plasma membrane . The first strain is a ras1/ras1 mutant reconstituted with a ras1 allele missing the last 67 amino acids ( ras1Δ67 ) [52] and the second is a ras1/ras1 mutant reconstituted with a ras1 allele only including the conserved N-terminal region of RAS1 ( ras1 N-term ) . Interestingly , the ras1Δ67 strain showed lower levels of GTP-Ras1 in control conditions compared to the ras1 N-term or wild type strain ( S6 Fig ) suggesting a possible GTP-binding inhibitory domain or function activated by Ras1 cleavage [52] . However , both truncated Ras1 variants showed a wild type reduction of GTP-Ras1 levels with MB ( S6 Fig ) . Thus , while Ras1 localization is controlled by its C-terminal lipid modifications , changes in these that might occur in the presence of MB were not responsible for altered GTP-Ras1 levels . Interestingly , the metabolomics pattern strongly resembled the response of mammalian cells to MB [53] . Furthermore , in mammalian cells the same metabolic shift due to respiratory inhibition is mediated by AMP kinase ( AMPK ) , an energy sensor that responds to relative ATP:AMP/ADP levels . Lipids are a rich source of ATP , and AMPK induces a lipid catabolic state when ATP levels are low [54 , 55] . These signatures suggest that the common signal in response to PYO , MB and AA is likely low intracellular ATP . GTP-Ras1 levels were not controlled by AMPK itself as a mutant lacking the γ-subunit of AMPK ( snf4/snf4 ) which is essential for AMPK activity in Saccharomyces cerevisiae [56] , still shows a reduction in GTP-Ras1 and intracellular ATP upon growth with MB ( Fig 5C and 5D and see S1A Fig panels 19 to 22 for cellular morphology ) . Under control conditions the snf4/snf4 mutant has very low levels of intracellular ATP and is unable to filament ( Fig 5C and S1A Fig panels 29 ) . Across eukaryotes , it has been shown that diverse cellular processes from proteome function to neurotransmitter responses are directly regulated by ATP levels . Thus , phenazine-mediated repression of C . albicans filamentation may occur through effects on ATP levels , as ATP is the precursor to cAMP , a second messenger that is a key positive regulator of hyphal growth ( Fig 1C ) [6] . PYO reduces levels of both cAMP and its precursor ATP in human epithelial cells due to its effects on respiration and oxidative phosphorylation [57] . To determine more directly if decreased ATP levels were impacting Ras1 signaling , we examined the effects of inhibitors of the proton gradient ( dinitrophenol ( DNP ) ) and the ATP synthase ( oligomycin ) which each caused a significant decrease of intracellular ATP ( Fig 6A ) . For both , relative levels of GTP-Ras1 were decreased and filamentation was repressed ( Fig 6B ) strongly suggesting that ATP levels were the connecting signal between mitochondrial activity and Ras1 signaling . To further test this hypothesis , we measured GTP-Ras1 levels in the ssn3/ssn3 mutant that had been previously shown to have increased intracellular ATP due to increased oxidative metabolism , without increased growth [44] , and found increased GTP-Ras1 levels compared to the reconstituted strain ( Fig 6C ) . In summary , our data show that GTP-Ras1 levels correlate with intracellular levels of ATP . C . albicans Ras1 GTP-binding has been genetically shown to be controlled by a GEF , Cdc25 , and a GAP , Ira2 [58 , 59] . The cdc25/cdc25 mutant had low levels of GTP-Ras1 , and was unable to filament , but MB caused a further reduction in GTP-Ras1 comparable to WT ( Fig 7A and see S1A Fig panels 23 and 24 for cellular morphology ) . In contrast , loss of Ira2 resulted in a hyperfilamentous phenotype and strongly increased levels of GTP-Ras1 which were unaffected by addition of MB ( Fig 7B and 7C ) , while ATP levels were decreased comparable to WT ( Fig 7D ) , showing that the decrease of GTP-Ras1 by MB is Ira2 dependent . In S . cerevisiae , Ira2 activity is negatively regulated through direct interactions with Tfs1 and positively regulated through protein stabilization by Gpb1/2 [60 , 61] . While C . albicans tfs1/tfs1 mutants displayed phenotypes consistent with increased Ira2 activity ( decreased filamentation and less GTP-Ras1 ) , the reduction of GTP-Ras1 levels upon growth with MB was similar to that of the WT ( S7A Fig ) . Deletion of the C . albicans Gpb1 homolog resulted in increased GTP-Ras1 levels under control conditions that were decreased with MB comparable to wild type ( S7B Fig ) . Together the Gpb1 and Tfs1 data suggest that new inputs into Ira2 may link ATP levels to GTP-Ras1 . We suspected this link may be the adenylate cyclase Cyr1 , which is known to be activated by Ras1 , and integrates diverse signals . The cyr1/cyr1 mutant , like a ras1/ras1 strain , forms smooth colonies consisting only of yeast ( Fig 8 and see S1A Fig panels 25 to 28 for cellular morphology ) . Surprisingly , the cyr1/cyr1 strain had a higher proportion of GTP-Ras1 , and this increase was complemented by addition of the native CYR1 gene . Furthermore , in the absence of Cyr1 , Ras1 GTP-binding was not decreased by MB or AA but rather increased ( Fig 8A and 8B ) . The cAMP signal itself appeared to be important , as a strain expressing only a catalytically-inactive Cyr1 ( cyr1/cyr1 +cyr11334 ) also had higher basal GTP-Ras1 levels that were increased and not decreased by MB ( Fig 8C and see S1A Fig panels 29 to 32 for cellular morphology ) . However , neither subunit of PKA , the only known cAMP sensor , was required for the control of GTP-Ras1 levels ( S7C Fig ) . In summary , our data suggest that low ATP causes Cyr1-mediated activation of Ira2 activity to reduce GTP-Ras1 levels . Thus , it appears that Ras1 and Cyr1 participate in a regulatory circuit that integrates multiple signals before triggering the expression of virulence related attributes .
In this study we identified a previously unknown link between total intracellular ATP levels and Ras1 signaling in C . albicans by characterizing the mechanism by which MB inhibits the C . albicans yeast-to-hypha switch ( Fig 9 ) . Interestingly , a recent study in the yeast S . cerevisiae found that dysfunctional mitochondria decrease cAMP-PKA signaling , adhesion production , and filamentous growth further emphasizing that the link between respiratory activity and Ras1-cAMP-PKA signaling is conserved beyond the Candida genus [62] . The same study also showed that the filamentous-growth-specific MAPK pathway is not involved in this signaling as this pathway retained functionality in respiratory-deficient S . cerevisiae yeast cells [62] . Furthermore , while it is not known whether Ras1 signaling is important for filamentation or virulence in C . tropicalis and C . parapsilosis , when grown on YNBAGNP media with and without MB , both Candida species had decreased Ras1 activation state with MB indicating that the link between respiratory activity and Ras1 signaling is conserved across Candida species . However , whether this decrease in Ras1 activation impacts filamentation and virulence of these fungal pathogens needs to be determined in future studies . In liquid conditions , in which C . albicans hyphal growth is fast , increased MB concentrations were necessary to see a decrease in GTP-Ras1 levels and filamentation . This requirement for higher levels of MB may be due to higher or altered respiratory activity , or differences in ATP homeostasis under well-mixed planktonic conditions . Interestingly , in our assays , GTP-bound Ras1 was lower and filamentation was inhibited by MB on both solid YPD + 5% serum medium and solid and liquid YNBAGNP . However , in liquid YPD + 5% serum conditions , the effect of MB on GTP-Ras1 levels was minor and no impact on morphology at concentrations that were not inhibitory . A recent publication by O’Meara and colleagues reported a global analysis of C . albicans morphology which showed that the role of different pathways in filamentation varied depending on the medium condition [63] , and we speculate that the effects of MB on ATP pools , Ras1 , and filamentation also varies in different media in ways that we cannot yet understand . Together , this variability shows the importance of understanding the interactions between nutrient sources and growth substrate and signaling inputs and outputs . The observation that MB had no impact on Ras1 GTP-binding under yeast growth conditions in C . albicans which could indicate that intracellular ATP pools serve as a “check point” in Ras1 signaling under hypha-inducing conditions . The differential effects of MB may be related to temperature ( yeast are grown at 30°C while hyphae are grown at 37°C ) though MB did modulate ATP levels at 30°C . Indeed , isolated mitochondria from C . albicans were shown to be more active at 37°C compared to 30°C [34] . At lower temperatures , mammalian mitochondria have a reduced respiratory rate and hyperpolarization of the mitochondrial membrane , potentially due to decreased ATPase activity [64–67] , which results in the accumulation of reduced flavins and cytochromes ( Fig 3A ) . This state may render cells less susceptible to the action of MB . Increased respiration upon growth at 37°C most likely result in an increase in intracellular ATP concentrations which may promote or permit the hyphal growth program . MB inhibits the accumulation of ATP ( Fig 3D ) causing cells to stay in the yeast morphology in conditions that would normally induce filamentation . Mutants defective in complex I and complex IV never establish a high intracellular ATP state , and are unable to undergo the yeast-to-hypha switch . Interestingly , in the presence of MB the ndh51/ndh51 and cox4/cox4 mutants showed an increase of GTP-Ras1 levels instead of a decrease ( Fig 3C ) . This is consistent with observations made in mammalian cells where it has been shown that MB can restore some electron flow to dysfunctional mitochondria [43] . There was only a small but measurable increase of ATP levels in these mutants with MB which could result in a small increase in GTP-Ras1 ( Fig 3D ) indicating some increased electron flow might occur in the presence of MB in these C . albicans mitochondria similar to mammalian cells . ATP production via respiration by the mitochondria is the main source of the chemical energy that fuels many different processes and pathways in the cell . Consequently , inhibition of respiration and ATP production will have a major impact on many aspects of cellular metabolism as shown by the metabolomics analysis in this study ( Fig 3A , S1 Table , and S5 Fig ) . However , the strongly reduced ATP levels observed in the presence of MB did not block filamentous growth in the constitutively filamentous tup1/tup1 mutant or UME6-OE strain suggesting that MB is not inhibiting other parallel signaling or metabolic pathways important for induction or maintenance of filamentation ( Fig 4C ) . Filamentation and wrinkled colony formation of these strains is not as robust as under control conditions , possibly due to reduction in other Ras1-controlled pathways or other effects of low ATP levels on growth dynamics that could be Ras1 independent , but hyphal growth is clearly evident in the strains even when MB is present . Furthermore , the NRG1-OE strain showed that filamentation is not required for higher ATP and elevated GTP-Ras1 under filamentation inducing conditions or the effects of MB on Ras1 signaling ( Fig 4A and 4B ) . Under the conditions tested , the NRG1-OE strain formed some wrinkles , however , it does not form true hyphae . Microscopy of the cells showed mainly budding yeast with some elongated yeast cells and short pseudohyphae ( S1A Fig panel 21 ) . Interestingly , the occurrence of elongated yeast cells and pseudohyphae is inhibited by MB . Previous publications with the NRG1-OE strain looked at YPD +serum , which in our hands is not an as strong an inducer of filamentation and wrinkle formation as YNBAGNP on plates and the overexpression level of the NRG1-OE strain might just not be enough to overcome this stronger induction completely ( Fig 1A and S1B Fig ) [3 , 42] . Wrinkle formation of C . albicans colonies by MB and other respiratory inhibitors may support the model in which wrinkles promote usage of and demand for oxygen and thus the regulation of wrinkle production is downregulated upon respiratory inhibition [68] . In the bacterium Pseudomonas aeruginosa wrinkled colony morphology has been shown to be a redox-driven adaptation that maximizes oxygen accessibility and increased oxygen is able to inhibit wrinkle formation [69] . It is very interesting that the response of C . albicans to respiratory inhibition is similar to what has been seen in mammalian cells [53] . We observed a metabolome profile typical for AMP kinase activation , which has also been shown in mammalian cells exposed to MB [70] . This activation is not surprising as AMP kinase is a known energy sensor that measures ratios of ATP to ADP/AMP [55] that is activated when cellular energy status is low in eukaryotes . To increase ATP availability , the cells increase the catabolism of energy stores , such as fatty acids , and the biosynthesis of “costly” fatty acids and ergosterol is decreased ( S1 Table , S5 Fig ) [54 , 55 , 71] . Our data indicate that AMPK is needed to sustain levels of ATP , as intracellular ATP levels were very low in the snf4/snf4 strain ( Fig 5D ) . We suspect this is the reason why GTP-Ras1 levels are so low and why this strain does not form filaments ( Fig 5C ) . In agreement , a recent study showed that loss of only the kinase activity due to a point mutation in Snf1 , an essential protein in C . albicans , inhibited the yeast-to-hyphal switch indicating how important AMPK activity is for energy homeostasis and filamentation [72] . Even though GTP-Ras1 and ATP levels are low in the snf4/snf4 strain , they are still responsive to MB showing that AMPK is not necessary for this signaling pathway . Overall , we believe that by understanding how C . albicans metabolism changes in different environments , we can use this fungus as an important probe for conditions within the host in states of health and disease . In many cells , ATP concentrations control diverse processes such as autophagy [39] , the retrograde response pathway [73] , activation of neutrophils [74] , and neurotransmitter responses [75] . ATP can act as an essential co-factor or can be recognized directly by binding to a receptor which triggers signaling . We do not yet know how ATP levels impact the Ras1 signaling cascade . One candidate ATP sensor is adenylate cyclase , Cyr1 , itself . Cyr1 catalyzes the conversion of ATP to cAMP and thus has an ATP binding site that could function as a sensor of ATP levels . In S . cerevisiae , studies indicate that Cyr1 acts as a scaffold protein for Ras2 ( homolog to Ras1 in C . albicans ) interactions with Ira2 [76] and one could imagine that this scaffold activity may be regulated by ATP concentration . We know that Cyr1 signaling is required for the effects of MB in C . albicans as a catalytically inactive Cyr1 , which can still serve some structural roles , was also insensitive to the Ras1-inhibiting effects of MB ( Fig 8C ) . Indeed the cyr1/cyr1 mutant and the strain expressing catalytically inactive Cyr1 showed an increase of GTP-Ras1 with MB . Together with previously published data showing that Ras1 signaling is important for mitochondrial activity in S . cerevisiae , this might indicate that the mitochondria are not functioning normally in these strains and that like in mammalian cells and the cox4/cox4 and ndh51/ndh51 strains MB is able restore some electron flow to these dysfunctional mitochondria [43 , 77] . In S . cerevisiae , Ira2 has been shown to interact with the protein kinase A regulatory subunit , and Cyr1 and Ira2 have both been found at the plasma membrane and on mitochondria [78] providing further support for the potential for interactions between Cyr1 and Ira2 , probably in ways that respond to mitochondrial activity [79] ( Fig 9 ) . These reports , with the data presented here , suggest that Cyr1 and Ras1 form a master regulatory circuit . Furthermore , the canonical Ras1 signaling pathway model has to be restructured from a pathway in which Cyr1 is just a factor downstream of Ras1 that is activated by GTP-bound Ras1 ( Fig 1C ) to a new model in which Cyr1 and Ras1 influence each other and together with Ira2 form a master-regulatory network necessary to coordinate the response to different environmental and intracellular signals in order to decide the fate of the cell ( Fig 9 ) . These data reveal important aspects of the regulatory cascade that controls the C . albicans switch to a state more capable of causing host damage ( Fig 9 ) . Our findings indicate that the energy status of the cell is one of the most important signals involved in the decision of C . albicans to undergo the yeast-to-hyphae switch as it is able to override an array of filamentation inducing signals ( Fig 9 ) Thus , host or host microbiome factors that impact energy levels will likely modulate C . albicans Ras1 signaling . Previous studies showed that Cyr1 can be directly activated by bicarbonate and muramyl dipeptides ( MDPs ) ( Fig 9 ) , though MDPs are only weak activators of filamentation in the absence of Ras1 [80 , 81] showing that Ras1 input is required for strong MDP-induced filamentation . Indeed , clinical data have linked the use of antibacterials to increased risk of C . albicans infections in multiple distinct body sites with very different bacterial community compositions [7–14] . In addition , numerous studies have shown that many bacteria inhibit C . albicans filamentation [13 , 14] . Future studies will focus on establishing whether repression in Ras1 activation , through modulation of ATP by competition with other microbes , contributes to the control of Candida behavior in a healthy mucosal microbial community . Furthermore , the future will show if known therapies or strategies can be used to favor benign host-Candida interactions by promoting low Ras1 activity .
All C . albicans strains were streaked from-80°C onto YPD ( 1% yeast extract , 2% peptone , 2% glucose ) plates every 8–10 days and maintained at room temperature . All strains used in this study can be found in S2 Table . Overnight cultures were grown in 5 ml of YPD , supplemented with uridine as necessary , and washed in distilled water ( dH2O ) prior to use . C . albicans cells were mostly grown under filament-inducing conditions which included 37°C on YNBAGNP ( 1 . 5% agar , 0 . 67% yeast nitrogen base medium with ammonium sulfate ( RPI Corp ) , 10 mM dextrose , 5 mM N-acetylglucosamine ( GlcNAc ) , and 2% [wt/vol] casamino acids ( BD Bacto ) , 25 mM potassium phosphate buffer ) . Cells were also grown on YPD + 5% fetal bovine serum when indicated . For yeast growth conditions C . albicans cells were grown at 30°C on YNBGP ( 1 . 5% agar , 0 . 67% yeast nitrogen base medium with ammonium sulfate ( RPI Corp ) , 10 mM dextrose , 25 mM potassium phosphate buffer ) . For filamentation-inducing liquid growth conditions media were prepared as described above without the addition of agar and cells were incubated in the roller drum for 12 hours at 37°C . Stock solutions were prepared of: methylene blue ( MB ) ( Fisher Scientific ) - 3 mM in dH2O; pyocyanin ( PYO ) ( Cayman Chemicals ) - 30 mM in 100% ethanol ( EtOH ) ; Antimycin A ( AA ) ( Sigma ) - 10 mM in 100% EtOH; oligomycin ( Sigma ) – 8 mg/ml in 100% EtOH; Dinitrophenol ( DNP ) ( Sigma ) – 100 mM in DMSO; menadione ( Sigma ) – 50 mM in 100% EtOH . All experiments were conducted in the dark to avoid light-induced ROS production . The deletion mutant strains were constructed in the BWP17 strain background using a previously described method [58 , 82] . Briefly , gene-disruption cassettes for transformation were amplified using ~75 bp primers and the plasmids , pRS-ARG4 or pGEM-HIS1 [82] which contain ARG4 and HIS1 for PCR-directed integration . The forward primer was designed to have homology to 50 bp sequence upstream of the gene of interest start codon while the reverse primer had homology to the 50 bp sequence following the stop-codon . Both the primers were flanked by a 20 bp sequence homologous to the plasmids , as mentioned above . Sequential transformations of these gene-disruption cassettes into C . albicans BWP17 strain yielded the deletion strain . Plasmid pSM2 and pSMTC were used to complement the cyr1/cyr1 strain only with URA3 or with CYR1-URA3 at the URA3 locus [80] . The ira2/ira2 strain was reconstituted with URA3 using the pClp10 plasmid at the RP10 locus [83] . Strain ras1/ras1 + ras1 N-term was generated by transforming DH482 with PacI linearized pAP13+ras1 N-term and integration at the endogenous RAS1 locus was confirmed by PCR . To construct pAP13+ras1 N-term a PCR product encoding the first 161 residues of Ras1 was amplified from pAP14 [51] with primers RAS1XhoIF [51] and Ras1delta129BamHI-R , digested with XhoI and BamHI and ligated into similarly digested pAP13 . All plasmids and primers used in this study can be found in S3 and S4 Tables . For wrinkled colony formation , 10 μl from overnight cultures re-suspended in dH2O at an optical density ( OD ) of 8 . 0 were spotted onto YNBAGNP unless otherwise specified . The medium was supplemented with methylene blue ( MB ) from a 3 mM stock solution to a final concentration of 1 . 5 μM . 5 μM MB was used for the metabolomics experiment . Pyocyanin ( PYO ) , Antimycin A ( AA ) , and oligomycin ( olig . ) were added to the medium to a final concentration of 20 μM , 2 . 5 μM , and 7 . 5 μg/ml , respectively , or an equivalent volume of 100% ethanol ( vehicle ) . Dinitrophenol ( DNP ) was added to the medium to a final concentration of 2 mM and menadione was added to the medium to a final concentration of 0 . 125 mM or an equivalent volume of vehicle solution . Cells were incubated at 37°C for 25 h . Colonies were imaged after 24 h with a digital camera . Unless otherwise noted , all spot assays were completed as at least three independent replicates and a representative data set is shown . Cell morphology in colonies was assessed using a ZeissAxiovert inverted microscope equipped with a 100x long working distance objective and Axiovision software . To image the morphology of cells within the colony , the cells were resuspended in water , then applied to an agarose-coated slide to immobilize cells of different morphologies . The images shown were representative of the make up of the entire colony . For western blot analysis spot colonies were scraped from agar plates after 24 h incubation at the conditions indicated , washed into a collection tube with dH2O and , after centrifugation , immediately snap-frozen in an ethanol/dry ice bath . Lysate preparation was conducted as previously published , with some modifications [51] . Whole-cell lysates were prepared by resuspending cells in Lysis/Binding/Wash Buffer ( Active Ras Pull-Down and Detection Kit , Pierce ) with protease inhibitors ( Halt Protease Inhibitor Single-Use Cocktail , Pierce ) and disrupting cells with glass beads in a Bio-Spec bead beater with six rounds of 50 seconds disruptions at 4°C and 1 minute rests on ice . Protein concentrations were determined by Bradford assay ( BioRad ) . Active or GTP-bound Ras1 was isolated utilizing the Active Ras Pull-Down and Detection Kit ( Pierce ) following the manufacturer’s instructions . In general , 200 μg of total protein were used for the pull-down unless otherwise specified . Due to the strong increase of GTP-Ras1 levels in ira2/ira2 strain only 100 μg of total protein were used ( indicated in the figure ) . 12 . 5 μl of the pull-down samples containing active Ras1 , and , for the input control , a total of 10 μg total protein diluted in SDS loading buffer were separated by SDS-PAGE , transferred to polyvinylidene difluoride ( PVDF ) with the Trans-Blot Turbo Transfer system ( BioRad ) , and detected with monoclonal anti-Ras clone 10 ( 1 . 5 μg/ml; Millipore ) , followed by secondary detection with goat anti-mouse ( Pierce ) and enhanced chemiluminescent visualization ( Pierce ) . As a control protein Pma1 was detected as described previously [52] . Densitometry analysis of Ras1 levels on Western blots was conducted with ImageJ [84] . Nanostring nCounter ( Nanostring Technologies ) analysis was used to quantify C . albicans gene expression . After 24 h spot colonies were harvested and fungal RNA was isolated using MasterPure Yeast RNA Purification Kit ( Epicentre ) . Each Nanostring reaction mixture contained 80 ng fungal RNA , hybridization buffer , reporter and capture probes . Overnight hybridization of RNA with probes at 65°C preceded sample preparation using Nanostring prep station . Targets were counted on the nCounter using 255 fields of view per sample [85] . Raw counts for hyphal and yeast specific transcripts ( HWP1 , ECE1 , HGC1 , HYR1 , ALS3 , YWP1 and ALS4 ) were normalized within each sample to the geometric mean of two C . albicans housekeeping genes ( ACT1 , PMA1 ) and scaled to WT control conditions; the numerical average was taken from three biological replicates . Heat maps were developed using Z-scoring of Nanostring counts of selected yeast- and hyphal-specific genes using the “heatmap . 2” function in the “gplots” package [86] in R ( R Foundation for Statistical Computing , Vienna , Austria ) . Spot assays were completed as previously described on YNBGNP agar plates and incubated at 37°C for 24 h . Cells were harvested , by scraping colonies from the surface of the agar using a coverslip , and then snap-frozen in an ethanol/dry ice bath . A total of 5 biological replicates were submitted to Metabolon for metabolite profiling , by GC/MS and LC/MS , of SC5314 wild type treated with vehicle ( EtOH ) , 5 μM MB , 20 μM PYO , or 2 . 5 μM AA . All metabolites with mean values that had significant differences ( p≤0 . 05 ) between treated and untreated samples were clustered into the 2 groups “UP” ( upregulated ≥1 . 00-fold ) or “DOWN” ( downregulated <1 . 00-fold ) . VennMaster ( http://sysbio . uni-ulm . de/ ? Software:VennMaster ) [87] was used to determine the overlap of biochemicals that were either “UP” or “DOWN” in any of the treated samples . To visualize the result of this overlap analysis the tool eulerAPE ( http://www . eulerdiagrams . org/eulerAPE ) [88] was used to generate the Euler diagrams . Spot assays were completed as previously described on YNBGNP agar plates with and without 1 . 5 μM MB or 7 . 5 μg/ml oligomycin and incubated at 37°C for 24 h . After harvesting by scraping colonies from the surface of the agar using a coverslip , the cells were disrupted with glass beads and 1x PBS in a Bio-Spec bead beater with 3 rounds of 60 seconds disruptions at 4°C and 1 minute rests on ice in between . A standard curve was prepared using Adenosine 5’-triphosphate disodium salt hydrate ( Sigma ) . ATP levels were measured using the CellTiter-Glo Luminescent Cell Viability Assay ( Promega ) following the manufacturer’s instructions . The luminescent signal , which is proportional to ATP levels , was measured using a Tecan Infinite 200 Pro equipped with Magellan software ( Tecan ) . All data were normalized to the protein concentration of each sample , which was determined using a Bradford Assay ( BioRad ) . Three independent biological replicates , each including three technical replicates , were conducted and a representative data set is presented . RAS1: C2_10210C_A; CYR1: C7_03070C_A; TPK1: C1_10220C_A; TPK2: C2_07210C_A; CDC25: C3_03890W; IRA2: C1_12450C_A; TFS1: C5_00930C_A; GPB1: C4_02150C_A; NDH51: C2_04550C_A; SDH1: C1_05260C_A; AOX1-A: C1_09160W_A; AOX1-B: C1_09150W_A; COX4: C2_01620W_A; SNF4: C6_03920W_A; NRG1: C7_04230W_A; UME6: C1_06280C_A; EFG1: CR_07890W_A; TUP1: C1_00060W_A; SSN3: C2_04260W_A
|
Candida albicans is a successful fungal commensal and pathogen of humans . It is a polymorphic organism and the ability to switch from yeast to hyphal growth is associated with the commensal-to-pathogen switch . Previous research identified the Ras1-cAMP-protein kinase A pathway as a key regulator of hyphal growth . Here , we report that mitochondrial activity plays a key role in Ras1 activation , as respiratory inhibition decreased Ras1 activity and Ras1-dependent filamentation . We found that intracellular ATP modulates Ras1 activity through a pathway involving the GTPase activating protein Ira2 and the adenylate cyclase Cyr1 . Based on our data the canonical Ras1 signaling model in C . albicans needs to be restructured in such a way that Cyr1 is no longer placed downstream of Ras1 but rather in a major signaling node with Ras1 and Ira2 . Our studies suggest that the energy status of the cell is the most important signal involved in the decision of C . albicans to undergo the yeast-to-hyphae switch or express genes associated with the hyphal morphology as low intracellular ATP or associated cues override several hypha-inducing signals . Future studies will show if this knowledge can be used to develop therapies that would favor benign host-Candida interactions by promoting low Ras1 activity .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2015
|
Mitochondrial Activity and Cyr1 Are Key Regulators of Ras1 Activation of C. albicans Virulence Pathways
|
Trypanosoma cruzi is a protozoan pathogen responsible for Chagas disease . Current therapies are inadequate because of their severe host toxicity and numerous side effects . The identification of new biotargets is essential for the development of more efficient therapeutic alternatives . Inhibition of sirtuins from Trypanosoma brucei and Leishmania ssp . showed promising results , indicating that these enzymes may be considered as targets for drug discovery in parasite infection . Here , we report the first characterization of the two sirtuins present in T . cruzi . Dm28c epimastigotes that inducibly overexpress TcSIR2RP1 and TcSIR2RP3 were constructed and used to determine their localizations and functions . These transfected lines were tested regarding their acetylation levels , proliferation and metacyclogenesis rate , viability when treated with sirtuin inhibitors and in vitro infectivity . TcSIR2RP1 and TcSIR2RP3 are cytosolic and mitochondrial proteins respectively . Our data suggest that sirtuin activity is important for the proliferation of T . cruzi replicative forms , for the host cell-parasite interplay , and for differentiation among life-cycle stages; but each one performs different roles in most of these processes . Our results increase the knowledge on the localization and function of these enzymes , and the overexpressing T . cruzi strains we obtained can be useful tools for experimental screening of trypanosomatid sirtuin inhibitors .
Acetylation is a ubiquitous protein modification present in prokaryotic and eukaryotic cells that participates in the regulation of many cellular processes . A limited set of acetyltransferases and deacetylases , and of the acetyl-lysine “reading” domain ( bromodomain ) are the principal components of the acetylation/deacetylation machinery . Among them , protein deacetylases are enzymes that catalyze the removal of acetyl groups from the ε-amino group of lysine residues and are classified into four classes . Sirtuins , the class III ( NAD+-dependent ) protein deacetylases , are homologous to the yeast transcriptional repressor , Sir2 [1] . Sir2 , as well as all sirtuins , deacetylates lysine residues in a unique chemical reaction that consumes nicotinamide adenine dinucleotide ( NAD+ ) and generates nicotinamide , O-acetyl-ADP-ribose ( OAADRr ) , and the deacetylated substrate [2] . Saccharomyces cerevisiae Sir2 , the founding member of the group , is a histone deacetylase ( reviewed in [3] ) involved in a range of chromatin-mediated processes; namely , gene silencing at telomeres and mating-type loci , DNA repair [4–5] , suppression of recombination within ribosomal DNA ( rDNA ) [6] , DNA replication [7] , chromosome stability [8] and plasmid segregation [9] . However , the identification and characterization of new members of this protein family in other organisms led to the discovery of more diverse functions and localizations . It is now recognized that sirtuins remove acetyl groups from lysines in nuclear , cytosolic and mitochondrial protein substrates [10] . Sirtuins are evolutionarily conserved enzymes present in all kingdoms of life , ranging from bacteria to higher eukaryotes including humans . Members of this family share a core domain of ~250 amino acids that exhibits 25–60% sequence identity between different organisms . Genes coding for seven sirtuins ( SIRT 1–7 ) have been found in the human genome , with subcellular distribution , substrate specificity , and cellular functions quite diverse [11] . Trypanosoma cruzi is a hemoflagellate protozoan parasite , branched early from the eukaryal lineage . It is an intracellular pathogen responsible for Chagas’ disease , or American Trypanosomiasis , a chronic infectious disease affecting 8 million people [12] . While Chagas disease is endemic in Latin America , a significant increase in confirmed cases of Chagas has recently been reported in the USA , Canada , Japan , Australia and Europe , indicating that it is an emerging disease [13] . Current therapies rely on a very small number of drugs , most of which are inadequate because of their severe host toxicity and numerous side effects . The identification of new biotargets is essential for the development of more efficient therapeutic alternatives . The structural basis for inhibition of sirtuins has been established through previous structural and functional studies [14–17] . Involvement of sirtuins in the cell cycle strongly suggests a role for these enzymes in cancer and the potential use of their inhibitors as anticancer drugs [18] . In addition , inhibition of sirtuins from Trypanosoma brucei and Leishmania ssp . showed promising results , indicating that these enzymes may be considered as targets for drug discovery in parasite infection [19–22] . T . cruzi belongs to the Kinetoplastida order , Trypanosomatidae family , as well as Trypanosoma brucei and Leishmania ssp . , and together they are termed TriTryps . Genes encoding three Sir2 related proteins ( SIR2RPs ) were found in the TriTryps . The trypanosomatid genes were designated SIR2-related proteins , SIR2RP1–3 . A previous phylogenetic analysis places SIR2RP1 in a group with ScSir2 , HsSIRT1 and HsSIRT2 , while SIR2RP2 and SIR2RP3 are more closely related to bacterial proteins and to HsSIRT4 and HsSIRT5 respectively [23] . However , a more recent extensive analysis places SIR2RP1 in the sirtuins subgroup Ib , together with cytoplasmic HsSIRT2 , and is now clearly differentiated from the nuclear HsSIRT1 [24] . SIR2RP1 from several Leishmania species and all three SIR2RPs from T . brucei have been characterized [16 , 23] . SIR2RP1 is found in cytoplasmic granules in different stages of L . major , L . infantum and L . amazonensis; and , under certain conditions , it is observed in the excreted/secreted fraction [25–27] . LiSIR2RP1 is also found associated with the cytoskeleton network and deacetylates α-tubulin , a function that resembles that of human SIRT2 and HDAC6 . In contrast , TbSIR2RP1 is a nuclear chromosome-associated protein . It is expressed throughout T . brucei life cycle , catalyses NAD+-dependent ADP ribosylation and deacetylation of histones and in the mammalian-infective bloodstream-stage controls DNA repair and repression of RNA polymerase I-mediated expression immediately adjacent to telomeres [16 , 23] . TbSIR2RP2 and 3 localize in the single mitochondrion of the parasite and it was reported that their interference do not produce growth or differentiation defects . In Trypanosoma cruzi , the gene coding for SIR2RP2 is lacking . This general landscape suggests that in spite of their sequence similarity , sirtuin variants from TriTryps have evolved to different functions . Here , we report the first characterization of the two sirtuins present in T . cruzi . In epimastigotes , TcSIR2RP1 localizes in the cytoplasm while TcSIR2RP3 localizes to the parasite’s single mitochondrion . Overexpression of TcSIR2RP1 causes no alteration to epimastigote growth , but it increases the number of trypomastigotes obtained in in vitro metacyclogenesis and the infectivity rate of Vero cells . In contrast , overexpression of TcSIR2RP3 slightly decreases epimastigote growth and the infectivity rate of Vero cells , it does not affect the in vitro differentiation to metacyclic trypomastigotes , and it increases the proliferation rate of intracellular amastigotes . Finally , overexpression of either of these sirtuins protects the parasite from the effect of sirtuin inhibitors .
All experiments were approved by the Institutional Animal Care and Use Committee of the School of Biochemical and Pharmaceutical Sciences , National University of Rosario ( Argentina ) ( File 6060/227 ) and conducted according to specifications of the US National Institutes of Health guidelines for the care and use of laboratory animals . Rabbits were only used for the production of polyclonal antibodies . The rabbits were immunized three times with the protein and an equal volume of Freund´s adjuvant , and bled two weeks after the final injection [28] . TcSIR2RP1 and TcSIR2RP3 genes were amplified using the following oligonucleotides: SIR2RP1ss ( 5’ AAAGGATCCATGAATCAAGATAACGCCAAC ) , SIR2RP1HAas ( 5’ AACTCGAGAGCATAATCCGGCACATCATACGGATATTTTCGGT CTGTCAG ) , SIR2RP3ss ( 5’ AAGGATCCATGAAGCCGCGGCGTCAGAT ) and SIR2RP3HAas ( 5′ AACTCGAGAGCATAATCCGGCACATCATACGGATACACCGCGT CTTGAAG ) . DNA purified from T . cruzi epimastigotes was used as template . The PCR products obtained with a proofreading DNA polymerase were inserted into pCR2 . 1-TOPO vector ( Invitrogen ) and sequenced . TcSIR2RP1 and TcSIR2RP3 coding regions were inserted into a pENTR3C vector ( Gateway system Invitrogen ) using the BamHI/XhoI restriction sites included in the oligonucleotides ( underlined ) and then transferred to pDEST17 ( Gateway system Invitrogen ) and pTcINDEX-GW vectors by recombination using LR clonase II enzyme mix ( Invitrogen ) . The pDEST17 constructs were transformed into Escherichia coli BL21 pLysS and recombinant proteins , fused to a six histidine tag , were obtained by expression-induction with 0 . 5 mM IPTG for 5 h at 37°C . The proteins were purified by affinity chromatography using a Ni-NTA agarose resin ( Qiagen ) following the manufacturer’s instructions . Rabbit polyclonal antisera against TcSIR2RP1 and TcSIR2RP3 were obtained by inoculating subcutaneously the recombinant proteins to these animals as described above . T . cruzi epimastigote forms ( Dm28c strain ) were cultured at 28°C in liver infusion tryptose ( LIT ) medium ( 5 g/L liver infusion , 5 g/L bacto-tryptose , 68 mM NaCl , 5 . 3 mM KCl , 22 mM Na2HPO4 , 0 . 2% ( w/v ) glucose and 0 . 002% ( w/v ) hemin ) supplemented with 10% ( v/v ) heat-inactivated FCS , 100 U/ml penicillin and 100 mg/l streptomycin . Cell viability was assessed by direct microscopic examination . For inducible expression of Sir2rp1-3 genes in the parasite , we first generated a cell line expressing T7 RNA polymerase and tetracycline repressor genes by transfecting epimastigotes with the plasmid pLew13 using a standard electroporation method . Briefly , epimastigote forms of T . cruzi Dm28c were grown at 28°C in LIT medium , supplemented with 10% FCS , to a density of approximately 3 × 107 cells/ml . Parasites were then harvested by centrifugation at 2 , 000 × g for 5 min at room temperature , washed once in PBS and resuspended in 0 . 35 ml of transfection buffer pH 7 . 5 ( 0 . 5 mM MgCl2 , 0 . 1 mM CaCl2 in PBS ) to a density of 1 × 108 cells/ml . Cells were then transferred to a 0 . 2 cm gap cuvette ( Bio-Rad ) and ~50 μg of DNA was added in a final volume of 40 μl . The mixture was placed on ice for 15 min and then subjected to 2 pulses of 450 V and 500 μF using GenePulser II ( Bio-Rad , Hercules , USA ) . After electroporation , cells were transferred into 3 ml of LIT medium containing 10% FCS , maintained at room temperature for 15 minutes and then incubated at 28°C . Geneticin ( G418; Life Technologies ) was added at a concentration of 200 μg/ml , and parasites were incubated at 28°C . After selection , transfected epimastigotes were grown in the presence of 200 μg/ml of G418 . This parental cell line was then transfected with pTcINDEX-GW constructs and transgenic parasites were obtained after 3 weeks of selection with 100 μg/ml G418 and 200 μg/ml Hygromycin B ( Sigma ) . To obtain metacyclic trypomastigotes , epimastigotes were differentiated in vitro following the procedure described by Contreras and coworkers [29] using chemically defined conditions ( TAU3AAG medium ) . Briefly , cells were washed with PBS and incubated in TAU medium ( 190 mM NaCl , 17 mM KCl , 2 mM MgCl2 , 2mM CaCl2 , 8 mM phosphate buffer pH 6 . 0 ) in the absence or presence of 0 . 25 μg/ml Tetracycline , reaching a density of 5 x 108 parasites/ml at 28°C for 2 hours . Then they were diluted 1:100 in TAU3AAG Medium ( TAU medium plus 10 mM Glucose , 2 mM L-Aspartic Acid , 50 mM L-Glutamic Acid and 10 mM L-Proline ) and incubated at 28°C for 72 hours , again in the absence or presence of Tetracycline . Finally , the parasites were fixed , stained with Giemsa , visualized with a Nikon Eclipse Ni-U microscope and counted using ImageJ software [30] . Only parasites with a fully elongated nucleus and a round kinetoplast at the posterior portion end of the parasite were considered metacyclic forms [31] . Five hundred parasites from each triplicate were counted and the experiment was repeated tree times . Vero cells were cultured in DMEM medium ( Life Technologies ) , supplemented with 2 mM L-glutamine , 10% FCS , 100 U/ml penicillin and 100 μg/ml streptomycin . Metacyclic trypomastigotes were obtained by spontaneous differentiation of epimastigotes at 28°C . Cell-derived trypomastigotes were obtained by infection with metacyclic trypomastigotes of Vero cell monolayers . After two rounds of infections , the cell-derived trypomastigotes were used for the infection and intracellular amastigotes proliferation experiments . Trypomastigotes were collected by centrifugation of the supernatant of previously infected cultures at 2 , 000 x g at room temperature for 10 minutes and incubated for 3 hours at 37°C in order to allow the trypomastigotes to move from the pellet into the supernatant . After this period , the supernatant was collected and trypomastigotes were counted in a Neubauer chamber . The purified trypomastigotes were pre-incubated in the presence or absence of 0 . 25 μg/ml Tetracycline for 3 hours and then used to infect new monolayers of Vero cells at a ratio of 10 parasites per cell . After 6 h of infection at 37°C , the free trypomastigotes were removed by successive washes using saline solution . Cultures were incubated in complete medium with or without Tetracycline ( 0 . 25 μg/ml ) for 2 days post-infection . Infections were performed in DMEM supplemented with 2% FCS . Cells were then fixed in methanol and the percentage of infected cells and the mean number of amastigotes per infected cell , were determined by counting the slides after Giemsa staining using a Nikon Eclipse Ni-U microscope , by counting ~1000 cells per slide . The significances of the results were analyzed by a two-way ANOVA using GraphPad Prism version 6 . 0 for Mac . Results are expressed as means ± SEM of triplicates , and represent one of three independent experiments performed . Exponentially growing epimastigotes were washed twice with cold PBS , pellets were resuspended in urea lysis buffer ( 8 M Urea , 20 mM Hepes pH 8 , 1 mM phenylmethylsulphonyl fluoride ( PMSF ) , and Protease Inhibitor Cocktail set I , Calbiochem ) , incubated at room temperature for 20 minutes and boiled for 5 minutes with protein loading buffer . Insoluble debris was eliminated by centrifugation . The same procedure was applied to amastigote and trypomastigote cellular pellets . Transfected Dm28c epimastigotes in exponential growth phase were centrifuged for 10 min at 2 , 000 x g and washed twice in homogenization buffer ( 25 mM Tris-HCl pH 8 , 1 mM EDTA , 0 . 25 M sucrose , 1 mM PMSF ) . Subcellular fractions were obtained following the procedure described by Opperdoes and coworkers [32] . The parasites were grinded in a pre-chilled mortar with 1 x wet weight silicon carbide until no intact cells were observed under the light microscope . The lysate was diluted and centrifuged at 100 x g for 10 min to remove the silicon carbide . Unbroken cells , nuclei and debris were sedimented at 1 , 000 x g for 10 min ( Fraction N ) . From the resulting soluble extract a large-granule fraction ( LG ) was separated at 5 , 000 x g for 15 min , a small-granule fraction ( SG ) at 20 , 000 x g for 20 min and microsomal fraction ( M ) at 139 , 000 x g for 1 h . All the sediments were resuspended in urea lysis buffer . Protein extracts were fractioned in SDS-PAGE and transferred to nitrocellulose membranes . Transferred proteins were visualized with Ponceau S . Membranes were treated with 10% non-fat milk in PBS for 2 hours and then incubated with specific antibodies diluted in PBS for 3 hours . Antibodies used were: rat monoclonal anti-HA ( ROCHE ) , rabbit polyclonal anti-TcSIR2RP1 and anti-TcSIR2RP3 , rabbit polyclonal anti-Acetyl-lysine ( Millipore ) , mouse monoclonal anti-acetylated α-tubulin clone 6-11B-1 ( Sigma Aldrich ) , mouse monoclonal anti-trypanosome α-tubulin clone TAT-1 , rabbit polyclonal anti-T . cruzi mitochondrial Malate Dehydrogenase ( TcMDHm ) , rabbit and mouse polyclonal anti-T . cruzi Tyrosine Amine Transferase ( TcTAT ) , mouse polyclonal anti-T . cruzi Aspartate Transaminase ( TcASAT ) and rabbit polyclonal anti-T . cruzi Bromodomain Factor 2 ( TcBDF2 ) . Bound antibodies were detected using peroxidase labeled anti-mouse , anti-rabbit IgGs ( GE Healthcare ) or anti-rat IgG ( Thermo Scientific ) and developed using ECL Prime kit ( GE Healthcare ) according to manufactures protocol . Trypomastigotes and exponentially growing epimastigotes were centrifuged , washed twice in PBS , settled on polylisine-coated coverslips and fixed with 4% para-formaldehyde in PBS at room temperature for 20 minutes . For the mitochondrial staining , parasites were resuspended in PBS and incubated with 1 μM MitoTracker ( Invitrogen ) for 30 minutes at 28°C , washed twice in PBS and fixed with 4% para-formaldehyde . Fixed parasites were washed with PBS and permeabilized with 0 . 1% Triton X-100 in PBS for 10 minutes . After washing with PBS , parasites were incubated with the appropriate primary antibody diluted in 5% BSA in PBS for 2 hours at room temperature . In colocalization experiments both antibodies were incubated together . Non-bound antibodies were washed with 0 . 01% Tween 20 in PBS and then the slides were incubated with fluorescent-conjugated anti-mouse ( FITC , Jackson Immuno Research ) or anti-rat ( FITC , Life Technologies ) and anti-rabbit ( Cy3 , Life Technologies ) IgG antibodies and 2 μg/ml of DAPI for 1 hour . The slides were washed with 0 . 01% Tween 20 in PBS and finally mounted with VectaShield ( Vector Laboratories ) . To analyze intracellular amastigotes , Vero cells monolayers were grown on coverslips and infected with T . cruzi trypomastigotes as described above . Two days post-infection cultures were washed with PBS and fixed with methanol at room temperature for 3 minutes . The same procedure described above was followed for immunodetection . Images were acquired with a confocal Nikon Eclipse TE-2000-E2 microscope using Nikon EZ-C1 Software . Adobe Photoshop CS and ImageJ software were used to process all images . To determine the IC50 values of the sirtuin inhibitors , epimastigotes of T . cruzi Dm28c strain were cultured at 28°C in liver infusion tryptose medium ( LIT ) supplemented with 10% FCS in the absence or presence of Nicotinamide , Cambinol and Ex-527 ( Sigma ) at various concentrations , in triplicates . Cell growth was determined after culture for 72 hours by counting viable forms in an automatized hemocytometer adapted to count epimastigotes ( WL 19 Counter AA , Weiner Lab ) . Then , Dm28c wt , Dm28c pTcINDEXGW-SIR2RP1HA and Dm28c pTcINDEXGW-SIR2RP3HA strains ( uninduced and induced with 0 . 5 μg/ml Tetracycline ) , were cultured at 28°C in LIT with FCS in the absence or presence of the sirtuin inhibitors at concentrations above their IC50 values . Experiments were performed in triplicate , and at least three independent experiments were performed . Data are presented as the mean ± SEM . Statistical analysis of the data was carried out using two-way ANOVA and unpaired Mann-Whitney , two-tailed Student t test . Differences between the experimental groups were considered significant as follows: p<0 . 05 ( * ) , p<0 . 005 ( ** ) , p<0 , 001 ( *** ) and p<0 . 0001 ( **** ) . To determine the IC50 values , we used nonlinear regression on Prism 6 . 0 GraphPad software . Student’s t test was applied to ascertain the statistical significance of the observed differences in the IC50 values .
Two protein coding sequences ( TcCLB . 507519 . 60 and TcCLB . 506559 . 80 ) corresponding to Sir2 related proteins were identified in the T . cruzi genome , termed TcSIR2RP1 and TcSIR2RP3 respectively ( http://www . tritrypdb . org/tritrypdb/ ) . TcSir2rp1 and TcSir2rp3 encode proteins of 359 and 241 amino acids , with predicted molecular weights of ~ 39 . 6 and 26 . 8 kDa and pIs of 6 . 39 and 6 . 51 , respectively . The alignment of T . cruzi sirtuins with human SIRTs and ScSir2 ( S1 Fig ) shows that although TcSIR2RPs lack the N-terminal portion , which is required for nucleolar localization in ScSir2 , they contain a complete catalytic domain ( Pfam: PF02146 ) . SIR2RP1 contains a Serine-rich motif towards the C-terminus and one of the Cys residues from the zinc-binding motif ( CX2CX20CX2C type ) is absent in SIR2RP3 . The GAD and NID motifs as well as other residues important for catalysis are conserved ( HG , arrowheads in S1 Fig ) . The catalytic domain of TcSIR2RP1 and TcSIR2RP3 share a sequence identity/similarity of 17 . 5%/26 . 3% with each other and 23 . 1%/33 . 9% and 20 . 8%/ 33 . 7% with ScSIR2 respectively . Similarly to what Greiss and Gartner observed [24] , TcSIR2RP1 grouped with the cytoplasmatic human SIRT2 whereas TcSIR2RP3 is more related to mitochondrial HsSIRT5 and bacterial sirtuins ( S2 Fig ) . Despite the discrepancies observed for each sirtuin from different Tritryp species regarding their localization and function , they seem to be conserved at the sequence level . In order to evaluate TcSIR2RP1 and TcSIR2RP3 expression in T . cruzi , antibodies were raised against the recombinant proteins and purified by affinity chromatography . After confirming the specificity of the antibodies ( S3 Fig showed a single band of the expected molecular weights ) , they were used in Western blots to test total lysates of epimastigotes , amastigotes and trypomastigotes . As can be observed in Fig 1 , the expression of sirtuins is developmentally regulated throughout T . cruzi life cycle . TcSIR2RP1 shows similar expression levels in epimastigotes and amastigotes , but lower in trypomastigotes . TcSIR2RP3 expression levels are higher in epimastigotes than in amastigotes and it is not detected in trypomastigotes under the conditions assayed . Overexpression of TcSIR2RP1 and TcSIR2RP3 enzymes was performed using the T . cruzi inducible vector pTcINDEXGW [33] . Epimastigote cell lines expressing each sirtuin with a C-terminal HA tag under the control of a Tetracycline-regulated promoter were generated ( Materials and Methods ) . The induction of the expression by Tetracycline was tested by western blot ( Fig 2A and 2B ) and immunofluorescense ( Fig 2C ) . Western blot analysis of whole-cell extracts with rat monoclonal anti-HA antibodies revealed the expression of both constructs after the addition of Tetracycline , at their expected molecular weights . No leaky expression was observed in the uninduced parasite lines ( Fig 2A ) . The western blots with the specific antibodies against TcSIR2RP1 and TcSIR2RP3 show a high degree of overexpression ( 20-fold ) in the induced lines ( Fig 2B ) . We also tested the inducible expression of the sirtuins in intracellular amastigotes and trypomastigotes by western blot of total lysates ( Fig 3A ) and immunofluorescense ( Fig 3B ) with anti-HA ( quantification of results from Fig 3A are shown in S4 Fig ) . The tagged sirtuins are expressed only in the presence of Tetracycline , throughout T . cruzi life cycle . Different technical approaches were performed to determine the localization of TcSIR2RPs , using the specific antibodies raised in rabbit against the two recombinant proteins and commercial anti-HA monoclonal antibodies . Confocal immunocolocalization microscopies performed with cytosolic ( anti-TAT ) [34] and mitochondrial ( MitoTracker ) markers , together with anti-TcSIR2RP1 and anti-TcSIR2RP3 , showed that TcSIR2RP1 co-localizes with TAT ( cytosol ) and TcSIR2RP3 with MitoTracker ( Fig 4A ) . In parallel , the subcellular distribution of tagged sirtuins was analyzed by immunofluorescense of induced epimastigotes of each cell line with cytosolic ( anti-TAT ) and mitochondrial ( anti-MDHm ) markers together with anti-HA ( Fig 4B ) . As can be observed in Fig 4B , SIR2RP1-HA colocalized with TAT and SIR2RP3-HA with MDHm , supporting the results obtained with specific antibodies . To study in further detail their localizations , we performed subcellular fractionation by differential centrifugation of the transfected lines . The fractions obtained were analyzed by western blot with different subcellular markers ( cytosolic TAT , mitochondrial MDH and nuclear BDF2 [35] ) , and with anti-HA ( Fig 4C ) . The nuclear marker was enriched in fraction N , the mitochondrial marker in fractions N and LG , and the cytosolic marker in fraction S , as reported by Opperdoes et al [32] . In agreement with our previous results , SIR2RP1-HA exhibits a cytosolic pattern while SIR2RP3-HA exhibits a mitochondrial one . To test the deacetylation activity of TcSIR2RPs , we performed western blot of uninduced and induced total lysates of each cell line with anti-Acetyl-lysine antibodies . Fig 5A shows the amount of protein loaded for each condition . Overexpression of both sirtuins reduced the acetylation levels of specific proteins ( Fig 5B ) . The differentially acetylated proteins are depicted with arrowheads . It is worth noticing that the deacetylated proteins are different for each overexpressed sirtuin . Alpha-tubulin is one of the most abundant acetylated proteins in trypanosomes . In fact , the rate of acetylated/non-acetylated α-tubulin in trypanosomatids is higher than in other eukaryotic cells like yeast or mammalian cells [36–37] . To test if any of the TcSIR2RPs can deacetylate α-tubulin , we measured acetylated α–tubulin in uninduced and induced parasites by western blot analysis . When normalized to total α-tubulin , the results reflect significant diminutions of the acetylated form of the protein of 27% and 35% in TcSIR2RP1HA and TcSIR2RP3HA overexpressing lines respectively ( Fig 5C ) . As already mentioned , deacetylation of α-tubulin mediated by SIR2RP1 was reported in Leishmania . However , even though our results are very confident , we consider that we cannot be conclusive enough to assign to TcSIR2RP1 nor to TcSIR2RP3 the function of being the T . cruzi tubulin deacetylase . The observed hypoacetylation could be an unspecific result due to the overexpression of a sirtuin . Furthermore , it has been demonstrated that in mammalian cells , tubulin is deacetylated not only by SIRT2 , but also by the non-sirtuin deacetylase HDAC6 [38] . Since there are more than one deacetylases with similarity to HDAC6 in trypanosomatids , a deeper study of the whole set of deacetylase enzymes is needed to determine the most relevant tubulin deacetylase in this organism . We monitored the effect of sirtuins overexpression on epimastigote growth by counting cell numbers daily after protein induction . Fig 6 shows that Dm28c TcSIR2RP1HA cell line grew at similar rates in the absence and presence of Tetracycline ( which was re-added every 5 days ) , but those harboring TcSIR2RP3HA showed a delay in the growth rate when induced . Even more surprising is the fact that the TcSIR2RP3HA expressing line reached stationary phase at a smaller number of parasites per ml . This culture continues in a stationary phase for the same period of time as the uninduced line . This phenomenon needs to be further studied in order to be completely understood , and the existence of a quorum sensing mechanisms recently described in T . brucei opens a novel possibility of interpretation [39] . Sirtuins are considered as fundamental life sustaining biocatalysts and under various conditions were found to be pro-survival [40] . Therefore , there is an increasing interest in sirtuins as therapeutic targets . Several structures of complexes involving sirtuins and their inhibitors have been reported [17 , 41] . In this work , we tested three sirtuin inhibitors: Nicotinamide ( NAM ) [42] , Cambinol and Ex-527 [43] . All of them inhibited T . cruzi Dm28c epimastigotes growth in a concentration-dependent manner in axenic cultures . The IC50 values obtained for each inhibitor are shown in Table 1 . Ex-527 IC50 ( 206 . 2 μM ) is significantly higher compared to the other tested sirtuin inhibitors . Since , Ex-527 is a potent and selective HsSIRT1 inhibitor , with a reported IC50 of 0 . 1–1 μM [43] , the low toxicity we observed might indicate that neither of the sirtuins present in T . cruzi is related to HsSIRT1 , consistent with our phylogenetic analysis . On the contrary , NAM and Cambinol exhibited IC50 values similar to those obtained for purified sirtuins of other organisms such as: NAM IC50 for PfSIR2: 51 . 2 μM [44]; NAM IC50 for hSIRT5: 46 . 6 μM , hSIRT1: 50–100 μM , hSIRT3: 30 μM; Cambinol IC50 for hSIRT1: 56 μM , hSIRT2: 59 μM [45] . We considered the possibility that sirtuin-overexpressing lines might be “protected” or less sensitive to the sirtuin inhibitors . To test this hypothesis , both parasite lines were treated with the inhibitors , in the absence and presence of Tetracycline . As seen in Fig 7 , overexpression of sirtuins protects epimastigotes from the growth inhibition of NAM and Cambinol . Moreover , the treatment with these inhibitors reversed the growth defect of the induced TcSIR2RP3HA cell line . In contrast , Ex-527-driven growth inhibition may be due to a pleiotropic effect and not to a reduction of the parasite’s sirtuin activities , since the inhibitory effect on trypanosomes is only observed at high concentrations . We also calculated the IC50 values of NAM and Cambinol for the overexpressing strains . The values obtained are shown in Table 1 , as expected they are higher than for the wild type parasites . In vitro metacyclic trypomastigotes were produced from epimastigotes using TAU medium , in the absence ( -Tet ) or presence ( +Tet ) of Tetracycline . TcSIR2RP1 overexpression resulted in a meaningful increase of metacyclogenesis ( 59% ) , whereas TcSIR2RP3 strain showed similar levels of metacyclic trypomastigotes formation in both the induced and uninduced conditions ( Fig 8 ) . To study the importance of sirtuins expression in trypomastigotes´ infectivity and in the replicative form present inside the mammalian host , we investigated how the transgenic lines induced with Tetracycline performed in vitro for invasion and replication in host cells . First , we performed the experiment with Dm28c wild type parasites to rule out any undesired effect of the Tetracycline treatment . Indeed , there was no significant difference in the infectivity rate nor in the number of amastigotes/cell ( S5 Fig ) . Trypomastigotes were pre-incubated in the presence or absence of 0 . 25 μg/ml Tetracycline and NAM ( 100 μM ) and then used to infect Vero cells at a ratio of 10 parasites per cell . After 6 h of infection at 37°C , the free trypomastigotes were washed out and replaced by complete medium alone or with Tetracycline ( 0 . 25 μg/ml ) for 2 days post-infection . Microscopic observation of Vero cells stained with Giemsa showed that treatment of uninduced T . cruzi trypomastigotes with NAM [ ( -/- , +NAM ) vs ( -/- ) ] caused a significant reduction in the percentage of infected cells ( as previously reported by Soares and coworkers [42] ) ( Fig 9A ) . The overexpression of both sirtuins protected the trypomastigotes from the negative effect of NAM [ ( +/+ , +NAM ) vs ( -/- , +NAM ) ] . These results suggest that sirtuin activity is necessary for an effective infection of mammalian cells . To analyze the effect of sirtuin overexpression on the infectivity rate of trypomastigotes , we focused on the condition in which the expression was induced only in the trypomastigote stage during infection ( +/- ) . As can be seen in Fig 9A , overexpression of TcSIR2RP1HA increased the infectivity rate of trypomastigotes [ ( +/- ) vs ( -/- ) ] , while overexpression of TcSIR2RP3HA slightly diminished it [ ( +/- ) vs ( -/- ) ] . To test the effect of sirtuins overexpression on the proliferation of intracellular amatigotes ( Fig 9B ) , we only added Tetracycline after the infection , when the trypomastigotes were washed out , for 48 hours post-infection ( -/+ ) . The number of amastigotes per infected cell is slightly decreased by the overexpression of TcSIR2RP1HA , but increased by TcSIR2RP3HA [ ( -/+ ) vs ( -/- ) ] . The reduction in the number of amastigotes per cell observed when expression of TcSIR2RP3 was induced during the infection [ ( +/- ) vs ( -/- ) ] , might be a consequence of the diminished infectivity or it might indicate that the overexpression of this enzyme at the trypomastigote stage results in an inefficient differentiation to amastigotes , thus delaying the amastigotes´ replication . In contrast , when inducing at all times ( +/+ ) , the overexpression of amastigotes increases the proliferation rate hiding the trypomastigote to amastigote differentiation delay .
We present herein the first experimental characterization of Trypanosoma cruzi sirtuins TcSIR2RP1 and TcSIR2RP3 . The expression of these enzymes is developmentally regulated throughout T . cruzi life cycle . TcSIR2RP1 is highly expressed in epimastigotes and amastigotes , but at lower levels in trypomastigotes . On the other hand , TcSIR2RP3 expression levels are higher in epimastigotes than in amastigotes , and it seems not to be expressed in trypomastigotes . The fact that the two sirtuins are differentially expressed along the life cycle of the parasite suggests that acetylation levels could play a role in the regulation of the biology of the parasite forms . It is remarkable that the life cycle pattern expression of Tetracycline induced/T7 transcribed sirtuins is similar to those of the wild type enzymes . This observation suggests that the protein levels could be regulated by a post-transcriptional mechanism independent of the 3´ and 5’ non-coding regions , which are absent in the pTc-INDEX-GW constructions or by a post-translational mechanism . Our results clearly demonstrate that TcSIR2RP1 and TcSIR2RP3 are , respectively , cytoplasmic and mitochondrial enzymes . These observations were expected , since both proteins lack the N-terminal portion responsible for the nuclear localization of ScSir2 and other related sirtuins . However , the possibility that under certain conditions , the trypanosomal sirtuins could be imported temporarily to the nucleus by an alternative targeting pathway cannot be completely ruled out . Taken together , our data suggest that sirtuin activity is important for the proliferation of T . cruzi replicative forms , for the host cell-parasite interplay , and for differentiation among life-cycle stages; but each one performs different roles in most of these processes . Considering its cellular localization , TcSIR2RP1 seems to be functionally more related to Leishmania than to T . brucei ortholog , even though TbSIR2RP1 is more similar at the sequence level ( 68% identity ) than LmSIR2RP1 ( 55% identity ) . The fact that T . brucei has a nuclear sirtuin that is absent in the other TriTryps can be explained by some well-known differences existing among these species . TbSIR2RP1 participates at the epigenetic-mediated silencing of RNA polymerase I-transcribed telomeric regions , but nothing similar occurs in T . cruzi or in Leishmania . In Plasmodium falciparum , a nuclear sirtuin ( PfSIR2A ) , phylogenetically unrelated to TbSIR2RP1 , is also implicated in telomeric gene silencing . Taken together , these results suggest that the participation of a sirtuin in histone deacetylation represents the exception , associated to telomeric gene silencing , rather than the rule of the function of sirtuins in these organisms , and it could be an example of convergent evolution . In spite of their cellular localization and way of action , sirtuins are considered pro-survival regulators of metabolism and lifespan . These general functions are related with the use of NAD+ as substrate , which together with acetyl-CoA , the acetyltransferases substrate , are considered sensors of the energetic state of the cell . Nuclear sirtuins , like HsSIRT1 , regulate transcriptional response to starvation or redox stress and under certain conditions , cytoplasmic and mitochondrial sirtuins are imported to the nucleus with the same purpose ( Reviewed in [46] ) . Since transcriptional regulation is absent in T . cruzi , it is reasonable to think that only non-nuclear functions will be found for these enzymes . One of the functions of cytoplasmic HsSIRT2 is to deacetylate the enzyme phosphoenolpyruvate carboxykinase ( PEPCK ) , increasing its stability , upon glucose deprivation [47] . Under caloric restriction , human mitochondrial sirtuins may also regulate gluconeogenesis from amino acids through glutamate dehydrogenase ( GDH ) , an enzyme that converts glutamate to α-ketoglutarate , thereby controlling glucose production via the TCA cycle [48–50] . HsSIRT3 and HsSIRT4 modulate the activity of GDH through deacetylation and ADP-ribosylation , respectively ( even though the existence of gluconeogenesis was only proved in amastigotes from Leishmania , it was already proposed that this pathway should be present in all trypanosmatids [51] ) . Human SIRT3 also decreases reactive oxygen species ( ROS ) production by stimulating superoxide dismutase 2 ( SOD2 ) , and enhances cellular respiration by increasing the activities of complex I , complex II ( via succinate dehydrogenase ( SDH ) ) , complex III and isocitrate dehydrogenase 2 ( IDH2 ) , affecting both glucose and lipid metabolism . Finally , mitochondrial HsSIRT3 stimulates β-oxidation and ketone body formation by targeting and activating long-chain acyl CoA dehydrogenase ( LCAD ) and 3-hydroxy-3-methylglutaryl-CoA synthase 2 ( HMGCS2 ) , respectively [52–53] . Although more specific research is needed to determine whether trypanosome sirtuins share these functions , some of our results can be interpreted under this general rationale: TcSIR2RP1 improves metacyclogenesis , a differentiation process induced by starvation of the parasites [54–55] . The increased infectivity of this strain resembles that observed for LmSIR2RP1 by Sereno and coworkers [56] . As this sirtuin was detected in the excreted/secreted material of parasites , they constructed LmSIR2RP1-overexpressing fibroblasts , which were more permissive towards Leishmania invasion than control ones . These results suggest a function for parasite cytoplasmic sirtuins at the host-parasite interplay . However , the secretion of TcSIR2RP1 is yet to be confirmed . In contrast , TcSIR2RP3 seems not to be implicated in metacyclogenesis or cell infection but improves the replication of intracellular amastigotes , which occurs in an oxidant environment , suggesting the participation of this enzyme in redox stress response . The activity of sirtuins in protecting cells from stress and starvation and so improving life span , led to the establishment of their role in cell proliferation . Currently , many sirtuin inhibitors are being assayed against different types of cancer . Conversely , sirtuin activators are considered as potential anti-aging drugs . It has also been proposed that sirtuins are promising targets for the development of anti-trypanosomal , anti-plasmodial and anti-leishmanial drugs and many well-characterized sirtuin inhibitors have already shown anti-parasitic activity [21–22 , 40 , 57] . In addition , an in silico structural and surface analysis of trypanosomal and human sirtuins determined potentially important structural differences in the corresponding inhibitor binding domains , indicating a possible selectivity of an inhibitor for a specific protein [58] . In a very recent in silico study , Sacconnay and coworkers [59] assayed a list of 50 phytochemicals previously described as anti-trypanosomals by docking into TcSIR2RP1 and TcSIRRP3 , revealing that the activity of four of these compounds could be explained by the inhibition of the sirtuins activity . The results presented herein , contribute to the knowledge of these enzymes localization and function , and our TcSIR2RP1HA and TcSIR2RP3HA overexpressing T . cruzi strains can be useful tools for experimental screening of trypanosomatid sirtuin inhibitors .
|
Sirtuins are a family of deacetylases , evolutionary conserved from bacteria to mammals . They participate in the regulation of a wide range of nuclear , cytoplasmic and mitochondrial pathways , and are considered pro-life enzymes . In the last years the search for sirtuin inhibitors was a very active field of research , with potential applications in a large number of pathologies , including parasitic diseases . We are interested in the study of the two sirtuins present in the protozoan parasite Trypanosoma cruzi , being our objective to understand their function . First , we determined the localization of these enzymes in the parasite: TcSIR2RP1 is a cytoplasmic enzyme and TcSIR2RP3 localizes in the mitochondrion . When we overexpress cytoplasmic TcSIR2RP1 , the transgenic parasites differentiate to metacyclic trypomastigotes and infect mammalian cells more efficiently . In contrast , the overexpression of mitochondrial TcSIR2RP3 does not affect metacyclogenesis but modifies epimastigotes growth and slightly increases the proliferation of the parasite in the intracellular stage . We also used these transgenic lines to test their sensibility to previously described sirtuin inhibitors .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"Methods",
"Results",
"Discussion"
] |
[] |
2015
|
Overexpression of Cytoplasmic TcSIR2RP1 and Mitochondrial TcSIR2RP3 Impacts on Trypanosoma cruzi Growth and Cell Invasion
|
The ability of adherent cells to form adhesions is critical to numerous phases of their physiology . The assembly of adhesions is mediated by several types of integrins . These integrins differ in physical properties , including rate of diffusion on the plasma membrane , rapidity of changing conformation from bent to extended , affinity for extracellular matrix ligands , and lifetimes of their ligand-bound states . However , the way in which nanoscale physical properties of integrins ensure proper adhesion assembly remains elusive . We observe experimentally that both β-1 and β-3 integrins localize in nascent adhesions at the cell leading edge . In order to understand how different nanoscale parameters of β-1 and β-3 integrins mediate proper adhesion assembly , we therefore develop a coarse-grained computational model . Results from the model demonstrate that morphology and distribution of nascent adhesions depend on ligand binding affinity and strength of pairwise interactions . Organization of nascent adhesions depends on the relative amounts of integrins with different bond kinetics . Moreover , the model shows that the architecture of an actin filament network does not perturb the total amount of integrin clustering and ligand binding; however , only bundled actin architectures favor adhesion stability and ultimately maturation . Together , our results support the view that cells can finely tune the expression of different integrin types to determine both structural and dynamic properties of adhesions .
As the linker between cytoskeletal adhesion proteins and extracellular matrix ligands , integrins play a vital role in the formation of adhesions and profoundly influence different phases of cell physiology , such as spreading , differentiation , changes in shape , migration and stiffness sensing [1–5] . Integrins are large heterodimeric receptors , with a globular headpiece projecting more than 20 nm from the cell membrane , two transmembrane helices , and two short cytoplasmic tails that bind cytoskeleton adhesion proteins ( see Fig 1A ) . In order to form adhesions , integrins undergo lateral diffusion on the cell membrane , switch conformation from bent to extended , and change chemical affinity for extracellular matrix ( ECM ) ligands , EIL . Integrins also assemble laterally , owing to interactions with talin [6 , 7] , kindlin [8] , or glycocalyx [9] , and can grow nascent adhesions into mature adhesions [10–12] . Integrin diffusion , activation , ligand binding , and clustering occur at the individual protein scale , but their effects can also be reflected on the cellular scale , resulting in a multiscale biological process . Simulations of adhesion assembly based on all-atom approaches are too detailed and computationally demanding to capture adhesion formation from multiple integrins . Instead , highly coarse-grained , CG , approaches based on Brownian Dynamics can condense the description of individual proteins into a few interacting CG “beads” that can recapitulate the emergent dynamics of complex biological systems from its individual components ( see , e . g . , Refs [13–17] for the example of cytoskeleton networks ) . Nascent adhesions are complex biological systems that form near the leading edge of protruding cells , appearing as spots of about 0 . 1 μm in diameter , with lifetimes of 2–10 min ( Fig 1B ) [18–22] . Unfortunately , the small size and short lifetime of nascent adhesions have made it challenging to study them experimentally . Among 24 different integrin isoforms , the αvβ3 and α5β1 integrins , have important , but potentially separate roles in the assembly of adhesions and the physiology of many cell types [23–32] . Nanoscale differences in physical properties between αvβ3 and α5β1 integrins can determine how nascent adhesions assemble [33] , their organization [34–36] , transmitted traction [37] and lifetime [38] , on account of their different properties . For example , it has been reported that the rate of integrin activation , ka , determines the number of integrins per adhesion [21 , 39] , while lateral clustering , or avidity , EII , increases the size of individual adhesions [40–42] . Single-protein tracking experiments combined with super-resolution microscopy and computational methods have helped extract physical properties of different integrin types . β-1 and β-3 integrins were found to have diffusion coefficients of 0 . 1 and 0 . 3 μm2/s , respectively [43] . β-1 integrins also maintain their active conformation longer than β-3 integrins . Free-energy energy differences between active and inactive states revealed activation rates for β-3 integrins about 10-fold higher that β-1 integrins [44 , 45] . The intrinsic ligand binding affinity , EIL , of β-1 integrins for soluble fibronectin is about 10–50 fold higher than β-3 integrins , spanning an overall range for the two integrins of 3–9 kBT [46] . β-1 integrins display a catch bond and adhesion strength-reinforcing behavior and are stationary within adhesions [47–49] . β-3 integrins , on the other hand , rapidly transit from closed to open conformations , break their bonds from ligands more easily under modicum tensions , and undergo rearward movements within adhesions [26 , 38] . How these differences in diffusion , rate of activation , ligand binding affinity , and bond dynamics reflect on the assembly of nascent adhesions and on the probability of adhesion maturation remains elusive . In this paper , we show that mixed populations of β-1 and β-3 integrins localize to both nascent and mature adhesions , suggesting that there could be important interactions between the two types of integrins . To address this question , we have developed a highly CG model of adhesion formation , based on Brownian Dynamics ( Fig 2 ) and study how nanoscale physical properties of different types of integrins interplay in the assembly of nascent adhesions . The CG model treats individual integrins as point particles within an implicit cell membrane and includes actin filaments as explicit semiflexible polymers ( Fig 2A ) . By incorporating nanoscale physical properties of individual integrins , sequential interactions and feedback mechanisms between integrin , ligands and actin filaments ( Fig 2B–2D ) , the model is used to characterize the formation of micrometer-size adhesions at the cell periphery in a multiscale fashion . Our calculations show that integrins with high EIL and enhanced bond lifetimes , such as β-1 integrins , facilitate ligand binding , transmission of traction stress , and engagement of actin networks . By contrast , integrins with low EIL and lower ligand bond lifetimes , such as β-3 integrins , are correlated with clustering , repeated cycles of diffusion and immobilization , and weak engagement of actin filaments . The architecture of actin filaments does not impact the amount of ligand binding and integrin clustering , but determines the probability of adhesions maturation , consistent with previous experimental findings [50] . Collectively , our data reveal important insights into adhesions assembly that are currently very challenging to obtain experimentally . The data supports the general view that cells , by controlling physical nanoscale properties of integrins via expression of specific types , can regulate structural , dynamical , and mechanical properties of adhesions .
Motivated by our recent work on integrin catch-bonds regulating cellular stiffness sensing [5] , we sought to investigate how interactions between different integrins could affect adhesion formation . Immunostaining in Human Foreskin Fibroblasts ( HFF ) for actin and either β-1 or β-3 integrins revealed that both types of integrins localize in nascent adhesions at the cell leading edge and in mature adhesions at the end of actin stress fibers ( Fig 1B ) . This suggests that potential interactions between the different adhesion populations could be important during adhesion formation . To address this question , we developed a computational model to investigate how the nanoscale properties of different integrins affect adhesion formation and stability . Since β-1 and β-3 integrins differ in ligand binding affinity , EIL , and strength of pairwise interactions , EII [44 , 45] , we use the CG model to test how variations in EII and EIL impact adhesion assembly in terms of the amount of integrin clustering , ligand binding , and spatial arrangement of adhesions . Different morphological arrangements of integrin adhesions are detected ( Fig 3A–3C ) . For high EII and low EIL , clustering is promoted ( Fig 3D ) , but only a few integrins are bound to ligands ( Fig 3E ) , resulting in few large integrin clusters ( Fig 3A ) . Conversely , for low EII and high EIL , only a few integrins cluster ( Fig 3D ) while ligand-binding is promoted ( Fig 3E ) , resulting in many ligand-bound integrins and few small integrin clusters ( Fig 3B ) . When EII and EIL have intermediate values , a mix of big clusters of integrins that are weakly bound to the substrate and smaller , ligand-bound clusters co-exist ( Fig 3C ) . By systematically varying EII and EIL , morphological regions differing in size and number of ligand-bound integrins versus clusters were precisely identified . A region of few large clusters exists for EII >3 kBT and EIL < 3 kBT; a region of many small adhesions exists for EII < 3 kBT and EIL > 3 kBT; the rest of the parameter space shows co-existence of intermediate-size clusters and ligand-bound integrins ( Fig 3D–3E ) . The fraction of ligand-bound integrins increases with EIL and is independent from EII . ( Fig 3E ) . By contrast , clustering is not independent from EIL and is promoted when EIL is low ( Fig 3D ) . In the model , when active , integrins can bind free ligands and cluster , when in close proximity of a ligand or another active integrin , respectively . Since the number of ligands is higher than the number of integrins , the probability for an integrin to find a free ligand is higher than that of finding an active integrin . Therefore , clustering increases less with EII when EIL is high than when EIL is low ( Fig 3D ) . This indicates that integrin clustering and ligand binding are competing mechanisms . Together , our results show that different arrangements of nascent adhesions can be achieved depending on EII and EIL . When we use high EII and low EIL , as for β-3 integrins , clustering is enhanced , and ligand binding reduced; when we use high EIL and low EII , as for β-1 integrins , clustering is reduced , and ligand-binding promoted . Thus , the competition between clustering and ligand binding can be determined by the integrin type . However , β-1 and β-3 integrins also differ in their rates of activation , which can lead to differences in this competition , by promoting clustering at high EIL . Therefore , we next aimed to understand how activation rates , combined with variations in EII and EIL , impact clustering and ligand binding . Competition between integrin clustering and ligand binding can be determined by the difference in activation rate between β-1 and β-3 integrins . By varying ka from 0 . 005 s-1 to 0 . 5 s-1 , our model shows that both clustering and ligand binding are promoted ( Fig 4A and 4B ) . Using EII = 5 kBT and varying EIL from 3 kBT to 11 kBT , clustering is independent from EIL ( Fig 4A ) , while overall ligand binding increases with EIL ( Fig 4B ) . Clustering is mostly set by the strength of pairwise interactions between integrins , EII . It can be promoted by low EIL . and high ka , leading to a higher number of integrins able to diffuse and cluster ( Fig 4A ) . Ligand binding is proportional to EIL at all ka . In experiments , variations in integrin activation rate are tied to variations in ligand binding affinity , making it unclear whether it is ka or EIL that determines organization of nascent adhesions . Our model shows that the rate of integrin activation set the level of the competition between ligand binding affinity and strength of pairwise interactions ( Fig 4A ) . Experimentally , Mn2+ or antibodies are typically used to modulate ligand binding affinity [51–54] . Both of these approaches , however , not only increase ligand binding affinity , but also the lifetime of the ligand bond . The increase of the ligand bond lifetime can be formally represented using a catch-bonds [55] , where ligand unbinding rates decrease under tension and promotes stress transmission from the adhesions [56] . Therefore , we next used the model to test how variations in catch bond kinetics , combined with differences in the relative amount of β-1 and β-3 integrins , modulate ligand binding and stress transmission . Since nascent adhesions transmit tension between the cytoskeleton and the ECM , we next asked how mixing integrins with different load-dependent bond kinetics impacts ligand binding and transmitted tension . The β-1 and β-3 integrins both behave as catch bonds that differ for unloaded and maximum lifetimes ( Fig 2C ) . In the model , an increase in the percentage of β-1 integrins while keeping the rest as β-3 integrins , increases ligand binding from about 5% to 35% when using actin flow speeds below 15 nm/s ( Fig 5A ) . The percentage of ligand-bound integrins is in direct proportion to the amount of β-1 integrins ( Fig 5A ) . At actin flow speeds below 15 nm/s , traction stress and flow rate are positively correlated , while at higher flows they are inversely correlated ( Fig 5B ) , in agreement with previous findings [60 , 61] . Interestingly , variations in the relative fractions of the two integrin types do not affect the average tension on each integrin-ligand bond ( Fig 5B ) . Below 10 nm/s actin flow , the minimum separation between ligand-bound integrins decreases from about 120 to 10 nm by increasing the fraction of β-1 integrins ( Fig 4C ) . Stable adhesions , with minimum separation between ligand-bound integrins of 70nm , form with at least 20% β-1 integrins ( Fig 5C ) . Together , our results show that the relative fractions of β-1 and β-3 integrins cooperate with actin flow to determine ligand binding and adhesion stability . Interactions of adhesions with a cytoskeleton network play important role in several cell activities , including spreading and migration . The actin cytoskeleton exists in different architectures , depending on the cell location and function . Therefore , we next considered how the architecture of the actin cytoskeleton can impact the formation of adhesions . We incorporated in the model explicit actin filaments , using random , crisscrossed , and bundled architectures ( Fig 6A–6C ) . The model assumes that ligand-bound integrins can interact with actin filaments , and that binding to actin increases integrin activation rate , as detected experimentally [63 , 64] . Increasing the fraction of β-1 integrins , ligand binding increases independent of network architecture ( Fig 6D ) . By contrast , integrin clustering remains at about 20–30% when a percentage of β-3 integrins is used . When only β-1 integrins are used , integrin clustering decreases of about 3-fold , independent from network architecture ( Fig 6E ) . The number of ligand-bound integrins with a separation less than 70 nm is enhanced using a bundled network architecture ( Fig 6F ) . This suggests that the probability of adhesion stability and ultimately maturation is higher with bundled architectures relative to both crisscrossed and random distributions of actin filaments ( Fig 6F ) . Collectively , our results indicate that the architecture of the actin cytoskeleton does not modulate the amount of ligand binding and integrin clustering . However , actin network architecture determines the physical distribution of ligand-bound integrins in adhesions , with bundled actin filaments increasing the probability of adhesion stability , consistent with previous experimental observations [50] .
Since our experiments show that different integrin types exist in nascent and mature adhesions ( Fig 1B ) , a computational model is here developed in order to understand if differences in nanoscale physical properties of integrins reflect on adhesions . This is largely untested by experimental approaches because it is very challenging to simultaneously distinguish between integrin types and isolate their nanoscale physical properties . The model is used to study how ligand binding affinity , rate of integrin activation , strength of pairwise interactions , bond kinetics , as well as the architecture of a network of actin filaments modulate integrin organization in adhesions and stress transmission . Our results collectively show that ligand binding and integrin clustering are competing mechanisms and that bundled actin networks favor adhesions stability , and ultimately maturation . The model is developed through three consecutive stages of increasing complexity: ( i ) simulations of single-point integrins diffusing on a quasi-2D surface and switching between active and inactive states , binding ligands , and interacting laterally; ( ii ) incorporation of an implicit actin flow and integrin/ligand catch bonds kinetics; ( iii ) binding of integrins to semi-flexible actin filaments in either random , bundled , or crisscrossed architectures . At all stages , we distinguish between β-1 and β-3 integrins , by using either exact , experimentally detected physical parameters , realistic fold differences between the two , or estimates from previous free energy calculations . For high EIL , many active integrins bind ligands and the fraction of integrins that can diffuse , and cluster , is reduced ( Fig 3D–3E ) . Accordingly , this happens when the fraction of β-1 integrins is higher than that of β-3 integrins ( Fig 5A ) , since β-1 integrins have higher ligand-binding affinity than β-3 integrins . By contrast , with many free diffusing integrins that have low EIL , and are less likely to bind ligands , the fraction of integrins that can encounter each other , and cluster is enhanced and reduces ligand binding ( Fig 3D–3E ) . This happens when the fraction of β-3 integrins is higher than that of β-1 integrins ( Fig 5A ) and also results from the higher diffusion coefficient of β-3 integrins with respect to β-1 integrins [43 , 44] . The result that ligand binding and integrin clustering are competing mechanisms is consistent with a kinetic Monte Carlo model showing that thermodynamics of ligand binding and dynamics of integrin clustering interplay [46] . Our model reproduces this competing process over the same range of ligand binding affinities and strength of pairwise interactions . The molecular mechanisms resulting in integrin lateral clustering remain controversial . However , several lines of evidence have suggested that β -3 integrins assemble clusters more easily than β-1 integrins . For example , activation of β -3 integrins induces formation of clusters with recruitment of talin [7] , while β-1 integrins require recruitment of many more signaling components in order to form clusters , such as FAK [57] . β-3 integrins cluster in response to talin binding without a concomitant increase in affinity [24] , while β-1 integrins cluster only when extended [58] . Moreover , previous studies in U2OS cells showed that β-3 integrins cluster on both β-3 and β-1 integrin ligands , while β-1 integrin clusters are present in adhesions only on β-1 ligands [59] . Our result that β-1 integrins are correlated with ligand-binding while β-3 integrins are mostly responsible for clustering is consistent with the observation that clusters of β-1 integrins are present only on β-1 ligands , possibly because , in this case , ligand binding and not pairwise interactions facilitate adhesions assembly . Further evidence that β-3 integrins assemble more easily than β-1 integrins is provided by the reported spatial and functional segregation of the two integrin types . β-1 integrins translocate from the cell periphery to the cell center to withstand higher tensions , whereas β -3 integrins remain at the cell edges to do mechanosensitive activities via dynamic breakage and formation of multiple bonds with the substrate and with one another [38] . Our model also shows that the number of integrins per cluster , computed as the fraction of clustered versus ligand bound integrins , is in the range of 2–15 particles , depending on EII and EIL ( Fig 3F ) . This value is comparable to the experimentally estimated number of integrins in nascent adhesions , between 5–7 [10] . By varying ligand density in the model , the ratio between clustered and ligand bound integrin particles does not vary significantly ( S1 Fig ) , suggesting that the average number of integrins per cluster in nascent adhesions is not modulated by ligand concentration , consistent with previous experimental observations [10] . In the presence of actin flow , the fraction of β-1 integrins is positively correlated with ligand binding ( Fig 5A ) . Above a threshold actin flow , however , ligand binding is almost suppressed , independent from relative amounts of β-1 and β -3 integrins ( Fig 5A ) , because of faster ligand unbinding from both integrins . This reduction in bound ligands corresponds to a drop in the average tension per integrin upon increasing actin flow ( Fig 5B ) . The biphasic response of tension to actin flow was previously observed experimentally [60] and is consistent with models of adhesion clutch assembly and rigidity sensing [61] . By increasing the fraction of β-1 integrins , a reduction of lateral integrin spacing is observed with our model ( Fig 5C ) . Previous studies on the lateral separation of integrins in adhesions reported that a minimum spacing of 70 nm is required to form stable adhesions [62] . This value corresponds in the model to a minimum of 20% β-1 integrins ( Fig 5B ) . This value represents a prediction from our CG model that can be experimentally tested in the future . When only β-3 integrins are used in the model , their lateral separation , upon binding ligands , is about 120 nm ( Fig 5C ) , supporting the notion that β-1 integrins are needed to form stable adhesions . This is consistent with the fast binding/unbinding dynamics of β-3 integrins previously observed in experiments [28] . By incorporating a positive feedback between actin filament engagement and integrin activation , as observed in [63 , 64] , the competition between clustering and ligand binding is maintained in all actin architectures ( Fig 6D–6E ) . This positive feedback represents the functional link between cytoskeleton and adhesions , where an increase in the probability of ligand binding results from binding actin , via an inside-out pathway [12 , 65] . Our model shows that the number of ligand-bound integrins with an average separation below 70 nm is enhanced with a bundled architecture ( Fig 6F ) , suggesting that this configuration favors adhesion stability , and ultimately maturation [50] . When integrins bind a bundled network , they are likely to re-bind in close proximity because bundled filament architectures present filaments that are spatially closer than filaments of crisscrossed or random networks , forming a spatial trap for the receptors . Of interest for future studies is mimicking conditions of actin filament turnover , in order to understand how a dynamic cytoskeleton can interplay with integrin mixing in forming nascent adhesions . This will help understanding outside-in pathways , where , for example , adhesions formation modulates actin filaments polymerization . A further extension of the model will incorporate dynamic ligands , interconnected by a fibrous extracellular matrix that deform under tension . We will study how adhesions formation can change ligand localization and how this , in turns , affects adhesions morphology . Previous computational studies of integrin dynamics range from all-atom simulations and enhanced sampling methods for understanding integrin activation at the level of individual molecules [66–68] , to lower resolution coarse-grained [61 , 69–74] , lattice-based [75 , 76] , diffusion-reaction algorithms [77] and theoretical models [78] for multiple integrins . With respect to the previous lower resolution models of multiple integrins , our new model allows us to directly incorporate properties of different integrin types , as detected experimentally ( Fig 1B ) . The particle-based implementation scheme of our model is similar to that of other software for modeling the cytoskeleton , such as Cytosim [79] and Medyan [80] . However , important differences exist . In contrast to Cytosim , an explicit implementation scheme is used here because our time step , combined with the limited number of simulated particles ( a few hundreds ) , allows us to achieve time scales of a few minutes , that are relevant for adhesion assembly , without excessive computational cost . In addition , in contrast to Medyan , our model does not have a scheme for solving stochastic reaction-diffusion equations , but instead focuses on the mechanics of particle interactions and displacements under deterministic and Brownian forces . To conclude , with our highly coarse-grained model based on Brownian Dynamics , we extend the scope of previous theoretical and computational studies of integrin-based adhesions formation , by testing how differences in nanoscale properties of β-1 and β-3 integrins impact ligand binding , clustering and transmission of traction stress . By coupling physical parameters ( such as diffusivity ) together with chemical ( i . e . , affinity and receptor pairwise interactions ) and mechanical ( bond kinetics ) parameters , and by using an explicit actin cytoskeleton , our model shows that nascent adhesions assembly can be finely tuned by differences in nanoscale physical properties of integrins . The CG model ultimately demonstrates that nanoscale differences in integrin dynamics are sufficient to determine ligand binding and integrin clustering . By incorporating dynamics of individual integrins in an explicit way , our model provides results that are consistent with a number of previous independent experimental observations , revealing important insight into the molecular origins of adhesion organization and mechanics . Taken together , our modeling results support the general view that a cell can control integrin expression to determine morphological and dynamic properties of adhesions .
The computational domain includes two systems: a square bottom surface , of 1 μm per side , and a rectangular 3D domain above the surface , with dimensions 1 x 1 x 0 . 04 μm ( Fig 2A ) . The bottom surface mimics the substrate; the lower side of the rectangular domain mimics the ventral cell membrane above the substrate , while its inside space represents a 40 nm thick cytoplasmic region where actin filaments diffuse beyond the ventral membrane ( Fig 2A ) . The cell membrane is separated from the substrate by 20 nm , a dimension characteristic of active integrin headpiece extension ( Fig 1A ) [82] . Within the cell membrane , integrins diffuse in quasi-2D and are restrained in the vertical direction by a weak harmonic potential with spring constant 100 pN/μm , mimicking membrane vertical friction . In the cytoplasmic region , a repulsive boundary is used on the top surface , to avoid filaments crossing the boundary . Periodic boundary conditions are applied on all lateral sides of the domain , in order to avoid finite size effects . The model considers a given number of ligands on the substrate , randomly distributed and fixed in space . We use a ligand density of 1000#/μm2 , of the same order of that used in a previous model of adhesions assembly [72] . Integrin density on the cell membrane is ~100#/μm2 [5] . Integrins are single-point particles , that are initially randomly distributed and diffuse over the course of the simulations . Integrin diffusion coefficient is D = 0 . 1 μm2/s for β-1 integrins and D = 0 . 3 μm2/s for β-3 integrins [43] . Introducing volume exclusion effects between integrins , in the form of a weak repulsion between nearby particles ( 1 pN force ) , does not change the fraction of ligand-bound integrins , their average separation , the mean tension per integrin and its distribution ( see S2A–S2D Fig ) . Increasing the magnitude of this repulsion ( 10 pN ) , however , affect the average separation of integrins ( S2E–S2F Fig ) . Semiflexible polymers represent actin filaments as spherical particles connected by harmonic interactions . Filaments have fixed length of 0 . 5 μm , corresponding to 6 beads separated by 0 . 1 μm equilibrium distance . The model of actin filaments is explained in detail in [83] . Actin filament beads are subjected to both stochastic and deterministic forces . Stochastic forces on the i-th bead are random in direction and magnitude in order to mimic thermal fluctuations and satisfy the fluctuation-dissipation theorem: 〈Fistochastic∙FistochasticT〉α , β=2 ( kBTμ/dt ) I^α , β ( 1 ) with I^α , β being the second-order unit tensor [84] and μ being a friction coefficient equal in three directions . Deterministic contributions come from bending and extensional forces on the filament beads . The bending force is computed as: Fibend=−dEspringdri=kBTlpl0∑j=1N−1d ( tj∙tj−1 ) dri ( 2 ) where lp = 10 μm is actin filament the persistence length , N is the number of beads in a filament ( N = 6 ) and ti= ( rj+1−rj ) |rj+1−rj| . The extensional force on filaments beads is computed as: Fiextension=−dEextensiondri=k2∑j=1N−1d ( |rj+1−rj|−l0 ) 2dr ( 3 ) where l0 is the equilibrium length of 0 . 1 μm , k is the spring constant of 100 pN/μm . Each spherical particle of a filament represents a binding site for integrin and each binding site can interact with multiple integrins . In order to mimic hierarchical formation of nascent adhesions [10] , the algorithm incorporates sequential interactions between integrins , ligands and actin filaments . First , we simulate a system composed of only integrins and ligands , in order to explore the ways in which integrins cluster and bind ligands in an actin-independent way . Then , we add actin filaments and study the effect of actin network architecture on adhesions formation . Integrins switch between inactive and active states , with rates of activation and deactivation ka = 0 . 5 s-1 and kd = 0 . 0001 s-1 , of the same orders of those previously estimated [44 , 45 , 72] . Activation probability corresponds to Pa = kadt , with time-step dt = 0 . 0001 s , as the time of the smallest simulated phenomena . This large time-step is allowed because the extensional stiffness of actin filaments , 100 pN/μm , is smaller than the real actin filaments stiffness , of about 400 pN/nm [85] . Upon activation , integrins can interact with free ligands , using a harmonic potential ( with equilibrium separation 20 nm and spring constant 1 pN/μm ) , and cluster with other active integrins , depending on relative distances . Ligand binding occurs within a threshold distance of 20 nm , which reflects the extension of the open conformation of αIIbβ3 integrin away from the membrane [82] . Each integrin can bind only one ligand , and each ligand can bind only one integrin , mimicking binding sites specificity . Clustering occurs below a threshold of 30 nm , a value of the same order of the integrin-to-integrin lateral separation observed experimentally [82] and one order of magnitude lower than the minimum separation between individual adhesions [86] . The probability of integrin deactivation is Pd = kddt . Once inactive , integrin loses its connections with ligands and other integrins . Integrins unbind ligands with dissociation probabilities depending on their affinities: P=λeEILdt , using a prefactor λ = 1 s-1 , for simplicity . They break later connections with probabilities inversely proportional to strength of pairwise interaction: P=λeEIIdt . For β-1 integrins , we use high affinity , ~9 kBT; for β-3 integrins we use lower affinity , 3–5 kBT . Ligand-bound integrins can establish harmonic interactions with semiflexible actin filaments below 5 nm , approximating the size of the intracellular integrin tails [82] . Since the exact molecular composition of the layer between integrin and actin can contain up to ~150 different proteins [87] , a detailed modeling representation is not possible . Therefore , interactions between integrin and actin are approximated by harmonic potentials with equilibrium distance of 3 nm and spring constant of 1 pN/μm . These interactions simplify the ~40 nm layer of adhesion molecules , including vinculin , talin and α-actinin , and is consistent with the level of details of the simulations , where harmonic interactions are used to connect particles within 20–100 nm . Displacements of integrin and actin filament particles are governed by the Langevin equation of motion in the limit of high friction , thus neglecting inertia: Fi−ξidridt+FiT=0 ( 4 ) where ri is a position vector of the ith element , ζi is a drag coefficient equal in three directions , t is time , Fi is a deterministic force , and FiT is a stochastic force satisfying the fluctuation-dissipation theorem [88] . Fi is the sum of forces resulting from interactions of integrins with a ligand and/or other particles in the system , and actin flow in a direction parallel to the substrate . Positions of the various elements are updated at every time step using explicit Euler integration scheme: ri ( t+dt ) =ri ( t ) +1ξi ( Fi+FiT ) dt ( 5 ) Since contraction forces are not needed for the assembly of nascent adhesions [5 , 10 , 89] , our computational model only incorporates forces mimicking actin retrograde flow . In order to simulate actin flow and characterize distribution of traction stress at various flow rates , a constant force is applied on ligand bound integrins , along y ( Fig 2B ) . Lifetime of the bond between integrin and ligand follows the catch-bond formalism ( Fig 2C ) , using: for β-1 integrins an unloaded affinity of 2 s and a maximum lifetime of 15 s; for β-3 integrins an unloaded affinity of 0 . 5 s and a maximum lifetime of 3 s . The parameters for the catch bond kinetics are from previous experimental characterizations [38 , 47 , 56] . Curves of bond lifetime versus tension are shown in Fig 2C . For β-1 integrins , we implemented an unbinding rate as a function of the force acting on the bond , F: ku ( F ) =0 . 4e−0 . 04F+4E−7e0 . 2F ( 6 ) For β-3 integrins , we used unbinding rate: ku ( F ) =2e−0 . 04F+4E−6e0 . 2F ( 7 ) The functional forms of the catch bonds were taken from a model that assumes a strengthening and a weakening pathway for the bond lifetimes , using a double exponential with exponents of opposite signs [90 , 91] . This model was also used for previous simulations of integrin-based adhesions [5] . To mimic promotion of integrin clustering upon ligand binding and actin filament engagement [81] , we introduce a positive feedback between binding of integrin to a filament and integrin activation rate . In the model , integrins can bind a filament only if already bound to a ligand . Upon binding to actin , integrin activation rate is increased by 2 to 4% relative to its initial value . This assumption is motivated by recent evidence from TIRF experiments on T-cells , where it was demonstrated that actin binding and correct ligand positioning are needed for integrin activation [81] . The positive feedback between actin binding and integrin activation rate also represents conditions of inside-out signaling , with increased affinity for ligand binding induced by the cytoplasm [65] . We use the model with the positive feedback ( schematics in Fig 2D ) to test the effect of different actin architectures on ligand binding and clustering ( Fig 6 ) . For bundled and crisscrossed actin filament architectures , we impose spatial restraints on filaments pairs . Bundled architectures have harmonic connections between beads of filament pairs that keep the filaments in parallel; crisscrossed architectures impose 90 deg angle between the axis of filaments pairs , that keep them almost perpendicular . Human Foreskin Fibroblasts ( HFF ) were purchased from ATCC and cultured in DMEM media ( Mediatech ) supplemented with 10% Fetal Bovine Serum ( Corning ) , 2 mM L-glutamine ( Invitrogen ) and penicillin-streptomycin ( Invitrogen ) . HFFs were plated on glass coverslips incubated with 10 μg/mL fibronectin ( EMD Millipore ) for 1 hr at room temperature . Cells were fixed 1 hr after plating by rinsing them in cytoskeleton buffer ( 10 mM MES , 3 mM MgCl2 , 1 . 38 M KCl and 20 mM EGTA ) and then fixed , blocked and permeabilized in 4% PFA ( Electron Microscopy Sciences ) , 1 . 5% BSA ( Fisher Scientific ) , and 0 . 5% Triton X-100 ( Fisher Scientific ) in cytoskeleton buffer at 37° for 10 minutes . Coverslips were subsequently rinsed three times in PBS and incubated with either a β1 antibody ( 1:100; Abcam product #:ab30394 ) or β3 antibody ( 1:100; Abcam product #:ab7166 ) followed by AlexaFluor 488 phalloidin ( 1:1000; Invitrogen ) and a AlexaFluor647 donkey anti-mouse secondary antibody ( 1:200; Invitrogen ) . Cells were imaged using a 1 . 2 NA 60X Plan Apo water immersion lens on an inverted Nikon Ti-Eclipse microscope using an Andor Dragonfly spinning disk confocal system and a Zyla 4 . 2 sCMOS camera . The microscope was controlled using Andor’s Fusion software .
|
Integrin-mediated cell adhesions to the extracellular environment contribute to various cell activities and provide cells with vital environmental cues . Cell adhesions are complex structures that emerge from a number of molecular and macromolecular interactions between integrins and cytoplasmic proteins , between integrins and extracellular ligands , and between integrins themselves . How the combination of these interactions regulate adhesions formation remains poorly understood because of limitations in experimental approaches and numerical methods . Here , we develop a multiscale model of adhesion assembly that treats individual integrins and elements from both the cytoplasm and the extracellular environment as single coarse-grained ( CG ) point particles , thus simplifying the description of the main macromolecular components of adhesions . The CG model implements sequential interactions and dependencies between the components and ultimately allows one to characterize various regimes of adhesions formation based on experimentally detected parameters . The results reconcile a number of independent experimental observations and provide important insights into the molecular basis of adhesion assembly from various integrin types .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"cell",
"motility",
"actin",
"filaments",
"spring",
"seasons",
"integrins",
"cellular",
"structures",
"and",
"organelles",
"cytoskeleton",
"dynamic",
"actin",
"filaments",
"contractile",
"proteins",
"actins",
"physical",
"properties",
"cell",
"adhesion",
"proteins",
"extracellular",
"matrix",
"cell",
"membranes",
"physics",
"biochemistry",
"cytoskeletal",
"proteins",
"cell",
"biology",
"earth",
"sciences",
"biology",
"and",
"life",
"sciences",
"physical",
"sciences"
] |
2019
|
Multiscale model of integrin adhesion assembly
|
An estimated 80% of genomic DNA in eukaryotes is packaged as nucleosomes , which , together with the remaining interstitial linker regions , generate higher order chromatin structures [1] . Nucleosome sequences isolated from diverse organisms exhibit ∼10 bp periodic variations in AA , TT and GC dinucleotide frequencies . These sequence elements generate intrinsically curved DNA and help establish the histone-DNA interface . We investigated an important unanswered question concerning the interplay between chromatin organization and genome evolution: do the DNA sequence preferences inherent to the highly conserved histone core exert detectable natural selection on genomic divergence and polymorphism ? To address this hypothesis , we isolated nucleosomal DNA sequences from Drosophila melanogaster embryos and examined the underlying genomic variation within and between species . We found that divergence along the D . melanogaster lineage is periodic across nucleosome regions with base changes following preferred nucleotides , providing new evidence for systematic evolutionary forces in the generation and maintenance of nucleosome-associated dinucleotide periodicities . Further , Single Nucleotide Polymorphism ( SNP ) frequency spectra show striking periodicities across nucleosomal regions , paralleling divergence patterns . Preferred alleles occur at higher frequencies in natural populations , consistent with a central role for natural selection . These patterns are stronger for nucleosomes in introns than in intergenic regions , suggesting selection is stronger in transcribed regions where nucleosomes undergo more displacement , remodeling and functional modification . In addition , we observe a large-scale ( ∼180 bp ) periodic enrichment of AA/TT dinucleotides associated with nucleosome occupancy , while GC dinucleotide frequency peaks in linker regions . Divergence and polymorphism data also support a role for natural selection in the generation and maintenance of these super-nucleosomal patterns . Our results demonstrate that nucleosome-associated sequence periodicities are under selective pressure , implying that structural interactions between nucleosomes and DNA sequence shape sequence evolution , particularly in introns .
Sequence-dependent differences in the physical properties of DNA influence its associations with the histone core , as well as the kinetics of nucleosome assembly and stability [2]–[9] . One of the most generalizable sequence affinities of the histone octamer is the periodic variation of dinucleotide frequencies across nucleosomal DNA . Alignments of nucleosomal sequences from diverse eukaryotes display a prominent ∼10 bp periodic enrichment of AT-rich dinucleotides , along with an anti-correlated periodicity of GC-rich dinucleotides [8] , [10]–[14] . The ∼10 bp spacing of AA/TT dinucleotides generates intrinsically curved DNA molecules with increased nucleosome binding affinity [5] , [8] , [14]–[17] . Peaks of AA/TT frequency are found specifically over positions where the minor groove bends interiorly , whereas GC dinucleotides peak where the major groove is facing the histone core . Structural data suggest that DNA shape , in particular the narrowing of the minor groove and the associated lowering of its electrostatic potential at AT-rich sequences facilitate contacts with key histone arginines [9] , [18] , [19] . GC dinucleotides contract the major groove , which also facilitates the tight winding of DNA around the core [9] , [20] . Although these broadly conserved dinucleotide patterns have been cited as evidence for a genomic “code” for nucleosome positioning [10] , the role of sequence in nucleosome function remains contested and unresolved [6] , [8] , [21] , [22] . Correlation between in vitro and in vivo nucleosome maps in yeast may reflect the influence of the inherent sequence preferences of the histone core on nucleosome positioning [7] , [23] . However , strong experimental evidence suggests that trans-acting factors ( e . g . RNA polymerase II , transcription factors and ATP-dependent remodelers ) are central to establishing nucleosome positions along genomic DNA ( translational positions ) , particularly in genic regions , with sequence providing a weaker contribution [7] , [8] , [21] , [24] , [25] . In cases where DNA sequence does impact translational nucleosome positions , its influence is largely attributed to GC content and anti-nucleosomal sequences , such as poly-dA/dT tracts , rather than dinucleotide patterns [5] , [8] , [25]–[27] . Dinucleotides are instead thought to play a distinct but integrally connected role in directing and preserving the ‘rotational positioning’ of nucleosomal DNA [8] , [9] , [20] , [28] , which refers to the orientation of DNA relative to the core . Due to the structural constraints inherent to nucleosome formation , a given translational position in the genome will assemble with a particular rotational alignment . This determines which bases face the nucleosome interior and exterior , and also the positioning of the major and minor grooves relative to the core . Nucleosomes tend to occupy translational genomic positions which are offset by ∼10 bp increments [13] , [28] , [29] . Thus , due to the helical structure of DNA , with ∼10 . 4 bp per turn , the rotational orientation of DNA relative to the core is thought to be unchanged as nucleosomes assume new favored translational positions ( Figure 1A ) . This 10 bp incremental movement leaves the exposure of sites at the surface unchanged [20] , [28] , and is in agreement with the reported step size of many chromatin remodelers [30] . By influencing the rotational positioning of DNA relative to the histone core , nucleotide changes at particular nucleosome positions ( or in flanking regions ) could have diverse functional impact , for example on nucleosome assembly , stability , remodeling efficiency , RNA and DNA polymerase processivities and transcription factor binding site access . However , despite considerable evidence that dinucleotide patterns impact nucleosome positioning and dynamics in vitro , in vivo evidence of function has remained elusive . One approach to discovering function is to look for evidence of natural selection in sequence polymorphism ( variation within species ) and divergence ( variation between species ) . Individual mutations influencing histone-DNA interactions may have only slight , undetectable phenotypic effects in the laboratory; in contrast , the associated fitness consequences in large natural populations can strongly shape rates of divergence and levels of polymorphism over many generations [31] . Of course , strongly selected variants will go to fixation quickly and be maintained in very high frequency against the weaker force of mutational reversion . However , observations of extensive DNA sequence polymorphism and divergence throughout the genome , including nucleosomal sequences , indicate that such systematic selection is not dominating stochastic effects ( mutation , genetic drift , and variation in selection coefficients ) in the evolutionary dynamic . Analysis of codon bias suggests that at equilibrium between selection , mutation and genetic drift , the ratio of the frequencies of two alternative synonymous codons throughout a single genome can be used to estimate the direction and magnitude of selection [32] . The action of natural selection can be inferred when synonymous codon pairs exhibit a strong “bias” towards one state relative to the other . This analysis extends to the distribution of polymorphic allele frequencies in genomes sampled from natural populations [33]–[35] . A similar approach can be applied to alternative nucleotides at particular positions within nucleosomal sequences . As the magnitude of selection increases , the expected frequency of preferred alleles increases . Consequently , the distribution of SNP frequencies ( or “site frequency spectrum” ) at a given nucleosome position ( analogous to a synonymous SNP ) is expected to shift towards relatively higher frequencies of preferred alleles . If the observed dinucleotide patterns reflect selectively favored states , ancestrally unpreferred base pairs across nucleosomes should diverge towards the “preferred” state along species lineages . Further , if the “preferred” divergence patterns reflect the average impact of natural selection , then frequencies of polymorphisms in natural populations should be more skewed at sites experiencing stronger selection . “Unpreferred” variants , specifically substitutions or polymorphisms away from favored nucleotides , such as substitution of an ancestral A with a G at nucleosome positions which are systematically enriched for AA dinucleotides , should diverge more slowly and be rarer when polymorphic in the population . In contrast , “preferred” variants , such as substitution of an ancestral A with a G at positions of enriched for the GC dinucleotide , should diverge more rapidly and be more common when polymorphic . At the lower resolution of an entire nucleosome and its nearby flanking regions , both divergence and polymorphism are observed to vary [36]–[39] , but evidence of a role for natural selection in the underlying evolutionary dynamics remains sparse [40]–[43] . Studies of human SNPs [37] , [38] and divergence in humans , yeast and medaka [36] , [38] , [39] , [43] show that both expected heterozygosity and divergence between species are elevated near the central dyad and depressed in the adjacent linker regions , though these patterns appear to differ by substitutional pathway [38] . One possibility is that patterns of variation relative to nucleosomes derive from nucleosome-specific mutational biases . This could result from suppression of mutation by a protective aspect of nucleosome occupancy [44] , or it could arise from an interaction between the histone core and DNA damage recognition or repair mechanisms [45]–[48] . Of course , natural selection mediated via DNA:nucleosome interactions may also strongly reshape the patterns of SNP variation and divergence between taxa [40]–[43] . Analysis of the site frequency spectrum promises to distinguish between these two alternatives . The whole-nucleosome-resolution analyses considered above cannot leverage the specific structural predictions of dinucleotide interactions with the core and their strong mechanistic implications . Examination of polymorphism and divergence at each base pair position across the nucleosomal DNA opens a rich and precise view , as well as powerful tests of alternative mechanisms such as biased mutation and natural selection . We report the discovery of fine-scale periodicities in inter- or intra-species sequence variation relative to nucleosomes and discuss their implications for the role of natural selection mediated through nucleosome function . Our analysis of DNA sequence polymorphism and divergence across isolated nucleosomal fragments from D . melanogaster embryos reveals that nucleosomal sequences are diverging towards “preferred” nucleotides . Regions where the minor groove is interior are becoming more AT-rich , and regions where the major groove is interior are becoming more GC-rich along the melanogaster lineage . Using a new index for quantitating the frequency spectrum ( Δπ ) , we identify clear signals associated with natural selection , which parallel the observed periodicities in divergence . This selection is strongest in intronic regions , where nucleosome assembly and positioning are expected to have greater functional impacts . These findings support the hypothesis that the widely observed sequence affinities of the core octamer have functional consequences that are subject to natural selection . Given the dominant role of nucleosomes in the packaging of the genome and their conserved sequence preferences , their interactions may broadly shape the sequence of melanogaster and other genomes .
To investigate the impact of nucleosomes on DNA sequence variation , we isolated nuclei from D . melanogaster embryos , performed Micrococcal nuclease ( MNase ) digestion , and used paired-end sequencing to position fragments on the genome ( Figure S1 ) . Previous studies in Drosophila identified a range of periodic dinucleotides in association with nucleosomes [10] , [11] . Our collections of 276 , 614 intergenic and 270 , 998 intronic autosomal 147 bp nucleosomal fragments ( hereafter n147 , Tables S1 and S2 ) cover 68 . 5% of the unique intronic and intergenic euchromatic autosomal genome and display a ∼10 bp periodicity for many dinucleotide frequencies ( Figures 1 and S2 ) . In these and subsequent analyses , the 5′-3′ sequence from bases −73 to −1 were joined to the reverse complement of bases 1 to 73 , to reflect the dyad symmetry of the nucleosome ( see Materials and Methods ) . AA , TT and GC showed the strongest periodicity of WW and SS ( where W = A|T , S = G|C ) dinucleotide pairs , respectively ( Figure 1B ) . These same dinucleotides show a distinct overrepresentation in the non-coding regions of the genome as a whole ( Figure 1C ) . As noted in previous studies , AA and TT are similarly periodic and occur where the minor groove is interior ( at superhelix locations , SHL , ± ( i+0 . 5 ) ; where i is 0 , 1 , …6 ) . However , noticeable differences between the distributions are apparent . For example , the frequency of TT displays a distinctly smaller peak at ∼SHL 4 . 5 , and AA frequency displays a stronger drop at ∼SHL 2 ( Figure 1B ) . GC frequency across n147 regions is anti-correlated with AA/TT and is characterized by a prominent upward concavity ( Figure 1B ) . These dinucleotide periodicities extend well beyond n147 edges into linker regions , consistent with the proposed translational step size of 10 bp . Upon examination of the dinucleotide frequencies flanking aligned n147 regions , we discovered an additional large-scale pattern in AA/TT and GC dinucleotide frequencies ( Figure 1D ) . This ∼180 bp periodic variation in frequency tracks with overall nucleosome “occupancy” in the regions flanking the n147 . Average AA/TT frequencies ( Figure 1D ) and overall A/T frequencies ( Figure S3 ) are higher in regions of greater nucleosome “occupancy” and lower in putative “linker” regions . Thus , the AA/TT sequence features that facilitate nucleosome formation are enriched over regions with higher nucleosome “occupancy . ” Conversely , GC frequency ( and overall G/C frequencies , Figure S3 ) peaks at the periphery of more nucleosome-dense regions and in “linker” regions . These surprising “super-nucleosomal” periodicities extend the observed n147 patterns to flanking multi-nucleosomal arrays , and suggest a contribution of sequence to translational positioning . Consistent with chemical mapping of nucleosomes , this result suggests that the observed experimental correlation between MNase nucleosome “occupancy” and GC content [1] , [8] , [22] , [26] , [49] , [50] reflects differential recovery , rather than positional preference [23] , [51] . If variations in dinucleotide frequencies relative to nucleosomes result from accumulated sequence divergence , we expect substitution patterns to parallel the observed base preferences . However , the timescale ( s ) at which these patterns evolve is unknown . Lineage-specific or “polarized” divergence is the proportion of nucleotide sites that are different in melanogaster while identical in its sister taxa simulans ( most recent common ancestor 2 . 5 MYA ) and the proximate outgroup ( yakuba or erecta; 6–7 MYA , see Materials and Methods ) . Overall genomic divergence on the melanogaster lineage shows a marked excess of G→A ( inferred ancestral G , derived A in melanogaster ) and C→T ( ancestral C , derived T ) substitutions compared to A→G and T→C ( Figure 2A ) . This is in agreement with earlier estimates of divergence on the melanogaster lineage [52] , [53] and with the observed two-fold greater mutation rate [54] . We next considered the average divergence at each site across n147 regions , normalized for underlying base frequencies . This analysis revealed a striking ∼10 bp periodicity in transitions ( GC→AT and AT→GC ) for two estimates of divergence; per-n147 in Figure 2B is weighted by the redundancy in the n147 set , while per-site in Figure S4 weights each site equally . Rates for GC→AT and AT→GC are anti-correlated and track with underlying dinucleotide frequencies . Thus , ancestral GC bases are more likely to become AT in nucleosomal regions where AA/TT dinucleotides are in higher frequency , and AT bases are more likely to become G or C at sites where GC is enriched . GC→AT divergence also shows a marked curvature , with a peak at the dyad axis . Given the substantial variation in individual substitution rates , we next examined specific pathways to determine their relative contributions . Of all pathways , G→A , C→T , A→G and T→C exhibited the most obvious periodicities in divergence ( Figure 2C; see Discussion ) . In some cases , divergence patterns reflect the subtleties observed for dinucleotide frequency patterns . For example , C→T rates are less peaked at SHL 4 . 5 , the location of the lowest peak in TT frequency ( Figure 2C ) . C→T also shows a greater difference in rates between the n147 periphery and linker regions ( compared to G→A ) . Interestingly , for each ancestral base , the periodicities of substitutions that do not change GC content , appear weaker , perhaps due to both scaling and weaker signal to noise ( Figure 2C ) . A subset of non-overlapping n147 regions showed similar patterns ( Figure S5A ) . When mapped onto the DNA from the nucleosome structure [55] , peaks of intergenic G→A divergence clearly occur within regions where the minor groove is interior and in contact with key arginines of the histone core ( Figure 2D ) . Note also the higher G→A divergence toward the central axis , as reflected by the downward concavity in Figure 2C . This is consistent with analyses of the impact of sequence variation on nucleosome structure , which identified this central region of H3/H4 interactions as most constrained [56] . Conversely , A→G substitution rates are highest in regions where the major groove is interior ( Figure S5B ) . This pattern is consistent with established SS dinucleotide patterns [8] , [10]–[14] and the observation that GC rich sequences are disfavored for minor groove compression and favor narrowing of the major groove [9] , [18] . Divergence patterns should also reflect the observed nucleosome-scale periodicities in base and dinucleotide frequencies ( Figures 1D and S3 ) . To increase signal , we combined complementary substitutions , G→A and C→T ( G→A:C→T ) and A→G and T→C ( A→G:T→C ) . Aligned n147 regions show substantially lower divergence rates than their immediate flanking sequences ( Figure 2E ) . Rates drop to the local background within ∼500 bp , following the skew of AA/TT dinucleotides ( and overall AT content; Figures 1D and S3 ) . In spite of this local variation in rates , due at least in part to MNase preferences ( Figure S6 ) , we observe a large-scale ( 180 bp ) periodicity in G→A:C→T divergence surrounding intergenic n147 nucleosomal regions ( Figure 2E ) . Introns showed a similar but weaker pattern , potentially due to the influence of flanking coding regions ( Figure 2E ) . Any periodicity of the A→G:T→C divergence in flanking regions is less obvious ( Figure S5C ) , at least partially due to a 50% lower rate of divergence and thus inherently weaker signal . These large-scale patterns allow us to resolve general trends in divergence relative to nucleosome occupancy . We find that , on average , G→A:C→T changes along the melanogaster lineage are fixed at higher rates across nucleosomes relative to linkers , mirroring underlying AA/TT dinucleotide frequencies . This is in apparent contrast to the report that the cytosine deamination mutational pathway ( a major source of G→A:C→T transitions ) and associated divergence is suppressed by nucleosome occupancy [44] . To clarify this discrepancy , we examined the interactions between divergence and “occupancy” of the n147 fragments , as estimated by depth of coverage by 142–152-bp nucleosomal fragments , n142-152 . Indeed , we observe a negative correlation between this metric and all substitutional pathways ( Table S3 , Figure S7A ) . However , we note that n142-152 coverage is correlated with GC content of the n147 region ( Figure S8D ) , as previously reported in other studies [1] , [5] , [8] , [26] , [49] , and that correlations between nucleosome fragment GC and divergence are even more striking ( Table S3 , Figure S7B , S8C ) . This is also true for 500 bp intergenic windows , independent of nucleosome coverage ( Table S3 ) . When we parse n147 by n142-152 “occupancy , ” we observe differences in AA/TT frequency , G→A:C→T divergence , and nucleosome phasing in flanking regions ( Figure S8A ) . The periodicities of these features are most obvious surrounding highly occupied ( GC-rich ) n147 regions , but they do not appear to be unique to them . Thus , we conclude that nucleosome bound regions in D . melanogaster embryos are generally more AT-rich and have higher rates of G→A:C→T substitution than their adjacent “linker” regions , inconsistent with the fundamental claim in Chen , et al . [44] ( mentioned above ) . The divergence patterns we observe are consistent with known nucleosomal dinucleotide preferences [5] , [8] , [10]–[17] . This is analogous to observations for codons , where substitutions mirror genome-wide codon usage biases and are attributed to natural selection for preferred codons [34] , [57] , [58] . However , divergence patterns alone cannot exclude the hypothesis that substitutional patterns result from biased mutation relative to nucleosomes . Mutation rates may vary across nucleosome-bound regions and could lead to compositional variation and different rates of divergence . Nevertheless , once a new selectively neutral allele arises , its dynamics and thus its distribution of frequencies are independent of type ( or rate ) of mutation [59] , [60] . While natural selection influences the probability of fixation ( thus the rate of divergence ) , mild differences in fitness will also shift the site frequency spectra of polymorphic alleles [61]–[63] . Neutral and deleterious mutations tend to spend much of their typically short lives as rare alleles , while weakly favored alleles will be found at higher frequencies as many more drift towards fixation . Although the impacts of varying demographic histories [64] and of linked selection [48] , [49] can lead to distributions of selectively neutral polymorphisms that mimic particular forms of selection , they should do so randomly across the genome and not show a positional relationship within nucleosomal sequences . The hypothesis that nucleosome structure and function impose natural selection on genomic sequence variation predicts periodicities in the frequency spectra . Indeed , the average per-n147 frequencies of G-A and C-T SNPs in a sample of 36 D . melanogaster genomes from Raleigh ( North Carolina ) exhibit nucleosomal patterns paralleling those observed for dinucleotides and polarized divergence ( Figures 3A and S9 ) . Frequencies of A alleles at G-A SNPs show clear periodicity across intergenic and intronic n147 regions , extending into linker regions ( Figure 3A ) . A alleles are relatively more common in SHL ± ( i+0 . 5 ) regions , and G alleles are higher in regions where the major groove faces the histone core . Removal of singleton SNPs ( cases where either allele is observed only once ) , which can mitigate the impact of possible sequencing errors , raises average A frequencies but does not eliminate the periodicity ( Figure S9 ) . Partitioning such SNPs by ancestral state can remove the impact of average mutation rate differences and reveal differences in the patterns of selection . Nucleosomal patterns of the average per-n147 frequencies of derived SNPs , such as G→GA ( ancestral G and a derived , polymorphic A ) , exhibit clear periodicities that generally parallel divergence and nucleosomal dinucleotide frequencies ( Figures S10 , intergenic , and S11 , intronic ) . To systematically assess the periodicity in the frequency spectra we calculated a new index , Δπ , ( closely related to Tajima's D [59] ) across n147 regions for the Raleigh sample [65] . Where p is the frequency of a SNP in the sample and is the estimate of the heterozygosity , we define Δπ as the average ( per SNP ) deviation in from expectation under equilibrium between genetic drift and mutation to selectively equivalent alleles ( see Materials and Methods ) . The “folding” of the frequency spectrum such that p is equivalent to ( 1−p ) mitigates the impacts of errors in the inference of the ancestral state [66] and emphasizes variation in the midrange of p . Weak positive selection is predicted to skew toward higher values ( more positive Δπ ) , while weak negative selection leads to more negative Δπ . Thus , systematic differences in selective forces at different positions across n147 regions should yield a pattern in Δπ that parallels that observed for divergence . These patterns of Δπ are superimposed on the observed genome-wide average negative skew [65] ( Table S4 ) that can be attributed to strongly deleterious mutations [67] , [68] , varying demographic history [59] , [64] or linked selection ( background selection [69] and hitchhiking [70] ) . Indeed , when we examined average Δπ for G→GA polymorphisms , we discovered a clear ∼10 bp periodic skew in frequency across nucleosomal regions , mirroring G→A divergence ( per-n147 in Figure 3B and per-site in Figure S12 ) . n147 G→GA Δπ are less negative in regions of higher AA dinucleotide frequency . Interestingly , intronic G→GA Δπ shows even more pronounced periodicity in the frequency spectrum , including the prominent drop at SHL 2 observed for AA frequency ( Figures 3B and S12 ) . Δπ for C→CT polymorphisms is also periodic in introns ( both per-n147 and per-site ) , with peaks aligning with regions of high TT frequency; while intergenic n147 share a subset of these peaks ( Figures 3B and S12 ) , the overall patterns show much weaker periodicity ( see below ) . Although peaks in intronic C→CT Δπ overlap roughly with those for G→GA sites , they show a more convex shape , similar to the C→T divergence . Substitutions in the complementary directions ( e . g . A→AG ) also show a periodic skew in allele frequencies . Introns display a striking periodicity in A→AG Δπ aligned with GC frequency ( Figure 3B , while intergenic n147 A→AG sites show only two peripheral Δπ peaks and several peaks ( valleys ) that are discordant with the GC dinucleotide periodicity . Like underlying GC frequency , intronic A→AG Δπ has a concave upward shape . We observe weaker but interesting indications of continued periodicity in linker regions , consistent with selection for the preservation of rotational positioning in association with translational repositioning . The patterns of Δπ for 5 non-overlapping subsets of n147 regions were similar ( Figure S13 ) . We conclude that for several substitutional pathways there is strong evidence of selection maintaining the observed nucleosomal ( di ) nucleotide preferences . The periodicity of nucleosomal Δπ is not limited to the Raleigh population . The strongest of these periodic patterns in Δπ are also apparent in a smaller , independent set of 21 sequenced genomes from a Rwandan ( Africa ) population [71] ( Figures 3C and S14 ) , which also exhibits more negative average Δπ values . This African sample is assumed to represent a larger , more stable population from the center of the species distribution , while the Raleigh sample represents the serial diasporas out-of-Africa and into North America . Notwithstanding differences in average Δπ , these strong and predicted periodicities in nucleosomal in Δπ support our hypothesis that direct interactions between the histone core and DNA sequence polymorphisms yield functional effects with fitness consequences . An alternative hypothesis to explain these periodicities holds that the sequences evolve independently of natural selection and that the in vivo positions of our isolated nucleosomes reflect the innate preferred rotational positions of the particular genome used . Derived SNPs detected in a single strain are likely to be in high frequency , and thus we might observe periodicity in the frequency spectra at such SNPs in the absence of natural selection . To test for the impact of this hypothesized ascertainment bias on the periodicity of Δπ , we filtered the n147 for those in which the source genome bore the ancestral alleles . Despite the unavoidable thinning of the data , we observed clearly periodic polarized Δπ for those pathways with the strongest initial signals , e . g . intronic G→GA , C→CT and A→AG ( Figure S15 ) . These results indicate that the observed periodicities in the frequencies of preferred bases ( parallel to the dinucleotide frequencies and the divergence ) cannot be attributed to biases in the ascertainment associated with the genotype from which the nucleosomal sequences were prepared . We next considered the values of Δπ surrounding n147 regions . The observed skew in intergenic G→GA:C→CT Δπ extends into adjacent sequence ( Figure 3D ) , tracking with the periodicity of G→A:C→T divergence . Interestingly , in the ∼500 bp flanking n147 regions , there appear to be major and minor Δπ peaks associated with each divergence peak . Given the shoulder of C→CT Δπ values in linker regions adjacent to n147 ( Figure 3A ) , this could represent a nucleosomal and a linker peak . Intronic regions show higher overall values of G→GA:C→CT Δπ and similar , but weaker , indications of increased G→GA:C→CT Δπ associated with nucleosome occupancy ( Figure 3C ) . Among other interesting patterns in Δπ and contrasts to divergence in these flanking regions are those associated with the complementary set of substitutional paths , A→AG:T→TC , which exhibits peaks over apparent linker regions in Δπ but no parallel pattern in A→G:T→C divergence ( Figure S16 ) . On average ( per-n147 ) , G→GA and C→CT are the most common polymorphisms and have among the most positive Δπ , indicating weak positive selection , in addition to being the most rapidly diverging bases ( Figure 3D ) . Although rates of A→G and T→C divergence ( and rates of associated polymorphisms ) are much lower , these types of polymorphic sites also have high average Δπ ( Figure 3E ) . Thus , substitutions with the most periodic divergence and Δπ also show the least overall negative skew in the frequency spectrum . Relative relationships of n147 average π , Δπ and divergence are quite similar to those of a non-overlapping subset and to the genome-wide averages ( Table S4 and Figure S17 ) . These broad genomic patterns appear inconsistent with equilibrium models and may reflect heterogeneity and/or recent ( transient ) shifts in selective forces [35] , [52] , [72] .
Histones are among the most ubiquitous and highly conserved eukaryotic proteins . Thus , it is not surprising that nucleosomal dinucleotide periodicities , which derive from key structural interactions between DNA sequence and the histone core , are shared widely across species . In spite of the near universality of these patterns among eukaryotes and decades of research , our understanding of their functional impact and evolutionary dynamics remain unsettled . In this work we examined genomic variation across regions defined by isolated nucleosomal DNA fragments . Our goal was to first determine if these regions showed interpretable variation in divergence between species , then to analyze population genomic variation for evidence of a role for natural selection in the generation and maintenance of nucleosome-associated sequence variation . We find that divergence on the melanogaster lineage mirrors the sequence preferences of the histone core . This periodic variation in substitution rates across nucleosomal regions indicates that interior minor groove regions display more rapid substitution of AT for GC , and that AT base pairs in regions where the major groove faces inward are more likely to become GC rich . These striking patterns align directly with dinucleotide patterns that stabilize associations between DNA and the histone core , as documented in numerous biochemical and structural studies [5] , [8] , [15] , [18] , [19] , [56] . If nucleosome-bound regions are evolving toward the observed nucleosome sequence preferences , a key question is whether this is the result of mutational bias relative to the positioning of chromatin proteins , or whether it is the consequence of natural selection based on functional differences . The available depth of population data and our new index Δπ allowed us to directly address this question . We find remarkable periodicities in Δπ that parallel the observed patterns of divergence . The spectra of SNP frequencies across n147 regions are variable , with higher Δπ when the inferred ancestral allele is unpreferred , and the derived allele is structurally favored . Therefore , we conclude that selection is , at least in part , driving the maintenance of nucleosome-associated sequence patterns on the melanogaster lineage . If the fitness differences associated with such histone:DNA interactions are largely arising from nucleosomal dynamics ( assembly , disassembly , movement and modification ) and rotational positioning of functional elements , then we can further hypothesize that transcribed ( intronic ) nucleosomal sequences should exhibit stronger periodicity than untranscribed ( intergenic ) nucleosomal sequences . Consistent with this hypothesis , correlations of lineage specific divergence and Δπ with the relevant underlying dinucleotide frequencies are stronger for intronic sequences . Table 1 shows that in each case where a large difference between intergenic and intronic is apparent , it is the intronic that is larger . The two exceptions , G→A & AA and A→G & GC , are those where both correlations are among the highest . As might be expected given the longer timescale and greater number of variable sites , divergence correlates more strongly with dinucleotide frequencies than Δπ . Interestingly , Figures 2C and 3B show that these intronic vs . intergenic differences in correlation of divergence and Δπ with dinucleotide patterns may be attributable to large deviations from expectation in specific regions of the nucleosomal sequences , while other regions follow the expected periodic patterns . While natural selection is the most direct interpretation of these results , interactions between chromatin proteins and DNA damage and repair are well documented [44]–[48] , [73]–[76] . Contextually biased mutation ( substitution ) pathways could underlie the observed periodicities in nucleosomal divergence . However , Drosophila does not have a significant level of 5-methylcytosine [77] , the deamination of which is thought to drive the strong contextual biases ( NpCpG ) in vertebrates [78] . Indeed , a recent genomic sequencing study of Drosophila mutation accumulation lines yielded no evidence for contextual biases [54] . Most importantly , such sequence-contextual as well as nucleosome-mediated biases in mutation rates are excluded as an explanation for the observed periodicities in the skew of the SNP frequency spectrum ( Δπ ) , since strictly neutral mutations should display the same frequency distribution across the genome [31] , [33] , [59] . Support for a role of natural selection maintaining these periodicities is bolstered by the stronger periodicities in intronic nucleosomal sequences , where transcription-associated remodeling and disruption of nucleosome-DNA interactions are more likely to have functional impacts . There is , however , one potential “selectively neutral” mechanism to explain the observed periodic patterns in Δπ . Biased gene conversion ( BGC ) , a process where heteroduplex regions formed between homologs are repaired in a direction favoring one base , can create SNP dynamics analogous to those of directional selection [79] . BGC systematically favoring GC over AT has been observed in a few species and indirectly implicated in others by associations of local GC content with estimated rates of crossing over [80] . However , evidence for such an association is not observed in Drosophila [81] . Given that the magnitudes of average Δπ and its periodicities for G→GA and C→CT SNPs are comparable to those of A→AG , any explanation of our results invoking BGC would have to involve multiple distinct gene conversional biases that depend on nucleosome position . While this is conceivable and worthy of further investigation , we conclude that the canonical GC-biased gene conversion is not a significant component of the evolutionary dynamics leading to these intricate nucleosomal patterns of polymorphism and divergence . Whether these periodic patterns are the product of natural selection or BGC , the magnitude of the average force shaping the dynamics of nucleosomal SNPs must be small compared to that affecting the evolution of nonsynonymous variants . The shifts in G→A divergence between peaks and valleys in the n147 are ∼0 . 001 against a background average of ∼0 . 01 , suggesting relatively weak constraint of 1 in 10 mutations . The non-synonymous rate of divergence on the melanogaster lineage , ∼0 . 006 , is about one tenth of that for synonymous divergence corresponding to 9 out of 10 mutations being selected against [61] . Comparable conclusions could be drawn from the modest magnitudes of periodic fluctuation in expected heterozygosities and , indeed , in the widely observed periodicities in dinucleotide frequencies of nucleosomal sequences . Still , by virtue of its four-fold greater genomic footprint , the net selective impact of just the selection associated with such nucleosomal periodicities could approach the magnitude of non-synonymous variants . As is the case for coding sequence , differences in the relative ( average ) rates reflect the aggregate impact of selection that must vary substantially among nucleosomes , as well as among sites . Evidence of natural selection supporting nucleosome-associated sequence periodicities and the implication of their biological impact casts the potential functions of non-protein-coding regions in a new light . Substantial portions of Drosophila , human and other genomes appear to be under evolutionary constraint , yet lack any functional annotation [67] , [68] . Further , SNPs identified by genome-wide association studies ( GWASs ) of interesting human phenotypes often have mild attributable effects and map to unannotated intronic or intergenic regions , where mechanistic hypotheses concerning the impacts of such genomic variation are lacking . We demonstrate that at least part of the constraint in Drosophila arises from interactions between histone proteins and DNA sequence . Our results suggest dinucleotide periodicities and the rotational positioning that they guide have significant biological consequences . Sequences affecting rotational positioning can influence the binding of transcriptional activators and participate in regulation of expression or gene splicing [25] , [28] , [82]–[84] . More generally , they impact nucleosome assembly and stability [2]–[7] , [9] , [17] , properties that broadly impact chromatin dynamics and may influence higher order chromatin structures . Further , the observed large-scale periodicities in dinucleotide frequencies ( and divergence and Δπ patterns supporting them ) demonstrate that sequences that facilitate rotational positioning are specifically enriched relative to adjacent nucleosomes . So , while periodic sequence patterns are considered more relevant to rotational positioning , they clearly interact with the translational positioning of arrayed nucleosomes in Drosophila . Going forward , deeper and more detailed population genomic analyses should provide a unique window into the complex in vivo interactions between DNA sequence and nucleosome function . The significance of these periodic patterns of polymorphism and divergence is amplified in light of the substantial proportion of the eukaryotic genome packaged in nucleosomes ( four-fold greater than that of coding sequence in Drosophila ) and the broad conservation of dinucleotide interactions with the histone core . Indeed , no other DNA-protein interaction remotely approaches the genomic density or structural impact of nucleosomes . The striking periodic variation we observe relative to nucleosomes fundamentally changes expectations about divergence and SNP frequency , particularly in non-protein-coding regions . Our results point to a layer of evolutionary forces across entire genomes , emanating from the interactions of DNA sequence variation with the structure and function of the histone core .
Embryos were collected from population cages [85] over a 1 hr period and aged at 25°C for 2–3 hr . Staged embryos were dechorionated in 50% bleach for 2 minutes , washed extensively , and then homogenized on ice in SEC buffer ( 10 mM HEPES , 150 mM NaCl , 10 mM EDTA 10% glycerol , 1 mM DTT ) with Protease Inhibitors ( PI ) ( 0 . 1 mM PMSF and 2X Roche EDTA-free Protease Inhibitor tablets ) . After lysate filtering and centrifugation , pelleted nuclei were resuspended in CIB ( 15 mM Tris pH 7 . 5 , 60 mM KCl , 15 mM NaCl . 0 . 34 M Sucrose , 0 . 15 mM Spermine , 0 . 5 mM Spermidine ) +PI and then repelleted . Centrifugation of nuclei in CIB was repeated 3 times , and the resultant pellet was flash frozen and stored at −80°C . Pelleted frozen nuclei were resuspended in CIB+PI , and chromatin was digested with 0 . 5 U/ml Micrococcal nuclease ( Sigma ) for 37°C for 15 min . MNase treated nuclei were pelleted , resuspended in 0 . 1% NP-40 PBS+PI , and incubated at 4°C for 3 hrs to release ( primarily ) mononucleosomes . Nuclei were re-pelleted , and chromatin from the supernatant was phenol-chloroform extracted . Digestion was analyzed on an agarose gel . Approximately 882 ng of DNA was used as starting material for paired-end sequencing library construction following the Illumina protocol ( PE-102-1001 ) . 10 µl of paired end adapter oligos were ligated to the end-repaired , A-tailed fragments in a 50 µl reaction . The adapter-ligation product was gel-purified to select molecules approximately 150–700 bp in length and re-suspended in 30 µl total volume . 1 µl of size-selected ligation product was used as template for 12 cycles of library enrichment PCR in a 50 µl reaction volume . The enriched library was purified using QIAGEN MinElute columns and sequenced ( 2×36 cycles ) on one lane of a flow cell ( FC42JB8 ) with an Illumina GAIIx running the Illumina software SCS v2 . 4 . 135/Pipeline v1 . 4 . 0 . Subsequently , 8 µl of the same size-selected ligation product was used as a template for 10 cycles of library enrichment PCR in a 100 µl reaction volume . To enrich for 147 bp fragments , the library was purified using QIAGEN MinElute columns , then size selected on an agarose gel to recover fragments approximately 273 bp in length ( as determined by an Agilent Bioanalyzer ) . This size-selected library was sequenced ( 2×36 cycles ) on four lanes of a flow cell ( FC61BGN ) with an Illumina GAIIx running the Illumina software SCS v2 . 5 . 38/Pipeline v1 . 5 . 0 . Reads that passed the Illumina pipeline's quality filters were then aligned to the Berkeley Drosophila Genome Project's Release 5 reference sequence [86] using Version 0 . 7 . 0 of the MAQ program [85] . Read pairs that mapped more than 1 , 000 bp apart and those for which the combined sum of the quality scores of mismatches exceeded 300 were filtered using the maq map –a 1000 -e 300 command . Otherwise , default map parameters were used . Mapped paired end clones of length 147 bp were then filtered based on Release 5 . 16 FlyBase Annotation of the D . melanogaster genome and classified as intronic or intergenic . For classification , all bases , including the flanking ±50 bp , were required to map entirely to a contiguous intronic or intergenic region . Heterochromatic reads were removed using cytogenomically-defined boundaries [87] . Downstream analysis was carried out on 276 , 614 intergenic and 270 , 998 intronic autosomal 147 bp nucleosomal fragments , referred to as intronic and intergenic n147 . These cover 61% and 79% respectively of the target intergenic and intronic regions in the euchromatic autosomes ( chr2 and chr3 ) . The coordinates of the n147±50 bp flanking regions are in Tables S1 and S2 . Average nucleosomal read depths , where represented , are a pileup representation of a set of similarly processed paired end clones with lengths ranging from 142–152 bp . In calculating dinucleotide frequencies , divergence , π and Δπ across n147 regions ( and , where relevant , ±50 bp flanking ) , both positional and average calculations took the dyad symmetry into account . Substitutional pathways were switched at the dyad axis , such that positions −73 to −1 were joined to the reverse complement of bases 1–73 . Where included , flanking regions ( ±50 bp ) were treated similarly . For larger scale positional divergence and genomic averages , data from complementary substitutional pathways were combined . The n147 dinucleotide frequencies are the averages ( in the reference sequence ) over all n147 fragments for each position . Genomic over/underrepresentations of dinucleotides were calculated by dividing the difference between observed and expected frequencies by the expected frequency . Estimates of expected intergenic and intronic dinucleotide frequencies were calculated based on underlying base frequencies . Observed frequencies were computed directly from the reference sequence . The sequences of the euchromatic portions of 36 D . melanogaster genomes from Raleigh , North Carolina were released by DPGP ( http://www . dpgp . org/1K_50genomes . html - Reference_Release_1 . 0 ) . The sequencing , alignment and assignment of estimated quality scores are described in Langley . et al . , 2012 [65] . The sequences of the 22 D . melanogaster genomes from Rwanda , Africa were released by DPGP ( http://www . dpgp . org/dpgp2/update_20Jan2012/dpgp2_v2_rg . ID5 . nohets . fastq . bz2 ) . The sequencing , alignment and assignment of estimated quality scores are described in Pool et al . , 2012 [71] . For both data sets , only bases with a minimum quality score of Q30 or greater were included in the analyses . Calculations of divergence on the melanogaster lineage , frequency , expected heterozygosity ( π ) and the index of skew in the frequency spectrum ( Δπ ) were based solely on sites that could be polarized using a multiple alignment of D . melanogaster , D . simulans , D . yakuba and D . erecta genomes [65] , i . e . , simulans and yakuba and/or erecta have the identical base . For consideration of the potential impact of ascertainment bias association with the isolation of nucleosomes , sites were subjected to a more stringent polarization ( simulans , yakuba and erecta have the identical base ) and calculations were done only on sites where the experimental genome had the ancestral allele . These statistics ( divergence , π and Δπ ) were estimated with two alternative weightings . The first , per-n147 , gave weight to genomic sites proportional to their occurrence at nucleosome positions among the n147 . Thus , in instances where a particular site was found multiple times in the n147 set , the divergences or SNPs at that site were give proportionally more weight . This in effect weights the signal by nucleosome site “occupancy . ” The second , per-site , counted normalized the weighting such that each site ( conserved , divergence or SNP ) contributed equally , independent of its recurrence in the n147 . The per-n147 estimates reflect those nucleosomal sequences that were readily isolated , while the per-site method treats each population genomic variant equally . A third representation of the mapping of divergence and Δπ consists of random non-overlapping regions sampled from the n147 . These analyses are presented to simply address whether the periodicities arise solely from a ∼10 bp periodicity in the overlap of n147 fragments . For divergence , non-overlapping n147 sets ( 72 , 710 intronic and 72 , 859 intergenic sequences ) were generated by random sampling of intergenic and intronic n147 without replacement . Newly drawn sequences were added to the non-overlapping subset only if they did not share any positions with prior sampled sequences . These non-overlapping subsets together cover 53% of the target intergenic and intronic regions in the euchromatic autosomes ( chr2 and chr3 ) covered by the full n147 . For non-overlapping Δπ , intergenic and intronic n147 ( not including the flanking ±50 bp ) were first filtered for only those nucleosomal regions containing the relevant SNP ( taking dyad symmetry into account ) . This produced non-overlapping intergenic and intronic sets of ∼90 , 000 regions each for G→GA and C→CT and ∼55 , 000 regions for A→AG . These sets were then subjected to random sampling without replacement . A new n147 was added to the set if it did not overlap any already in the set . Non-overlapping G→GA and C→CT intronic and intergenic sets contained ∼46 , 000 regions each and A→AG sets contained ∼30 , 000 . All three of these methods for calculating divergence and Δπ yielded similarly periodic patterns reflecting the fact that while the genomic coverage of the n147 is not deep , it is also relatively uniform ( cv 0 . 65 ) and the periodicities are not arising from a small subset of the n147 or interactions from overlapping n147 sequences . We require a sample-based index of the skew in the site frequency spectrum . Tajima [59] proffered the test statistic , D , a normalization of d , the difference between two estimates of the same population parameter , 4Nμ , where μ is the mutation rate to selectively neutral alleles and N is the population size . These estimates areandin a sample of size n , where xi is the frequency of one of the two alleles at each of the S segregating sites in an arbitrary genomic segment of length L base pairs . Thus ( expected heterozygosity ) and are per site ( base pair ) estimates of 4Nμ . But here we seek not a test statistic for a genomic segment , but an index of the same deviation , that can be aggregated across heterogeneously sampled data and compared across classes of genomic annotation . To that end considera simple rescaling of Tajima's ( little ) d , Tajima [59] also presents the distribution of the proportion of S segregating sites with frequency i/n in the sample , Gn ( i ) /S . Gn ( i ) can not be used to compute properties of a sample unless one can argue that the sites evolved independently and are sampled independently . If we choose a set of S segregating sites ( assumed to be independently sampled from a population , i . e . no linkage disequilibrium ) , rather than a genomic segment , we have the expected heterozygosity in a sample of size n , and so our index , This index is thus a measure of the deviation of the population ( expected ) heterozygosity per segregating site from its predicted value under the assumptions of equilibrium between selectively neutral mutation and genetic drift in a Wright-Fisher population . To estimate this deviation across sites with different sample sizes , we can calculate the weighted average , weighting by the reciprocal of the variance . The variance of isNotice that this is the theoretical variance in at a single segregating site in a sample of size n under the assumptions of the neutral model ( above ) . If the frequencies at different sites are independent then we can estimate the sample variance of Δπ ( n ) for S sites with sampling depth n , simply as Var[Δπ ( n ) ]/S . Assuming again that these SNPs are sampled independently both within and over sample sizes ( i . e . , no linkage disequilibrium ) , the average Δπ can be estimated by the weighted averagewhere nmin>3 and nmax = largest sample size . Δπ for each position in the n147 regions ( and its average across positions ) was calculated over polarized sites with sampling sizes ( n ) between 32 and 34 in the Raleigh [65] and 18 to 21 for Rwandan data [71] . For n147 average Δπ the n147 regions were trimmed as described above for divergence . Data from complementary pathways were merged for larger scale positional Δπ . n142-152 coverage of intergenic n147 regions was defined as the sum of the coverage of the region by the larger set of 142–152 bp nucleosomal fragments . This corresponds to what some authors call “occupancy . ” GC frequency of intergenic n147 regions was calculated based on the nucleotide frequencies in the D . melanogaster reference sequence from bases 6 to 141 of the n147 regions ( to minimize the potential impact of MNase sequence bias ) . For density plots and Spearman's ρ , only n147 regions with at least 50 ( out of 147 ) intergenic bases polarizable were included in the analyses . As above , the interior 5 bp on each end of n147 regions were trimmed from the analysis . For correlations between divergence and genomic intergenic GC frequencies , Spearman's ρ were reported for non-overlapping 500 bp windows in which at least 166 intergenic bases were polarizable and no more than 250 bp in the reference were “N” . Plots were generated using R [88] . Prior to plotting , all calculations were symmetrized around the dyad axis . n147 ( ±50 bp ) divergence and Δπ plots ( Figure 1D , Figure 2B , C , Figure 3A ) were smoothed using running average in a window of 5 bp ( weights: 0 . 125 , 0 . 250 , 0 . 250 , 0 . 250 , 0 . 125 ) . For large scale plots of dinucleotide frequency , divergence and Δπ ( Figure 1C , Figure 2D , Figure 3B ) , flanking regions were smoothed using running average smoothing with a window of 50 bp of equal weights . In those plots , the central n147 regions were smoothed separately using a 30 bp window of equal weights . Regions of 5 bp upstream and downstream of the n147 edges were trimmed prior to smoothing for large-scale plots . To elucidate the distribution of the smoothed ( as above ) divergence on DNA from the structure of the nucleosome we colored-coded values of each base pair in a schematic rendering of the DNA strands of pdb1kx5 [55] using PyMOL [89] . Only the “top” turn of the DNA ( base pairs 73 to −6 ) is shown . Bases 73 to 70 were rendered as grey , due to extreme values induced by MNase sequence bias .
|
In eukaryotic cells , the majority of DNA is packaged in nucleosomes comprised of ∼147 bp of DNA wound tightly around the highly conserved histone octamer . Nucleosomal DNA from diverse organisms shows an anti-correlated ∼10 bp periodicity of AT-rich and GC-rich dinucleotides . These sequence features influence DNA bending and shape , facilitating structural interactions . We asked whether natural selection mediated through the periodic sequence preferences of nucleosomes shapes the evolution of non-protein-coding regions of D . melanogaster by examining the inter- and intra-species genomic variation relative to these fundamental chromatin building blocks . The sequence changes across nucleosome-bound regions on the melanogaster lineage mirror the observed nucleosome dinucleotide periodicities . Importantly , we show that the frequencies of polymorphisms in natural populations vary across these regions , paralleling divergence , with higher frequencies of preferred alleles . These patterns are most evident for intronic regions and indicate that non-protein coding regions are evolving toward sequences that facilitate the canonical association with the histone core . This result is consistent with the hypothesis that interactions between DNA and the core have systematic impacts on function that are subject to natural selection and are not solely due to mutational bias . These ubiquitous interactions with the histone core partially account for the evolutionary constraint observed in unannotated genomic regions , and may drive broad changes in base composition .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"genomics",
"cell",
"biology",
"chromosome",
"biology",
"genome",
"evolution",
"genetics",
"biology",
"and",
"life",
"sciences",
"population",
"genetics",
"epigenetics",
"computational",
"biology",
"evolutionary",
"biology",
"chromatin"
] |
2014
|
Nucleosomes Shape DNA Polymorphism and Divergence
|
Cryptosporidiosis has emerged as a leading cause of non-viral diarrhea in children under five years of age in the developing world , yet the current standard of care to treat Cryptosporidium infections , nitazoxanide , demonstrates limited and immune-dependent efficacy . Given the lack of treatments with universal efficacy , drug discovery efforts against cryptosporidiosis are necessary to find therapeutics more efficacious than the standard of care . To date , cryptosporidiosis drug discovery efforts have been limited to a few targeted mechanisms in the parasite and whole cell phenotypic screens against small , focused collections of compounds . Using a previous screen as a basis , we initiated the largest known drug discovery effort to identify novel anticryptosporidial agents . A high-content imaging assay for inhibitors of Cryptosporidium parvum proliferation within a human intestinal epithelial cell line was miniaturized and automated to enable high-throughput phenotypic screening against a large , diverse library of small molecules . A screen of 78 , 942 compounds identified 12 anticryptosporidial hits with sub-micromolar activity , including clofazimine , an FDA-approved drug for the treatment of leprosy , which demonstrated potent and selective in vitro activity ( EC50 = 15 nM ) against C . parvum . Clofazimine also displayed activity against C . hominis–the other most clinically-relevant species of Cryptosporidium . Importantly , clofazimine is known to accumulate within epithelial cells of the small intestine , the primary site of Cryptosporidium infection . In a mouse model of acute cryptosporidiosis , a once daily dosage regimen for three consecutive days or a single high dose resulted in reduction of oocyst shedding below the limit detectable by flow cytometry . Recently , a target product profile ( TPP ) for an anticryptosporidial compound was proposed by Huston et al . and highlights the need for a short dosing regimen ( < 7 days ) and formulations for children < 2 years . Clofazimine has a long history of use and has demonstrated a good safety profile for a disease that requires chronic dosing for a period of time ranging 3–36 months . These results , taken with clofazimine’s status as an FDA-approved drug with over four decades of use for the treatment of leprosy , support the continued investigation of clofazimine both as a new chemical tool for understanding cryptosporidium biology and a potential new treatment of cryptosporidiosis .
Cryptosporidium species are apicomplexan protozoans that are important causes of diarrhea in humans and some domestic animals . The parasite relies on an oral-fecal route of transmission , and ingestion of water or food contaminated with Cryptosporidium oocysts may lead to infection . Upon ingestion , oocysts are activated , releasing four sporozoites which then invade host epithelial cells in the small intestine [1] . In some severe cases , infection may expand beyond the gastrointestinal tract and into the respiratory tract—a complication most often seen in patients with human immunodeficiency virus ( HIV ) [2–4] . The life cycle is not well understood , although Cryptosporidium undergoes both asexual and sexual replication within a single host , ultimately leading to the generation of environmentally-hardy infectious thick-walled oocysts that are excreted with the host feces [5] . Acute and persistent watery diarrhea with concomitant oocyst shedding are hallmarks of cryptosporidiosis [6] . Clinically , the two most relevant species that cause human cryptosporidiosis are C . hominis and C . parvum , and distribution of each species varies greatly depending on the region and geography . A large molecular epidemiology study carried out by the Global Enteric Multi-Center Study ( GEMS ) revealed Cryptosporidium spp . to be the second-leading cause of life-threatening diarrheal disease in young children [7] . A second study by MAL-ED confirmed the significant contribution of Cryptosporidium to diarrheal disease burden in infants 12 months and younger [8] . While rotavirus continues to be the most common cause of severe pediatric diarrheal disease , 8–30 . 5% of cases , dependent upon location and age range , are now attributed to cryptosporidiosis [9–13] . In general , severe infectious diarrheal diseases cause dehydration and malnutrition due to low retention of nutrients [14] . Young children are particularly vulnerable to the untoward effects of severe diarrhea , which can result in death or stunted development [8 , 15] . Currently , there is only one approved drug to treat Cryptosporidium infections—nitazoxanide [16] . The efficacy of nitazoxanide has been questioned [17] and appears to be dependent upon a competent immune system . This is notable because young children and immunodeficient individuals are disproportionally affected by cryptosporidiosis [18] , and nitazoxanide demonstrates very poor efficacy in AIDS patients [19] , highlighting an urgent unmet medical need among this comorbid patient population . Drug discovery efforts against cryptosporidiosis have been limited to a few targeted mechanisms in the parasite and whole cell phenotypic screens against small , focused collections of compounds [20–22] . The most advanced compound from these efforts , bumped kinase inhibitor 1294 ( BKI-1294 ) , is a putative inhibitor of C . parvum calcium-dependent protein kinase 1 ( CDPK1 ) [23] which is a validated target in other protozoans , including Toxoplasma gondii [24] and Plasmodium spp . [25] . BKI-1294 has demonstrated efficacy in a mouse model of chronic cryptosporidiosis when dosed once per day for ten consecutive days at 100 mg/kg [23] . Active compounds have also emerged from screens for inhibitors of inosine-5’-monophosphate dehydrogenase [26] , and polyamine analogues [27] . Despite these efforts there remains a dearth of compounds in the drug discovery pipeline , a limited biological understanding , and a lack of available tools to study this parasite . Therefore , we undertook a high-throughput screen to identify compounds that may lead to effective new treatments for cryptosporidiosis .
As a starting point , we developed a whole-cell phenotypic screening platform for testing a library of small molecules . A high-content screen of C . parvum-infected human intestinal epithelial HCT-8 cells was previously developed in a 384-well format by Bessoff et al . [20] to measure C . parvum proliferation over a 48-h period . We modified this assay for automated dispensing and further miniaturization to 1536-well format , both critical factors to support a high-throughput screening campaign ( Fig 1 ) . A small pilot screen was initiated and assay quality was determined by calculating the Z’ value . Despite miniaturization , the 1536-well assay had comparable Z’ values ( 0 . 2–0 . 5; S1 Table ) to the published 384-well assay format when comparing control wells containing DMSO to wells treated with the positive controls nitazoxanide ( NTZ ) and 5-fluoro-2’deoxyuridine ( FDU ) . Like the 384-well assay , a strict hit cut-off for parasite proliferation was applied to avoid high false-positive hit rates created by higher than average well-to-well variability ( Fig 2 ) . Initial pilot screens with a final compound concentration of 1 . 88 μM confirmed that typical hit reconfirmation rates of 40–50% were observed when applying a 70% inhibition cut-off . Because a significant decrease in hit reconfirmation , reflecting a significantly higher false-positive rate , was observed when the hit cut-off was reduced , the 70% inhibition cut-off was applied for all primary screens . The miniaturized , fully automated design of the C . parvum proliferation assay enables a screen throughput of approximately 40 , 000 compounds per assay . We screened two modestly sized compound libraries to validate this assay and identify new anticryptosporidial compounds: a Bioactive ( 10 , 000 ) set assembled by the California Institute for Biomedical Research ( Calibr ) , and the Global Health Chemical Diversity Library ( GHCDL; 69 , 000 ) provided by the University of Dundee Drug Discovery Unit . A primary screen of these libraries revealed highly divergent hit rates that varied by 70-fold and yielded 106 reconfirmed hits ( 75 from the Bioactive set; 31 from the GHCDL; Table 1 ) . These hits were resupplied from high-purity powder stocks and assayed in dose-response . Compounds with a half maximal effective concentration ( EC50 ) against intracellular C . parvum of less than 1 μM ( 49 from Bioactives and 18 from GHCDL ) and no discernable HCT-8 cytotoxicity ( ≥ 10-fold the observed EC50 value ) were deemed selective , potent hits ( 10 from Bioactives and 2 from GHCDL; S2 Table ) . The high attrition of screen hits at this stage was directly attributable to false positives from compounds that demonstrated poor selectivity and were generally cytotoxic to the host HCT-8 line and/or against additional counter-screened mammalian cell lines ( HepG2 and HEK293T ) . The remaining selective compounds were advanced for hit validation against Cryptosporidium hominis . Methods for continuous , in vitro cultivation of Cryptosporidium spp . have not been well established . Instead , in vivo oocyst propagation of C . parvum is performed in calves , a more promiscuous Cryptosporidium species for animal and human infections . For C . hominis , gnotobiotic piglets are the primary means of propagation . The calf is the more accessible model , making C . parvum the more convenient species to screen . However , C . hominis is equally clinically relevant with regards to human cryptosporidiosis . Thus , compounds identified against C . parvum were counter-screened against C . hominis to confirm that these compounds were active against both species . Given that the genomes of these two species exhibit extremely high synteny and sequence identity ( 95–97% ) [28] , compound potency was expected to also be highly correlated . High purity powders of the 10 filtered C . parvum Bioactives screen hits and both GHCDL screen hits were tested against C . hominis ( TU502 ) to confirm activity against both species . Surprisingly , the species-specific activities the screening hits and control compounds , NTZ and FDU , were determined to be only modestly correlated ( R2 = 0 . 7445; Fig 3 ) . While no compounds demonstrated species-exclusive activity , there were notable outliers with differential activities between species , including the known compounds clofazimine and cyclosporine ( ~20-fold more potent against C . parvum ) . Validated compounds that had the greatest potency were all derived from the Bioactive compound collection comprised of known drugs and annotated compounds . The five most potent compounds against C . parvum were Gö6976 ( EC50 = 2 . 5 nM; S1 Fig ) , monensin ( EC50 = 7 nM ) , clofazimine ( EC50 = 15 nM; Fig 4A ) , cyclosporine ( EC50 = 48 nM ) , and MST-312 ( EC50 = 61 nM ) . Three of these compounds were designed to target mammalian enzymes: Gö6976 is a potent inhibitor of protein kinase C ( isotypes α and β ) and of the tyrosine kinases JAK 2 and FLT3 [29 , 30] , whereas MST-312 , also known as Telomerase Inhibitor IX , is a potent , reversible inhibitor of telomerase activity and arrests cells in the G0-G1 phase during the cell cycle [31] . Finally , cyclosporine is an inhibitor of the phosphatase activity of calcineurin with potent immunosuppressive properties that is indicated for graft-versus-host disease , rheumatoid arthritis , psoriasis , and other immune-related diseases [32] . The other two compounds , monensin and clofazimine ( CFZ ) , are both antibiotics . Monensin is an ionophore antibiotic commonly added to cattle feed [33 , 34] and CFZ is an FDA-approved drug for the treatment of leprosy , a disease caused by chronic infection of the bacteria Mycobacterium leprae and Mycobacterium lepromatosis . The inherent limitation of phenotypic screens prevents elucidation of whether the inhibition of C . parvum proliferation is attributable to modulation of the human host cell or to direct inhibition of a Cryptosporidium target . Regardless , Gö6976 , MST-312 , cyclosporine , and monensin all demonstrated a limited therapeutic index when activity against C . parvum was compared to toxicity for mammalian cell lines , or had undesirable effects , and these compounds were therefore deprioritized . The remaining compound , CFZ , is of particular interest based on a very good safety profile in humans with dosage regimens lasting up to three years [35] . We further characterized CFZ in an assay evaluating the cidal activity throughout the parasite lifecycle . We examined short , three-hour compound treatments of in vitro cultures , followed by compound washout , that covered the complete asexual life cycle , believed to be 13–15 h post invasion [36] . The first treatment group had compound added to the cell monolayers 2 h prior to infection with C . parvum sporozoites from excysted oocysts . The treatment interval lasted 3 h before washout with fresh medium; infected cells were allowed to grow until 48 h post infection before fixation . We treated assay plates with NTZ , FDU , BKI-1294 , and CFZ at their EC50 , EC99 , or 3×EC99 value at the designated intervals throughout the lifecycle ( EC99 data shown; Fig 4B ) . Strikingly , CFZ inhibited parasite growth >75% at the EC99 at every time point , indicating that the effect of CFZ is irreversible within 3 h of treatment at any point throughout the parasite’s asexual life cycle . FDU , a cytotoxic pyrimidine analog , also showed strong inhibition of growth across all time points . Though FDU is generally cytotoxic , it has been previously shown to be a substrate of C . parvum thymidine kinase and inhibit C . parvum growth without affecting HCT-8 host cells [37] . Interestingly , both nitazoxanide and BKI-1294 showed the greatest effect on parasite growth at the end of the life cycle ( 10–13 h post infection ) , and at all other time points showed <60% inhibition , even when concentrations were increased to 3×EC99 ( S2 Fig ) . As CFZ is very lipophilic , we sought to determine whether intracellular accumulation was affecting the strong inhibition across each time point . Upon pre-treatment with each compound in a dose-response titration , we observed that both FDU and CFZ were retained by host cells and exhibited near equipotent inhibition of C . parvum proliferation despite compound removal from the assay media by repeated washes ( S3 Table ) . This retention effect likely contributes to the high inhibition of C . parvum proliferation observed at each time point in the washout assay . However , high retention , and subsequent high activity against parasite proliferation , may be a desirable property to promote high in vivo efficacy . The excellent activity profile of CFZ compared to BKI-1294 supported in vivo pharmacokinetic and efficacy studies in a mouse model of cryptosporidiosis . CFZ has been reported to have two remarkable properties: a long drug half-life ( ~70 days ) [38] and very high tissue distribution . This leads to a notable accumulation of CFZ throughout the body [39] , including the skin , producing a strong but reversible pigmentation with chronic administration . In anticipation of advancing CFZ to a mouse model of cryptosporidiosis to determine its efficacy , we characterized the pharmacokinetic properties of this compound solubilized in corn oil or prepared as a suspension in methylcellulose and Tween 80 ( MC-Tween ) . Healthy CD-1 mice dosed with 20 mg/kg CFZ ( 4 mg/mL ) prepared in either formulation demonstrated comparable pharmacokinetic profiles . Plasma concentrations of CFZ peaked at 445 ± 61 ng/mL ( corn oil ) and 288 ± 2 ng/mL ( MC-Tween ) within the first 24 h after oral administration ( Fig 5A ) . The extended biological half-life was evident as plasma concentrations remained stable for at least 20 hours after the initial maximal concentration . Mice dosed with an equivalent amount of BKI-1294 reached a higher peak ( 1282 ± 154 ng/mL ) , though the plasma concentration rapidly dropped to 6 . 7 ± 0 . 3 ng/mL within the first 24 h post dosing . Pharmacokinetic analysis was further extended to a three-day evaluation of CFZ and BKI-1294 elimination in the feces and urine of mice following a single oral dose . Within the first 24 h after oral administration , the mean ( ± SEM ) fecal excretion was 14 . 8% ± 3 . 1% for CFZ in corn oil and 25 . 7% ± 13 . 3% for CFZ in MC-Tween , while BKI-1294 was only excreted unchanged at 8 . 9% ± 0 . 5% ( Fig 5B ) . Continued evaluation of feces for the next 48 h showed mean excretion values < 2% for either formulation of CFZ . Parallel collection and analysis of urine over the first 48 h period revealed < 0 . 15% of the CFZ dosage versus 1 . 6% of BKI-1294 excreted in the urine ( S3 Fig ) . In the absence of a clearly defined pharmacokinetic profile for anticryptosporidial compounds , we surmise that the detectable presence of CFZ in the feces confirms CFZ exposure throughout the entire gastrointestinal tract . The juvenile IFNγ-/- mouse is susceptible to C . parvum infection via the oral route [40] . Despite the previously reported lethality , in our hands , C . parvum-infected IFNγ-/- mice remained overtly asymptomatic , with infection marked by an acute period of intense oocyst shedding ( 3- to 8-days post infection ( p . i . ) ; reproducibly peaking at approximately one thousand oocysts per milligram of feces ) . This reproducible fecal oocyst shedding was measured by flow cytometry to quantify therapeutic efficacy of CFZ benchmarked against the recently published anticryptosporidial compound BKI-1294 ( Fig 6A ) . Three daily oral doses of 10 mg/kg were given over a three-day period beginning at the onset of high-level oocyst shedding ( day 4 p . i . ) . Within 24 h of the end of treatment ( day 7 ) , BKI-1294 had reduced oocyst shedding to 18 . 1% ± 2 . 2% of mock-dosed mice , while CFZ treatment reduced shedding to below 1% of the control group and below the reliable limit of detection ( Fig 6A , inset ) . Rapid efficacy of CFZ was evident . Within two days of the treatment onset , shedding was reduced to 2 . 5% ± 1 . 7% of the control group , suggesting that single-dose treatment might be efficacious . To evaluate the potential for single-dose efficacy , mice were orally gavaged with an augmented infective titer of C . parvum ( 1×106 oocysts ) and treated with one 100 mg/kg oral dose of CFZ on day 4 ( Fig 6B ) . A rapid reduction in oocyst shedding was observed for the CFZ group , with counts decreasing to the reliable limit of detection by the second day after treatment ( 1 . 5% ± 0 . 4% of mock-treated mice; Fig 6B , inset ) . In contrast , the control group of mice continued shedding close to 1×103 oocysts per milligram of feces with the gradual onset of self-resolution not apparent until day 8 p . i . ( 4 days post treatment ) .
Phenotypic screens have been used extensively for neglected tropical diseases to spearhead drug discovery efforts and generate tool compounds for chemical biology . Here we describe the automation and miniaturization of a previously established high-content imaging phenotypic assay to identify 12 anticryptosporidial compounds . The ultimate objective of this work is to discover novel leads for the treatment of cryptosporidiosis and to identify diverse small molecules with anticryptosporidial activity that may become valuable chemical probes to study Cryptosporidium biology . The most promising hit compound , clofazimine ( CFZ ) , is an FDA-approved riminophenazine antibiotic used for the treatment of leprosy that is recognized by the World Health Organization as an essential medicine . The discovery of CFZ as an anticryptosporidial compound , and the subsequent demonstration of efficacy in a mouse model of acute cryptosporidiosis has provided a novel chemical tool to help study and define the pharmacokinetic and pharmacodynamic properties that drive in vivo efficacy , and a potential drug repurposing candidate for the treatment of human cryptosporidiosis . Aside from CFZ , the ability to prioritize anticryptosporidial screening hit compounds currently presents a challenge because a consensus profile of predictive in vitro ADME and in vivo pharmacokinetic characteristics has yet to be established . Recent insights from Gorla et al . [26] showed that in vivo efficacy of C . parvum inosine-5’-monophosphate dehydrogenase ( CpIMPDH ) inhibitors correlated well with the compounds’ uptake and intracellular accumulation in Caco-2 cells , an established model for human intestinal epithelial cells . Conversely , high Caco-2 permeability and systemic exposure in blood plasma within this series had no correlation to efficacy in a mouse model of acute cryptosporidiosis . While these correlations are rational given the intestinal pathogenesis of this parasite , validation for whether intestinal bioaccumulation is a general characteristic of efficacious anticryptosporidial compounds or a specific driver of efficacy for the CpIMPDH inhibitor series evaluated by Gorla et al . remains to be determined . To this end , the in vitro and in vivo pharmacokinetic profiles of CFZ become of considerable interest . CFZ is known to have limited oral bioavailability , requiring it to be encapsulated as a micronized suspension in a lipid-wax base ( Lamprene®; Novartis ) to promote higher absorption in the treatment of the Mycobacterium spp . responsible for leprosy . Its non-optimal systemic exposure has been particularly restrictive for the development of CFZ in the treatment of Mycobacterium tuberculosis , the causative agent of tuberculosis [41] . While compounds for the treatment of tuberculosis require high systemic exposure , this same pharmacokinetic property may not be a limitation for the treatment of Cryptosporidium infections , which primarily affect the epithelium of the small intestine . The intestinal permeability and high lipophilicity of CFZ ensures retention and exposure throughout the gastrointestinal tract , reportedly leading to the highest deposition of CFZ in the jejunum and ileum [39] . Furthermore , the high bioaccumulation in the small intestine and limited compound solubility can lead to crystalline-like CFZ deposits within the cytoplasm of macrophages [42] and intestinal epithelial cells . These intestinal epithelial crystalline-like CFZ deposits are observed with as few as five doses of 200 mg/kg administered over a two-week period in a mouse model [43] . Chronic dosing ( 8 weeks at ~10 mg/kg/day in mice ) has also been shown to lead to CFZ distribution and bioaccumulation in the mesenteric lymph nodes , spleen and fatty tissue [39] . Whereas the selective partitioning of CFZ into specific tissues may be limiting for the treatment of systemic disease , these properties appear to be favorable for efficacy against C . parvum in a mouse model of acute cryptosporidiosis . We have not as yet established whether CFZ directly targets a parasitic pathway or modulates a host-mediated pathway essential to parasite proliferation . However , the differential in vitro activity between C . parvum ( EC50 = 15 nM ) and C . hominis ( EC50 = 340 nM ) assayed in the same in vitro host cell line are suggestive of a parasite-directed mechanism . Interestingly , CFZ has been noted to have both antibiotic and anti-inflammatory effects , and perhaps not surprisingly , it does not have a clearly defined mechanism of action ( MOA ) associated with either activity . CFZ has been implicated in the disruption of the Kv1 . 3 potassium channel , suggesting a potential mechanism by which CFZ selectively modulates immunosuppression involving Kv1 . 3-expressing T cells [44] . Mechanistic studies in mycobacterium suggest CFZ may directly compete with menaquinone , an essential cofactor of the electron transfer chain [45] . Interestingly , Cryptosporidium has a unique , atypical mitochondrial assembly and genes encoding proteins for electron transport or oxidative phosphorylation have yet to be identified [46] . Previously no known homologs to human potassium voltage-gated ( Kv ) channels have been identified in the relevant Cryptosporidium spp . [47] , thus confounding the mechanism by which CFZ may inhibit parasite proliferation . Regardless of a defined MOA , its status as an FDA-approved drug and over four decades of use for the treatment of leprosy warrants further investigation of CFZ as a possible repurposing candidate for treatment of cryptosporidiosis . Recently , a target product profile ( TPP ) for an anticryptosporidial compound was proposed by Huston et al . [48] and highlights the need for a short dosing regimen ( < 7 days ) and formulations for young children ( 0–24 months ) . CFZ has a long history of use and has demonstrated a good safety profile for a disease that requires chronic dosing for a period of time ranging 3–36 months . The safety profile for young children and infants is not yet well defined and will require additional investigation . However , the targeted dosing regimen of < 7 days is a fraction of the minimal time frame for which this drug is administered to leprosy patients , and the once-daily oral administration of 10 mg/kg CFZ for three days in a mouse model of cryptosporidiosis is likely to translate favorably to human treatment . Importantly , it may be possible that all three doses could be delivered within a single 12 to 24–hour period to achieve rapid efficacy , comparable to the single-dose cure . The differential activity of CFZ between Cryptosporidium spp . raises an additional consideration that will require testing to determine the dose and regimen to treat C . hominis . These parameters will continue to be relevant as CFZ is evaluated in cryptosporidiosis clinical trials , and as the remaining 11 hits from this screen are investigated for further preclinical development .
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health . The protocol ( S13013 ) was approved by the Institutional Animal Care and Use Committee of the University of California , San Diego ( Animal Welfare Assurance Number: A3033-01 ) . All efforts were made to minimize suffering of animals employed in this study . Two compound libraries were assayed against Cryptosporidium parvum in vitro . The first , the Bioactives Set , was a compilation of 10 , 453 known bioactive compounds , including approved drugs and annotated tool compounds . The second was a diversity set designed by the University of Dundee Drug Discovery Unit , generously gifted for screening against C . parvum , totaling 68 , 689 unique drug-like compounds . The Global Health Chemical Diversity Library is a lead-like diversity library prepared for screening against the Bill and Melinda Gates Foundation priority pathogens . The compounds were commercially sourced from reputable vendors , selected to have lead-like physicochemical properties , enhanced Fsp3 character , enrichment in novel chemotypes , and filtered to remove unwanted groups . 200 compounds from this library were not tested due to insufficient source volume or transfer failures . Data from the GHCDL portion of the screen will be made available online from the University of Dundee Drug Discovery Unit , published at ChEMBL-Neglected Tropical Disease ( https://www . ebi . ac . uk/chemblntd ) . Resupplied compounds for confirmation were high-purity powders ( > 95% purity ) from Selleck Chemicals , Calbiochem , Ambinter , Enzo Life Sciences , NCI , ChemDiv , or Life Chemicals . Screen control compounds were 30 μM nitazoxanide ( Sigma-Aldrich ) and 0 . 5 μM floxuridine ( Sigma-Aldrich ) as controls for C . parvum inhibition , and 10 μM puromycin ( Sigma-Aldrich ) for host cell cytotoxicity . Clofazimine ( USP reference standard ) was purchased from Sigma-Aldrich . All source plates were 384-well acoustic transfer-compatible plates ( Labcyte ) with compounds pre-diluted in dimethyl sulfoxide ( DMSO ) at either 1 mM ( Bioactives ) or 2 mM ( all others ) . For single-point testing , compounds were transferred into 1536-well tissue culture-treated , black-walled , clear-bottomed , low base microwell plates ( Greiner ) with an ECHO liquid handler ( Labcyte ) to a final concentration of 1 . 88 μM . For dose-response testing , compounds were serially diluted 1:3 in 11-points and then transferred into triplicate 1536-well plates , with a top concentration of 12 . 5 μM . Primary screening was done with a single replicate , whereas single-point reconfirmation and dose-response testing was carried out in triplicate ( 3 technical replicates ) . The number of biological replicates for filtered hits is indicated in S2 Table . Human ileocecal adenocarcinoma ( HCT-8; ATCC CCL244 ) cells were maintained in T-75 tissue culture flasks with RPMI 1640 medium supplemented with L-glutamine , 10% heat-inactivated fetal bovine serum ( HI-FBS ) , 100 IU penicillin , and 100 mg/mL streptomycin . At confluency , cells were trypsinized , washed , and resuspended in assay medium: RPMI 1640 ( - ) phenol red , supplemented with 2% heat-inactivated horse serum ( ATCC ) , 100 IU penicillin , and 100 mg/mL streptomycin . Cells were then plated ( 5 μL/well ) into 1536-well assay plates at a density of 5 . 5×105 cells/mL ( 2 , 750 cells/well ) . Cells were allowed to grow for 24 h at 37°C with 5% CO2 in a humidified tissue culture incubator . C . parvum oocysts ( Iowa strain , isolated from infected calves ) were purchased from the Sterling Parasitology Laboratory , University of Arizona , and stored at 4°C for ≤ 3 months in an antibiotic solution ( 0 . 01% Tween 20 containing 100 U/mL penicillin and 100 μg/mL gentamicin ) . C . hominis oocysts ( TU502 , isolated from gnotobiotic piglets ) were purchased from the Tzipori Laboratory , Cummings School of Veterinary Medicine , Tufts University , and stored at 4°C for ≤ 2 weeks in an antibiotic solution ( 100 U/mL penicillin and 100 μg/mL streptomycin ) . For infection of assay plates , 24 h after cell seeding , oocysts were excysted and prepared for inoculation as previously described [49] . Briefly , the oocysts were diluted into 1 mL pre-warmed 10 mM HCl and then incubated in a 37°C water bath for 10 min . Oocysts were then pelleted at 3 , 000×g for 1 min in a microcentrifuge ( Thermo ) , and the supernatant was carefully aspirated . Oocysts were resuspended in 2 mM sodium taurocholate in DPBS++ ( with 0 . 9 mM CaCl2 and 0 . 5 mM MgCl2 ) and incubated at room temperature for 20 min . After incubation in bile salts , the oocysts were diluted 1:100 and free sporozoites and intact oocysts were enumerated to determine the percentage of excystation . The oocysts were then diluted with assay medium to 1 . 04×106 oocysts/mL ( 3 , 125 oocysts/well ) and dispensed ( 3 μL/well , 8 μL/well final volume ) with a MultiFlo FX Multi-Mode Dispenser ( Biotek ) . Plates were then spun at 150×g for 3 min in a Sorvall Legend XTR benchtop centrifuge ( Thermo ) . Infected cells were incubated at 37°C with 5% CO2 in a humidified tissue culture incubator covered with metal assay lids ( The Genomics Institute of the Novartis Research Foundation ) for 48 h . Following incubation , infected cells were fixed with 4% paraformaldehyde for 15 min at room temperature . Plates were then washed twice with phosphate buffered saline ( PBS ) . Prior to staining , cells were permeabilized with 0 . 25% Triton X-100 in PBS for a maximum of 15 min at room temperature , and then washed twice with PBS-T ( 1× PBS with 0 . 1% Tween 20 ) . To prevent non-specific binding , the cells were blocked with SuperBlock™ T20 blocking agent ( Thermo ) for 1 h at room temperature . Cryptosporidium parasites were stained with 1 μg/mL fluorescein isothiocyanate ( FITC ) -conjugated Vicia villosa lectin ( Vector Laboratories ) in 1:10 diluted SuperBlock™ in PBS-T , supplemented with 3 μM 4' , 6-diamidino-2-phenylindole ( DAPI ) to visualize host cell nuclei . Staining was for 1 h at room temperature in the dark . Finally , cells were washed twice with PBS-T and the plates were sealed with adhesive foil . The plates were imaged with a CellInsight CX5 High Content Screening Platform ( Thermo ) with a 10× objective . Two channels were used: 384/440 nm for DAPI-stained nuclei , and 485/521 nm for FITC-lectin-labeled Cryptosporidium parasites . For 1536-well format assays ( primary screening , triplicate reconfirmation , and dose-response reconfirmation , one microscopic field ( 802 , 511 . 39 μm2 ) per well , and for the 384-well washout assay , 4 fields per well were captured . The software identified primary objects ( HCT-8 host cells ) and spots within allowed distances to the nuclei ( Cryptosporidium ) . Both cytotoxicity against HCT-8 cells ( number of nuclei relative to DMSO-treated controls ) and Cryptosporidium inhibition ( spot counts relative to DMSO-treated controls ) were assessed . Images were processed by the HCS Studio Scan software , and Selected Object Count ( HCT-8 cells ) and Spot Count ( Cryptosporidium ) were analyzed in Genedata Screener ( v13 . 0-Standard ) . Spot Count and Selected Object Count were normalized to neutral controls minus inhibitors ( floxuridine for Spot Count , and puromycin for Selected Object Count ) . For single-point primary assay plates , an additional run-wise median correction was applied to reduce screen artifacts ( e . g . uneven dispense ) , whereas no correction was applied to triplicate reconfirmation and dose-response plates . For primary screening , a compound is considered a hit by a ≥ 70% reduction in Corrected Spot Count . Compounds that caused moderate cytotoxicity were filtered out ( i . e . , excluding all Corrected Selected Object Counts ≤ -60% ) . Dose-response curves were fit with Genedata Analyzer using the Smart Fit function . Final filtered hits included those with an EC50 ( half-maximal effective concentration ) ≤ 1 μM , with a CC50 ( half-maximal cytotoxic concentration ) ≥ 10-fold greater than the EC50 value . Two mammalian cell lines were used for counter-screening for general cytotoxicity of hit compounds: human embryonic kidney cells ( HEK293T; ATCC CRL-3216 ) and human hepatocellular carcinoma cells ( HepG2; ATCC HB-8065 ) . Each were maintained in T-150 tissue culture flasks with DMEM supplemented with 10% HI-FBS , 100 IU penicillin , and 100 mg/mL streptomycin . At 80% confluency , cells were trypsinized , washed , and resuspended in assay medium: DMEM supplemented with 2% HI-FBS , 100 IU penicillin , and 100 mg/mL streptomycin . Compounds were pre-spotted into tissue culture-treated white solid-bottomed 1536-well plates ( Greiner ) in a 1:3 dose-response dilution ( top concentration 20 μM ) . HEK293T and HepG2 cells were diluted to 75×103 cells/mL and 150×103 cells/mL , respectively , and 5 μL/well were dispensed into assay plates with a MultiFlo FX Multi-Mode Dispenser ( Biotek ) . Cells were incubated with metal lids ( The Genomics Institute of the Novartis Research Foundation ) at 37°C with 5% CO2 in a humidified tissue culture incubator for 72 h . At the completion of the assay , CellTiter-Glo ( Promega ) was prepared at 1:2 ( reagent:water ) of the manufacturer’s instructions , and 2 μL were dispensed into each well . After a 5 min incubation at room temperature , luminescence readings were measured with an EnVision Multilabel Plate Reader ( Perkin Elmer ) . Relative fluorescence units were uploaded into Genedata Screener ( v13 . 0-Standard ) , and data normalized to DMSO- and puromycin-treated wells . A four-parameter non-linear regression curve fit was applied to dose-response data using Genedata to determine the half maximal cytotoxic concentration ( CC50 ) of each compound . Infected cells were treated in six 3-h increments over the course of 15 h ( the presumed in vitro asexual lifecycle of C . parvum ) . HCT-8 cells were seeded ( 25 μL/well ) into clear-bottomed 384-well plates ( Greiner ) at a density of 6 . 0×105 cells/mL ( 15 , 000 cells/well ) in assay medium , and allowed to grow 24 h at 37°C with 5% CO2 in a humidified tissue culture incubator , covered with a custom metal lid ( The Genomics Institute of the Novartis Research Foundation ) to reduce evaporation . After 24 h , and 2 h before infecting with C . parvum oocysts ( time -2 hpi ) , the treatment time course was initiated . Compounds ( NTZ , FDU , BKI-1294 , and CFZ ) were used at their EC50 , EC99 , and 3×EC99 value to determine the extent of parasite proliferation inhibition resulting from a 3-h exposure to each compound . DMSO was used as a negative control . Cell medium was removed from designated wells and replaced with medium containing either compound or DMSO . At time 0 ( infection ) , C . parvum ( Iowa strain ) oocysts were excysted as above , diluted in assay medium to 1 . 04×106 oocysts/mL ( 15 , 625 oocysts/well ) and dispensed ( 15 μL/well , 40 μL/well final volume ) . Plates were then spun at 150×g for 3 min , and kept at 37°C with 5% CO2 in a humidified tissue culture incubator during drug treatment intervals . At the end of each 3 h time point , infected cells were carefully washed three times with fresh , pre-warmed assay medium , and the next set of wells for the subsequent time point were treated . The final time point extended to 18 h , to cover re-invasion of merozoites into new HCT-8 cells . After the final compound treatment was washed off , infected cells were allowed to grow until 48 h post infection . To examine if cellular accumulation of compound was inhibiting parasite proliferation , host cells were pretreated for 3 h with NTZ , FDU , BKI-1294 , or CFZ ( in 1:3 11pt dose-response; top concentrations were 30 μM for NTZ; 0 . 5 μM for FDU; 12 . 5 μM for BKI-1294 and CFZ ) , washed , and then infected with C . parvum . A set of parallel wells were also treated and not washed prior to infection . 48 hpi , infected cells were imaged as described above , and EC50s for each compound pre- and post-wash were determined . Male CD-1 fasted mice ( three per group ) were dosed per os with 20 mg/kg CFZ , formulated at 4 mg/mL in either 100% corn oil or 0 . 5% methylcellulose/0 . 5% Tween 80 , or 20 mg/kg BKI-1294 formulated at 4 mg/mL in 7% Tween 80 , 3% ethanol , and 90% water , and then monitored for plasma concentration , renal excretion , fecal excretion , as well as other clinical markers of adverse events ( severe weight loss , lethargy , hunched posture , social isolation ) for a total of 72 h . Four week old female C57BL/6 IFNγ-/- mice were purchased from the Jackson Laboratory and acclimated for four to ten days in specific pathogen-free conditions in the Health Sciences Biomedical Research Facility at the University of California , San Diego . Mice were provided water and chow ( Teklad 2920X ) ad libitum . For inoculation , C . parvum oocysts were adjusted to a final density of 5×104/mL in cold , sterile , distilled water . Mice were infected via oral gavage with 200 μL ( 104 oocysts ) using a 20G×1 . 5” feeding needle . On days 4 , 5 , and 6 post-infection , mice were gavaged with 10 mL/kg clofazimine ( USP reference standard , Sigma Aldrich ) solubilized in food-grade corn oil . Treatment regimens included 10 mg/kg , 100 mg/kg , and 0 mg/kg ( vehicle only; n = 4 per group ) . Certain assays also involved gavage of characterized anticryptosporidial BKI-1294 ( a kind gift from W . Van Voorhis , U Washington ) , administered at 10 mg/kg mouse as a 10 mL/kg emulsified suspension in 0 . 5% methylcellulose + 0 . 5% Tween-80 . At intervals between 3 and 30 days post-infection mice were weighed and temporarily placed in isolation to allow for collection of feces ( 3 pellets/mouse/sampling point ) . Pellets were weighed and placed in 0 . 5 mL 2 . 5% potassium dichromate solution , and stored at 4°C until processing . Oocysts were extracted via a modified discontinuous sucrose gradient technique [50] . Briefly , Sheather’s solution was freshly prepared by dissolving 156 . 25 g sucrose and 2 . 5 mL phenol in 100 mL water . Fecal pellets were removed from storage then homogenized by vortexing and pipetting . In microcentrifuge tubes , 0 . 2 mL fecal homogenate was overlaid on 0 . 75 mL of solution with a specific gravity of 1 . 064 ( 20% Sheather’s and 0 . 1% Tween 80 in PBS ) , overlaid on 0 . 75 mL of solution with a specific gravity of 1 . 103 ( 33% Sheather’s and 0 . 1% Tween 80 in PBS ) . Oocysts were floated from feces by centrifugation at 1 , 000×g for 20 min and collected with a pipet tip from the 1 . 064/1 . 103 specific gravity interface . Oocysts were rinsed once in cold PBS , pelleted by centrifugation at 15 , 000×g for 10 min , and resuspended in PBS . 50 μL aliquots were incubated for 30 min at room temperature with 0 . 25 μg fluorescein isothiocyanate-conjugated mouse anti-Cryptosporidium antibody ( OW50-FITC , BioRad 2402–3007 ) , then diluted to 200 μL with PBS . Samples were analyzed using a Guava EasyCyte flow cytometer and CytoSoft Data Acquisition and Analysis software ( v5 . 3; Guava Technologies , Inc . ) , using a 100 s sampling interval , 0 . 59 μL/s flow rate , and logical gating of forward and side light scatter and OW50-FITC fluorescence signals . Each experimental run included positive and negative controls to calibrate region settings discriminating the oocyst-FITC population from background signals . Oocyst counts/mL sample values were exported to Excel ( Microsoft Corp . ) for normalization to counts/mg feces . Final graphing and statistical analyses of data were done using GraphPad Prism software ( v6 , GraphPad Software , Inc . ) . Two-way ANOVA with Dunnett’s correction ( Fig 4A ) and unpaired multiple t-tests with False Discovery Rate approach ( Fig 4B ) were performed on % Vehicle data to determine significance between vehicle-treated and compound-treated mice .
|
Diarrheal diseases cause significant morbidity and mortality in the developing world . A recent multi-site study investigating diarrheal disease identified Cryptosporidium as one of the leading causes , responsible for upwards of 25% of cases in some regions . Currently approved therapy for cryptosporidiosis is limited to nitazoxanide , which is approved only for children ages 1–11 years . Nitazoxanide demonstrates moderate efficacy in immunocompetent patients and poor efficacy in immunodeficient patients . The limited treatment options underscore an opportunity to reduce the global burden of diarrheal disease by developing novel medicines with increased efficacy against Cryptosporidium spp . We report the largest high-throughput screen for anticryptosporidial compounds to date , leading to the identification of the FDA-approved drug clofazimine as a potential new treatment for cryptosporidiosis , and as a tool to probe the biology of this pathogen .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"parasite",
"groups",
"medicine",
"and",
"health",
"sciences",
"oocysts",
"viral",
"transmission",
"and",
"infection",
"microbiology",
"cryptosporidium",
"parasitic",
"diseases",
"parasitic",
"protozoans",
"parasitology",
"animal",
"models",
"apicomplexa",
"model",
"organisms",
"protozoans",
"pharmaceutics",
"experimental",
"organism",
"systems",
"molecular",
"biology",
"techniques",
"cryptosporidiosis",
"research",
"and",
"analysis",
"methods",
"mouse",
"models",
"cryptosporidium",
"parvum",
"molecular",
"biology",
"molecular",
"biology",
"assays",
"and",
"analysis",
"techniques",
"host",
"cells",
"library",
"screening",
"virology",
"biology",
"and",
"life",
"sciences",
"drug",
"therapy",
"organisms"
] |
2017
|
A high-throughput phenotypic screen identifies clofazimine as a potential treatment for cryptosporidiosis
|
The accurate diagnosis of sporotrichosis and identification at the species level are critical for public health and appropriate patient management . Compared with morphological identification methods , molecular diagnostic tests are rapid and have high sensitivity and standardized operating processes . Therefore , we designed a novel multiplex real-time polymerase chain reaction ( PCR ) method based on the calmodulin ( CAL ) gene for the identification of clinically relevant Sporothrix species: S . globosa , S . schenckii s . str . , and S . brasiliensis . We evaluated the assay with clinical and spiked samples and assessed its diagnostic performance by comparing the results to those of culture and species-specific PCR . Thirty-three DNA templates were used to detect assay specificity , and three plasmids were constructed to create a standard curve and determine the limits of detection ( LODs ) . For nucleic acid detection , the sensitivity and specificity reached 100% . The LODs were 10 copies , 10 copies and 100 copies for S . globosa , S . schenckii s . str and S . brasiliensis , respectively . For the clinical samples , the positive detection rates by culture , species-specific PCR and the multiplex real-time PCR assay were 87 . 9% ( 29/33 ) , 39 . 4% ( 13/33 ) , and 93 . 9% ( 31/33 ) , respectively . For the spiked samples , the positive detection rates were both 100% for S . schenckii s . str and S . brasiliensis . Based on the above results , compared with culture and other molecular diagnosis methods , the novel multiplex real-time PCR assay is effective , fast , accurate , and highly sensitive . It has a lower reaction cost and lower sample volume requirements , can detect co-infections , and allows for standardized operation and easier interpretation of results . In the future , this assay could be developed into a commercial kit for the diagnosis and identification of S . globosa , S . schenckii s . str , and S . brasiliensis .
Sporotrichosis is a subacute or chronic infectious disease caused by dimorphic fungi of Sporothrix spp . , which are distributed worldwide , especially in tropical and subtropical regions[1–2] . The infection generally occurs through traumatic inoculation of contaminated plant debris[3–4] or through bites and scratches from infected animals , mostly felines[5–6] . The disease can occur sporadically or in outbreaks[7–8] . The primary lesion is usually restricted to the skin and subcutaneous tissue but can subsequently affect adjacent lymphatic vessels[9] . Rarely , this fungus can disseminate through the blood or the lymphatic system and eventually lead to a systemic infection[3] . The pathogen of sporotrichosis , S . schenckii sensu lato , was recognized as the sole agent for more than a century following its first isolation in 1898 by Benjamin Schenck[10] . However , based on macroscopic characteristics and physiological and molecular aspects , S . schenckii sensu lato is considered to be a complex of several distinct species , including S . brasiliensis , S . mexicana , S . globosa , and S . schenckii s . str . [11] , S . luriei[12] , S . pallida[13–14] , and S . chilensis[15] . Further , S . globosa , S . schenckii s . str , S . brasiliensis are medically important species within the Sporothrix genus . These species differ in ecology , distribution and epidemiology . Furthermore , widespread variations in virulence and antifungal susceptibility among these species have been demonstrated . S . brasiliensis , which is associated with severe clinical forms of sporotrichosis[16] , is considered the most virulent species , followed by S . schenckii s . str . and S . globosa . Therefore , identification at the species level is critical for public health and appropriate patient management[17 , 18] . Sporotrichosis can be diagnosed through a combination of clinical manifestation and epidemiological and laboratory tests , including direct examination , culture , histopathological examination , molecular detection , sporotrichin skin tests and antibody detection[18] . It is difficult to detect the parasitic budding yeast cells by direct examination , likely because they are too small ( 2 to 6 μm in diameter ) . Yeast cells can be observed in tissue by histopathological examination using Haematoxylin and eosin ( H&E ) , Gomori methenamine silver ( GMS ) or Periodic acid-Schiff ( PAS ) stain[1] . However , neither direct examination nor histopathological examination can identify the pathogen at the species level . The “gold standard” for diagnosing sporotrichosis is based on conventional culture of clinical specimens obtained from active lesions , pus , secretions or biopsies . After culture on Sabouraud agar ( SDA ) for 5 to 7 days at 28°C , filamentous hyaline colonies start to grow and then develop a dark colour , especially from the centre of the colonies[19] Positive cultures provide the strongest evidence for sporotrichosis , and isolates obtained from culture can be tested for antifungal susceptibility and phenotypic characterization . With the development of molecular biology , an increasing number of methods based on nucleic acid detection have been applied for the rapid diagnosis of infectious disease . Many molecular diagnostic tests , including DNA sequencing of “barcoding” genes[20–23] , nested PCR[24–25] , PCR fingerprinting[26] , restriction fragment length polymorphism ( RFLP ) of different gene targets[27] , random amplified polymorphic DNA ( RAPD ) [7] , amplified fragment length polymorphism ( AFLP ) [8] , rolling circle amplification ( RCA ) [28] and species-specific primers[29] , have been developed for Sporothrix spp . detection . However , there are still some shortcomings , such as time-consuming procedures ( PCR sequencing for at least 12 h ) ; complicated operation steps , which can increase the chance of contamination ( nest PCR ) ; insufficient sensitivity ( RCA , 3 × 106 copies ) ; and so on . Most of them can only identify isolates from culture , and only a few methods have been evaluated with clinical samples[24–25] . In addition , none of the above methods can detect co-infection simultaneously . In the present study , we developed a novel multiplex real-time PCR assay to identify the mainly clinical pathogenic species S . globosa , S . schenckii s . str . and S . brasiliensis , and we evaluated the assay with clinical and spiked samples .
The analytical specificity was examined using 33 DNA templates , including from fungi ( 28 ) , bacteria ( 3 ) , a human ( 1 ) and a mouse ( 1 ) ( S1 Table ) . None of the 33 templates yielded positive signals in the assays; furthermore , nonspecific amplification was not detected . Excluding the negative controls , all 25 Sporothrix templates , including S . globosa ( 21 ) , S . schenckii s . str . ( 3 ) and S . brasiliensis ( 1 ) , yielded positive signals , and the positive detection rate for the nucleic acid templates of the assay was 100% . Standard curves ( Ct vs . log CFU ) for S . globosa , S . schenckii s . str . and S . brasiliensis were constructed using the plasmid DNA template by serial 10-fold dilution ( S1 Fig ) . In addition , the results indicated that the LODs of this assay were 10 copies , 10 copies and 100 copies for S . globosa , S . schenckii s . str . and S . brasiliensis , respectively . For the mixed templates , the multiplex real-time PCR assay could detect the corresponding fluorescent signals in the same tube without nonspecific amplification . The Ct values of different amount templates are shown in Table 1 . The Ct values of the single fluorescence real-time PCR for S . globosa , S . schenckii s . str . , and S . brasiliensis were 21 . 67±0 . 15 , 24 . 64±0 . 15 , and 27 . 00±0 . 11 , respectively , while under the same templates and condition , the Ct values obtained from the multiplex real-time PCR were 21 . 80±0 . 07 , 24 . 77±0 . 07 , and 27 . 32±0 . 08 , respectively . There was no significant difference in Ct values between multiplex and single fluorescence real-time PCR except for S . brasiliensis ( t-test , p = 0 . 03 ) . A total of 40 samples from patients suspected of sporotrichosis were collected ( S2 Table ) . Seven samples were eliminated based on culture and histopathological examination results . The results of the multiplex real-time PCR and species-specific PCR were negative for these 7 samples . Of the 33 selected samples , the positive detection rates of the culture , species-specific PCR and multiplex real-time PCR assays were 87 . 9% ( 29/33 ) , 39 . 4% ( 13/33 ) , and 93 . 9% ( 31/33 ) , respectively ( Table 2 ) . The positive detection rates of the culture and multiplex real-time PCR assays were not significantly different ( paired χ2 , p = 0 . 4142 ) . Differences were observed between the multiplex real-time PCR assay and species-specific PCR ( paired χ2 , p<0 . 0001 ) . The isolates from culture were identified by sequencing the CAL gene , and all 29 strains were S . globosa . Among the 33 samples detected by the multiplex real-time PCR assay , the Ct values of 31 were less than 40 and were judged as positive; the Ct values of 2 samples were greater than 40 and were judged as negative . Only FAM fluorescence was detected; therefore , all of the samples were identified as S . globosa infections , consistent with the sequencing results from the cultured isolates . Among the 33 samples , 11 were positive by all three methods , and 27 were positive by both culture and multiplex real-time PCR . In addition , culture was positive and multiplex real-time PCR was negative for 2 samples , and culture was negative and multiplex real-time PCR was positive for 4 samples , of which two were also positive by species-specific PCR . No false positives were detected from the negative control group of spiked samples , and the positive detection rates of S . schenckii s . str . , and S . brasiliensis were both 100% ( 6/6 ) , while the Ct values were 33 . 03–38 . 57 ( S . schenckii s . str . ) and 30 . 23–34 . 84 ( S . brasiliensis ) .
In 2006 , Marimon et al . [21] reported Sporothrix complex for the first time , which led to many studies of the differences between species in the Sporothrix genus[16–17 , 22 , 30–32] . These studies showed the variations between different species and highlighted the need to identify the species level of Sporothrix spp . . Calmodulin ( CAL ) , internal transcribed spacer ( ITS ) , and elongation factor ( EF ) , which are recognized as fungal "barcoding" genes , are widely applied in the identification of Sporothrix spp . [20–23] . However , all of the methods are based on conventional PCR . Compared to conventional PCR , real-time PCR has many extraordinary advantages , such as rapidity , sensitivity and low risk of contamination . Due to these strong points , real-time PCR is widely used in pathogen detection . However , the application of real-time PCR to Sporothrix spp . identification has not yet been reported [18] . In this study , by comparing sequences of the various “barcoding” genes , a target sequence , which can be used to design the primers and probes for three pathogenic species , was found on the calmodulin gene of Sporothrix spp . . Based on this finding , we established a multiplex real-time PCR to detect S . globosa , S . schenckii s . str and S . brasiliensis simultaneously , which not only can improve the sensitivity of Sporothrix spp . detection but also could save detection time and costs . Furthermore , the detection ability of co-infections was assessed by 9 combinations of different amounts of plasmids . The results showed that , when the amounts were different , the amplification of the smaller amount plasmid was affected , and because of the competitive inhibition , the greater the difference in the proportion of plasmids was , the greater the effect was on the Ct value of the small amount of plasmid . For the same template , the Ct values were not significantly different between multiple and single fluorescence for S . globosa and S . schenckii s . str . Only the intensity of the fluorescent signal was weakened , but this decrease did not affect the judgement of the result . However , a difference in Ct value was observed for S . brasiliensis ( p = 0 . 03 ) , likely due to the weak amplification efficiency compared to the other two . Considered together , these results demonstrated that the assay was capable of detecting co-infections and that there was almost no mutual interference between the primers and probes in the multiplex system . In addition , the clinical samples were used to evaluate the performance of the assay compared with the “gold standard” diagnostic and the latest molecular diagnosis methods of culture and species-specific PCR . Culture , as the “gold standard” for sporotrichosis diagnosis , is widely used in clinical practice . However , species-level identification requires further phenotypic identification and physiological tests , which require at least 3–4 weeks . During this period , contamination by fast-growing fungi or bacteria is likely happening . Some patients undergoing antifungal treatment can also have negative culture results . Therefore , histopathological examination results , such as detection of Sporothrix spp . yeast cells or typical pathological structures , are also combined with culture results for clinical diagnosis . In this experiment , these diagnostic criteria were followed when the clinical samples were collected . Results were considered negative only if the culture and histopathological examinations both excluded the diagnosis . In the present study , the positive rate of culture was 87 . 9% ( 29/33 ) . Of the 4 negative results , 2 samples had contamination from other microorganisms and were judged to be negative , while the multiplex real-time PCR and species-specific PCR results of the 2 samples were both positive . Another 2 samples did not grow after 30 days in culture , while the multiplex real-time PCR results were positive . It is speculated that these 2 cases might have been treated with antifungal therapy or affected by the location of the biopsy . Based on the different lengths of the CAL gene sequences , species-specific PCR[29] , reported by Rodrigues et al . , could identify S . brasiliensis , S . schenckii s . str , S . globosa , S . mexicana , S . pallida , and its relative , Ophiostoma stenoceras . The reaction conditions for this species-specific PCR included 35 cycles , and amplification could not be detected after 35 cycles . In this experiment , according to the results of the multiplex real-time PCR , which was performed for 45 cycles , most of the Ct values were greater than 35 ( 22/33 ) . In this situation , the results of species-specific PCR were often negative . Until now , no reports of S . brasiliensis have appeared , and there have been only four known isolates of S . schenckii s . str . found in China[33] . To compensate for the singularity of pathogens in clinical specimens , an evaluation of the spiked samples was performed . Yeast cells , the pathogenic phase of Sporothrix spp . , were selected for mixing with negative tissue samples , and positive results were obtained from all of the spiked samples , which were mixed with different numbers of yeast cells . In addition , the standard curves of assay were established on direct dilution of plasmids , whereas the clinical samples were prepared using DNA extraction kit . Since DNA lost is unavoidable during the extraction process , the sensitivity of the multiplex real-time PCR was applied to the extracted DNA samples , not the original clinical samples . Considering that different extraction methods result in varying degrees of DNA lost , it is recommended to use a method with high yield of DNA for clinical samples extraction . In conclusion , the novel multiplex real-time PCR assay was effective , fast , accurate , and highly sensitive . It had a lower reaction cost and sample volume requirements , could detect co-infections and allowed for standardized operation and easier interpretation of results . However , the assay must still be validated with clinical samples of S . schenckii s . str and S . brasiliensis . In the future , the number of clinical samples used to validate the assay must be increased , and the assay could be further verified using pus or secretions from active lesions as the templates .
Twenty-five Sporothrix spp . isolates ( including 21 S . globosa , 3 S . schenckii s . str , and 1 S . brasiliensis ) , twenty-eight other fungal strains ( including agents of superficial , subcutaneous , and systemic mycoses in humans and animals ) , three bacterial strains , one human genomic DNA sample and one BALB/c-mouse genomic DNA sample were used to develop the PCR assays ( S1 Table ) . The fungi were obtained from the Collection of Pathogenic Fungi at the Research Centre for Medical Mycology , Peking University ( BMU , Beijing , China ) , the bacterial DNA were obtained from the National Institute for Communicable Disease Control and Prevention , Chinese Centre for Disease Control and Prevention ( Beijing , China ) , the human DNA was obtained from a healthy volunteer , and the mouse DNA was obtained from a BALB/c mouse . All of the fungal strains were previously characterized at the species level via morphological analysis and sequence analysis of the rDNA operon ( ITS1-5 . 8S-ITS2 ) and the CAL gene . A total of 40 tissue biopsies were collected between September 2017 and August 2018; the clinical data are shown in S2 Table . These samples were collected from patients in the Dermatology Department of the Second Hospital of Jilin University , whose clinical manifestations indicated suspected sporotrichosis . The clinical symptoms of the subjects were examined by professional physicians . In addition , 6 other negative human samples were collected from different volunteers and were used as artificially contaminated samples to simulate clinically infected specimens of S . schenckii s . str and S . brasiliensis . All of the specimens were skin and subcutaneous tissue harvested by surgery . When the specimens were collected , informed consent was obtained based on the guidelines and agreements of the institutional ethics committee . All of the fungal isolates were subcultured on 2% potato-dextrose agar ( PDA ) slide medium at 28°C for 7–14 days . All of the tissues were cut into small pieces with sterile scissors , and then all of the pieces were placed in liquid nitrogen and ground thoroughly with a mortar and pestle . DNA was extracted and purified with the QIAamp DNA Mini Kit ( QIAGEN , Hilden , Germany ) in accordance with the manufacturer’s instructions; detailed steps are provided in the supplemental methods ( S1 Text ) . The quality of the extracted DNA was assessed by amplification of part of the rDNA operon or β-globin using universal primers[34] . The amplified products were visualized via agarose gel electrophoresis and UV detection . A single amplification product indicated that the sample was free of PCR inhibitors . Three plasmids were constructed to create a standard curve and to determine the LODs of the multiplex real-time PCR assay . The CAL regions of S . globosa ( BMU 09028 ) , S . schenckii s . str ( CBS498 . 86T ) and S . brasiliensis ( CBS 120339T ) were cloned into pMD-18T vectors ( Takara , Dalian , China ) . The plasmids were transformed into E . coli DH5a competent cells , and the cells that contained recombinant plasmids were cultivated in lysogeny broth for 24 h . The plasmids were then extracted from the cultured E . coli suspension with a Qiagen Plasmid Mini Kit ( QIAGEN , Hilden , Germany ) . Plasmid concentrations were measured with a NanoDrop 2000 spectrophotometer ( Thermo Fisher Scientific , USA ) , and the copy numbers of the plasmids were calculated from their total base lengths and DNA concentrations using the equation of Godornes et al . [35] . The DNA samples and plasmids were stored at -20°C until use . All of the CAL sequences belonging to Sporothrix spp . were selected from the National Center for Biotechnology Information ( NCBI ) database to develop specific primers targeting the conserved sequence of Sporothrix spp . and probes marked by different fluorescent signals targeting the divergent sequences of S . globosa , S . schenckii s . str and S . brasiliensis ( details in Table 3 ) . Primer Express software ( version 3 . 0; Life Technologies-Applied Biosystems ) was used to design the primers and probes and to evaluate melting temperatures , GC content , dimers , and mismatches in the candidate primers and probes . Each PCR mixture consisted of 2 . 5 μL of 10x Platinum Buffer ( Life Technologies-Invitrogen ) , 4 . 0 μL of MgCl2 ( 50 mM ) , 0 . 2 μL of each primer ( 25 μM ) , 0 . 1 μL of each probe ( 25 μM ) , 0 . 25 μL of Platinum Taq DNA polymerase ( 5 U/μL; Life Technologies-Invitrogen ) , 1 . 5 μL of PCR nucleotide mix ( 10 mM ) , 5 μL of DNA template , and nuclease-free water to achieve a final volume of 25 μL . Multiplex real-time PCR was performed in a CFX96 Real-time PCR Detection System ( Bio-Rad , Hercules , CA , USA ) under the following conditions: predenaturation at 95°C for 3 min , followed by 45 cycles of 95°C for 15 s and 60°C for 30 s . The data were analysed with CFX Manager software ( version 3 . 1; Bio-Rad ) . The analytical specificity of the assays was tested by analysing 33 DNA samples derived from other pathogenic fungi , bacteria and tissues from a human and a mouse . The analytical sensitivity , standard curves and LODs of the assays were determined by using three 10-fold dilutions of the previously constructed plasmids , ranging from 2 . 0×105 copies/μL to 0 . 2 copies/μL . S . globosa , S . schenckii s . str and S . brasiliensis were detected by FAM , VIC , and CY5 fluorescence , respectively . The detection limit was noted for each probe . Each dilution of the plasmids was assayed in triplicate . The detection ability for a mixed template was determined by analysing the Ct values from 9 compositions of plasmid mixtures . The amount of detected plasmid was set at one gradient larger than LOD ( i . e . , 100 copies for S . globosa; 1000 copies for S . brasiliensis; 100 copies for S . schenckii s . str ) . The amount of the other two plasmids were 10-fold , 100-fold , and 1000-fold greater than that of the detected plasmid . Each template mixture was comprised of three plasmids of different proportions ( details in Table 1 ) . The mixed templates were detected by multiplex real-time PCR . The obtained Ct values were compared with those of single plasmid detection in the same reaction system under the same conditions . The detection ability of the multiplex and single fluorescence assays was tested by comparing the Ct values from four different reaction systems ( one multiplex and three single fluorescence values of FAM/VIC/CY5 ) with the same templates under the same conditions . Each reaction was assayed in triplicate . A total of 40 specimens from biopsies were collected , and each was divided into three parts . One part was used for culture ( PDA , 28°C ) , one for histopathological examination ( HE and PAS ) and one for DNA extraction . The performance of the multiplex real-time PCR assay was evaluated by comparison with the culture method and the species-specific PCR[29] . After 4 weeks of culture , no fungal growth or growth of contaminating microorganisms was judged as negative[18] . The results of the histopathological examination showed a mixed suppurative and granulomatous inflammatory reaction in the dermis and subcutaneous tissue , and the detection of asteroid bodies or Sporothrix spp . yeast cells by PAS was suggestive of sporotrichosis . The clinical diagnosis was made by combining the results of the culture and histopathological examination . To evaluate the assay detection ability for S . schenckii s . str and S . brasiliensis infectious samples , we used artificially contaminated ( spiked ) samples to simulate clinically infected specimens . Each of the 6 negative human samples was divided into three parts , two of which were used to simulate infected specimens and one of which was used for negative controls . S . schenckii s . str ( CBS498 . 86T ) and S . brasiliensis ( CBS 120339T ) were subcultured on brain heart infusion ( BHI ) agar medium and incubated at 35°C for 7 days to obtain the yeast cells of Sporothrix spp . The yeast cell suspensions of S . schenckii s . str and S . brasiliensis were prepared with sterile saline solution , and the OD was adjusted at 520 nm to 0 . 2 , approximately corresponding to a concentration of 106 cells/mL . [28] . Then , 10 , 50 , 100 μL each of the suspensions were mixed with 2 different negative human samples , respectively . DNA extraction procedures were the same as described above . A no-template control ( NTC ) , a negative control ( NEG ) and a positive control ( POS ) were established for each test of the multiplex real-time PCR assay . When the amplification result showed NTC ( - ) , NEG ( - ) , and POS ( + ) , the test was considered a valid amplification . A Ct value from the valid amplification of less than 40 was judged as positive; otherwise , it was negative . The species-specific PCR was performed according to the literature[29] , and a single clear band shown by gel electrophoresis and UV detection was judged as positive; otherwise , it was negative . The quantitative data are presented as the mean ± standard deviation ( SD ) . The differences in Ct values between multiplex and single fluorescence were tested with the independent samples t-test . The differences in positive detection rates between the multiplex real-time PCR and culture and between the multiplex real-time PCR and species-specific PCR were tested with the paired chi-square test . Statistical significance was defined as a p value <0 . 05 . All of the calculations were performed with the Statistical Analysis System software package ( version 9 . 3; Cary , NC , USA ) . This study was performed in strict accordance with recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health . The protocol was approved by the Ethics in Research Committee of the Second Hospital of Jilin University , protocol number 2018–018 . Written informed consent was obtained from patients or at least one guardian of the patient before enrolment .
|
Sporotrichosis is a subacute or chronic infectious disease caused by dimorphic fungi of Sporothrix spp . The genus Sporothrix consists of several species with different geographic distributions , virulence , and antifungal susceptibilities , making species-level identification necessary . S . brasiliensis , S . globosa , S . schenckii s . str and S . luriei make up the “pathogenic clade” of the genus Sporothrix . Importantly , S . luriei has a low clinical-epidemiological impact within this genus . Therefore , we designed a novel multiplex real-time PCR method using fluorescent probes for the identification of S . globosa , S . schenckii s . str , and S . brasiliensis . We designed a pair of primers based on the conserved sequence of the calmodulin gene of Sporothrix spp . and probes with different fluorescent signals based on the divergent sequences of S . globosa , S . schenckii s . str and S . brasiliensis . Through the verification of nucleic acid , clinical and spiked sample detection , the multiplex real-time PCR could quickly and accurately identify the three clinically relevant species of Sporothrix spp . with high sensitivity . This new assay could be applied in epidemiology , clinical diagnosis , and experiments with sporotrichosis to control new outbreaks , reduce diagnostic and identification time , and improve test efficiency .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"medicine",
"and",
"health",
"sciences",
"pathology",
"and",
"laboratory",
"medicine",
"pathogens",
"microbiology",
"plasmid",
"construction",
"organisms",
"fungi",
"sporotrichosis",
"dna",
"construction",
"molecular",
"biology",
"techniques",
"fungal",
"diseases",
"sporothrix",
"schenckii",
"fungal",
"pathogens",
"extraction",
"techniques",
"research",
"and",
"analysis",
"methods",
"infectious",
"diseases",
"mycology",
"artificial",
"gene",
"amplification",
"and",
"extension",
"medical",
"microbiology",
"microbial",
"pathogens",
"molecular",
"biology",
"dna",
"extraction",
"eukaryota",
"diagnostic",
"medicine",
"biology",
"and",
"life",
"sciences",
"polymerase",
"chain",
"reaction"
] |
2019
|
Fast diagnosis of sporotrichosis caused by Sporothrix globosa, Sporothrix schenckii, and Sporothrix brasiliensis based on multiplex real-time PCR
|
Maintaining levels of calcium in the cytosol is important for many cellular events , including cell migration , where localized regions of high calcium are required to regulate cytoskeletal dynamics , contractility , and adhesion . Studies show inositol-trisphosphate receptors ( IP3R ) and ryanodine receptors ( RyR ) , which release calcium into the cytosol , are important regulators of cell migration . Similarly , proteins that return calcium to secretory stores are likely to be important for cell migration . The secretory protein calcium ATPase ( SPCA ) is a Golgi-localized protein that transports calcium from the cytosol into secretory stores . SPCA has established roles in protein processing , metal homeostasis , and inositol-trisphosphate signaling . Defects in the human SPCA1/ATP2C1 gene cause Hailey-Hailey disease ( MIM# 169600 ) , a genodermatosis characterized by cutaneous blisters and fissures as well as keratinocyte cell adhesion defects . We have determined that PMR-1 , the Caenorhabditis elegans ortholog of SPCA1 , plays an essential role in embryogenesis . Pmr-1 strains isolated from genetic screens show terminal phenotypes , such as ventral and anterior enclosure failures , body morphogenesis defects , and an unattached pharynx , which are caused by earlier defects during gastrulation . In Pmr-1 embryos , migration rates are significantly reduced for cells moving along the embryo surface , such as ventral neuroblasts , C-derived , and anterior-most blastomeres . Gene interaction experiments show changing the activity of itr-1/IP3R and unc-68/RyR modulates levels of embryonic lethality in Pmr-1 strains , indicating pmr-1 acts with these calcium channels to regulate cell migration . This analysis reveals novel genes involved in C . elegans cell migration , as well as a new role in cell migration for the highly conserved SPCA gene family .
Calcium is one of the most versatile and important small molecules in the cell: it is involved in cell processes starting at the beginning of new cell life during fertilization and ending with a role in cell death during apoptosis [1] . One developmental role for calcium is in cell migration . Studies have determined that migrating cells establish a relatively stable calcium gradient , with higher levels in the rear of the cell [2] . The increased calcium levels at the rear of the cell are likely important for regulating cell detachment and contraction [3] , [4] . More recently , studies identified transient flickers of high calcium enriched at the leading edge of migrating cells and preceding directional changes . These calcium flickers may provide localized cues that help direct migration [4]–[6] . Calcium-sensitive molecules important for cell migration can be divided into two broad categories: those that sense and respond to differences in calcium levels and those that control levels of calcium by transporting it across membranes [1] . In this second class , channels that release calcium into the cytosol , such as the plasma-membrane-associated transient receptor potential ( TRP ) channels , partnering with the secretory pathway inositol-trisphosphate channels ( IP3R ) and ryanodine receptors ( RyR ) , have been shown to play a key role in producing calcium flickers and in cell migration [1] , [4] , [7] , [8] . The localized movement of calcium out of the cytosol must also be important for producing calcium flickers and gradients . Calcium uptake from the cytosol requires Ca2+ ATPase transporters found at three distinct subcellular locations: the plasma membrane ( PMCA ) , the Sarco ( endo ) plasmic reticulum ( SERCA ) and the secretory pathway/Golgi ( SPCA ) [1] , [9] , [10] . SERCA , which localizes to both the endoplasmic reticulum ( ER ) and Golgi , helps to control cytosolic calcium levels in a variety of cells and is integral for proper IP3 signaling [1] . The Golgi-localized SPCA , which has not been as extensively characterized , transports both Ca2+ and Mn2+ , playing an important role in homeostasis for both molecules . SPCA is also critical for protein processing in the Golgi , Ca2+ release in response to IP3 signaling , and stress tolerance [1] , [9] , [11] . While at least one of these Ca2+ ATPase proteins is likely to be involved in maintaining subcellular calcium gradients important for cell migration , this link has yet to be firmly established . Genetic studies show the highly conserved SPCA1 protein has essential roles in embryonic development and disease . Loss of a single copy of the SPCA1 gene in humans causes Hailey-Hailey disease ( MIM# 169600 ) , a skin disorder characterized by recurrent lesions or blisters of the skin in areas subject to high stress [12]–[16] . SPCA1 is likely essential in humans , as more severe phenotypes are found in patients who suffer clonal loss of both copies of the gene [17] , [18] . Mice embryos homozygous for null mutations in SPCA1 die with defects in neural tube closure , while heterozygotes show susceptibility to squamous cell tumors , a phenotype observed occasionally in humans with Hailey-Hailey [19]–[22] . The primary cellular defect in Hailey-Hailey disease patients is keratinocyte acantholysis or loss of cell adhesion , marked by aggregation of keratin intermediate filaments that have retracted from desmosomes [23] , [24] . In mice , cell adhesion appears normal , but cells show signs of Golgi dysfunction , which likely induces the high levels of apoptosis observed in the neural tube and mesenchyme [19] . Based on this genetic analysis , it is not evident that SPCA , which has a conserved role at the molecular level , has a similarly conserved role in development , although each phenotype may be the result of secretory pathway stress [25] . Caenorhabditis elegans has served as a good model system for studying calcium signaling because it has the full complement of membrane-associated calcium channels . Research has revealed a role for the calcium channel ITR-1/IP3R in ovulation , gastrulation , pharyngeal pumping , defecation and other developmental processes [26] . It is expressed in the embryo and is the only channel shown to play a direct role in cell migration , as itr-1 mutants fail to complete ventral closure during C . elegans development [26]–[28] . The UNC-68/RyR localizes to both ER and Golgi and plays a key role in muscle function [7] , [29]–[32] . The single SPCA1 gene , pmr-1 , encodes a protein that , like its vertebrate counterpart , co-localizes with Golgi markers , when expressed in COS cells or in C . elegans [33]–[35] . Based on RNA in situ hybridization and microarray analysis , pmr-1 is expressed in the adult germ line and at moderate levels throughout embryogenesis in most blastomeres [36]–[38] . pmr-1 transcriptional and translational fusion constructs show expression starting from the 3-fold embryo and continuing until the adult stage in the nervous system , intestine , intestinal valve , hypodermis , spermatheca , and gonad [35] , [39] . The absence of germ line and early embryo expression for these fusion constructs may indicate they are silenced [40] or contain incomplete regulatory sequences . PMR-1 protein also shows functional conservation , with important roles in Ca2+ spiking in response to IP3 signaling , stress response , thermotolerance , pathogen resistance , and metal homeostasis [26] , [33]–[35] , [41] , [42] . Using alleles of pmr-1 obtained in forward genetic and deletion screens , we have identified a new , essential role for pmr-1 in embryonic development in C . elegans . Strains homozygous for mutations in pmr-1 show temperature-sensitive embryonic lethality caused by defects in ventral enclosure , apical enclosure , and morphogenesis . By identifying the temperature-sensitive period we have learned that pmr-1 plays an important role during the period preceding ventral enclosure . While pmr-1 mutants show normal cell fate specification , lineage , and cell ingressions during gastrulation , we see defects in the migration of cells along the surface of embryo following ingression in the C-lineage , ventral neuroblasts , and in anterior lineages , a cell migration process that appears similar to migrations during neurulation in vertebrates . Phenotypes in pmr-1 mutants can be suppressed or enhanced by changes in the activity of the RyR unc-68 and IP3R itr-1 genes , indicating that these cell migrations are sensitive to calcium levels . This study identifies new molecules important for embryonic cell migration in C . elegans and reveals a new role in cell migration for the calcium channel PMR-1/SPCA protein family .
Using two independent forward genetic screens designed to identify conditional mutants essential for embryonic enclosure and morphogenesis , we obtained two alleles of the C . elegans SPCA1 homolog pmr-1 , jc10 and ru5 . Strains homozygous for either the jc10 or ru5 allele produce inviable progeny when grown at 25°C , although these strains show increased embryonic viability at lower temperatures ( Table 1 ) . The tm1840 and tm1750 alleles [43] show similar temperature sensitivity , although overall embryonic viability is significantly lower at both 15°C and 20°C . For all four alleles examined , Pmr-1 embryos that die during mid-embryogenesis display a range of terminal phenotypes , including ventral closure defects , head ruptures , morphogenesis defects , and pharynx unattached ( Pun ) phenotypes ( Figure 1; Table 2 ) . We established that the embryonic lethal phenotypes observed in jc10 and ru5 strains are due to mutations in the pmr-1 gene by a series of criteria . Using conventional and snp-snp mapping strategies , the jc10 and ru5 alleles were mapped to a small interval on L . G . I ( Figure S1A ) . Whole genome and conventional sequencing revealed that pmr-1 was the only gene in the mapped region that carried a mutation in both the jc10 and ru5 strains . Complementation analysis showed that jc10 and ru5 failed to complement each other and also failed to complement the pmr-1 deletions tm1750 and tm1840 , consistent with the interpretation that all four mutations are allelic ( Table 1; data not shown; Figure S1B ) . Transformation of ru5 strains with fosmids that contain the pmr-1 gene ( Figure S1B , S1C ) rescued the high temperature embryonic lethality . In contrast , we did not observe rescue in strains transformed with any other fosmids or cosmids from the mapped region ( Figure S1B , S1C ) . Taken together , these experiments confirm that the defects in embryogenesis that we observe are due to disruption of pmr-1 gene activity . The pmr-1 ( tm1840 ) deletion allele most likely represents a null , as the deletion eliminates an early exon , found in all of the major identified isoforms ( Figure S1B ) [33] . Based on strong sequence homology between PMR-1 and SERCA , whose crystal structure has been solved [44] , the deletion would eliminate a transmembrane domain and a portion of the E1/E2 ATPase domain , thus rendering the expressed protein non-functional . However , we were unable to confirm whether the pmr-1 ( tm1840 ) allele is a null . Genetic tests with the dxDf2 deficiency strain were uninformative due to strong dominant phenotypes in the parental strain . Injection of dsRNA did not enhance or diminish the defects in the pmr-1 ( tm1840 ) strain ( Table 1 ) , which is consistent with expectations for a null mutation . However , injection of dsRNA into controls did not phenocopy the pmr-1 ( tm1840 ) allele , a result consistent with those reported by others ( Table 1 ) ; [35] , [45] . Alternative experiments to test for the null , such as use of the antisera generated against PMR-1 [34] , were also ambiguous because of background activity of the antisera ( data not shown ) . While the tm1840 allele has the highest mortality at low temperatures , other alleles of pmr-1 show a range of phenotype severity . The pmr-1 ( tm1750 ) allele contains a deletion of coding sequence from just one isoform , but the severity of the phenotype , which is similar to pmr-1 ( tm1840 ) , suggests either the isoform is critical or other isoforms are also disrupted ( Table 1 and Table 2 ) . The pmr-1 ( jc10 ) allele has a premature stop in an exon found in all isoforms ( exon 4 of the pmr-1b isoform ) that would eliminate transmembrane helices and the E1/E2 ATPase domain ( Figure S1 ) . The pmr-1 ( ru5 ) allele is a G142D substitution in a residue conserved from yeast to humans in PMR-1 , but also conserved in SERCA ( Figure S1 ) [46] . Based on homology with SERCA , this residue , in a beta strand that forms a portion of the A domain , would likely result in localized disruption of this domain , interfering with key interactions between the A and N domains of the E1/E2 ATPase [44] . Because all four alleles show the same embryonic lethal defects , varying only in the frequency of inviable embryos observed at a given temperature , the simplest interpretation is that they represent reduced or complete loss of function of pmr-1 gene activity , and are indicative of a key role for pmr-1 in embryogenesis . In humans , Hailey-Hailey disease has a dominant inheritance pattern , likely due to haploinsufficiency of the pmr-1 homolog SPCA1 [10] . In C . elegans , analysis of dominance is complicated because we observe both maternal and zygotic rescue for pmr-1 alleles . pmr-1 ( ru5 ) hermaphrodites crossed to control males produce live progeny at 25°C , showing zygotic activity is sufficient to rescue the mutant phenotype . F1 hermaphrodites from a cross between pmr-1 and control strains produce between 96%–100% viable F2 embryos at 25°C , depending on the allele ( Table 1 ) . Since 25% of these F2 progeny should be homozygous for the pmr-1 loss-of-function alleles , these data suggest the maternally provided gene product is sufficient to rescue most of these embryos . However , for pmr-1 ( ru5 ) , we were able to determine that the inheritance pattern is weakly semi-dominant . Of F2 progeny from a cross between pmr-1 ( ru5 ) and controls , 39% show a later developmental phenotype at 25°C , including larval lethality ( 3% ) , sterility ( 13% ) , small broods ( 3% ) , and inviable F3 progeny ( 20%; n = 236 ) . These data show that some individuals heterozygous for pmr-1 loss-of-function alleles exhibit a phenotype . Although the literature reports that pmr-1 is expressed in most blastomeres throughout embryonic development [36]–[38] , these genetic data suggest the wild type pmr-1 gene product is essential during a developmental period when both maternal and zygotic gene products are active . During C . elegans embryonic development the epidermal ( hypodermal ) cells that are born on the dorsal surface migrate to the ventral midline , enclosing the body of the embryo . The anterior-most epidermal cells also migrate to enclose the head , ultimately connecting epidermal tissue with the anterior foregut , creating the buccal cavity ( Figure 1A ) [47]–[49] . pmr-1 mutant embryos die with variable terminal phenotypes that include failure to enclose the body ( Figure 1B , 1C ) and head Figure 1D , 1E ) , as well as a detached pharynx ( Pun ) ( Figure 1F , 1G ) . pmr-1 embryos that elongate also show defects in epidermal cell organization ( 1H , I ) . Additional phenotypes include reduced brood size and some larval lethality ( Table 1 ) . The terminal phenotypes due to pmr-1 disruption , such as ventral enclosure defects and head enclosure failures , are at least superficially similar to the neural tube closure failures in mice lacking the PMR-1 homolog SPCA , as well as the cell adhesion defects observed in Hailey-Hailey disease patients [19] , [23] , [24] . When pmr-1 is disrupted , the frequency of embryonic lethality varies depending on both allele and temperature . Given that all four alleles carry different types of genetic lesions , from point mutation to deletion , it is unlikely that they are temperature sensitive , per se , but rather there is a developmental process in which normal pmr-1 activity is increasingly important as temperature is increased . All alleles are also pleiotropic , showing a range of embryonic lethal phenotypes ( Table 2 ) . The frequency with which a specific terminal phenotype is observed is similar for all alleles , at all temperatures , and independent of whether the temperature shifts occur early or close to the temperature-sensitive period in development , as described below ( Table 2 ) . These data are consistent with the hypothesis that pmr-1 is playing a role in developmental events that affect cells positioned throughout the embryo , resulting in the variable effects . To better understand the defects leading to the Pmr-1 terminal phenotypes , we observed embryos during gastrulation using Nomarski optics . Using this approach , the first observable differences between control and pmr-1 ( ru5 ) embryos are in the position of the C blastomeres . In control embryos , the C-lineage blastomeres are positioned along the ventral surface at the posterior end of the embryo just prior to ingression ( Figure 2A ) . In pmr-1 ( ru5 ) mutant embryos at the same developmental stage , these cells occupy a more dorsal position ( Figure 2E ) . At the end of gastrulation , the gastrulation cleft closes in control embryos ( Figure 2B ) . However , in pmr-1 ( ru5 ) embryos , the cleft remains open ( Figure 2F ) , reflecting mispositioned ventral neuroblasts . We also saw later differences in cell position , with anterior blastomeres properly positioned in controls ( Figure 2C ) , but altered in pmr-1 ( ru5 ) embryos ( Figure 2G ) . Finally , we observed that the basement membrane , which initially surrounds the pharynx but is lost along the anterior most cells of the pharynx in control embryos ( Figure 2D ) [49] , remains in place in pmr-1 ( ru5 ) embryos ( Figure 2H ) . This analysis suggests the terminal phenotypes we observe are actually caused by problems in cell positioning earlier in embryogenesis . To better pinpoint the developmental time point when pmr-1 activity is required , we took advantage of the temperature sensitive phenotypes of pmr-1 alleles . Using the pmr-1 ( ru5 ) allele , which has viable progeny at 15°C but not at 25°C and is less fragile than the pmr-1 deletion alleles , we performed temperature shift experiments . In reciprocal temperature shifts , we found that embryos grown at the restrictive temperature 2 to 3 hours after the two-cell stage are not viable , while those grown at the permissive temperature during this time period survive ( Figure 3 ) . Temperature pulse experiments show a one-hour pulse from the restrictive to permissive temperature can significantly rescue viability only if the pulse occurs between 2 and 3 hours after two-cell stage , consistent with the hypothesis that pmr-1 is required only during this specific time period ( ) . In contrast , one-hour pulses from 15°C to 25°C are insufficient to cause high lethality ( ) . The temperature-sensitive period for pmr-1 is during gastrulation , during a period when anterior , c-lineage , and ventral blastomeres migrate ( Figure 3; Figure S2 ) , and significantly earlier than the events directly associated with the enclosure failure , Pun , and morphogenetic terminal phenotypes . Given that the temperature-sensitive period is during a stage of development when cells are rapidly dividing and differentiating , we wanted to determine if the defects observed in pmr-1 embryos were due to changes in cell lineage . In C . elegans , the lineage pattern is essentially invariant in control embryos [47] . We took advantage of this to compare the lineage of pmr-1 ( ru5 ) embryos to controls , from the 2-cell to ∼300-cell stage , using StarryNite software [50]–[53] . Our analysis of pmr-1 ( ru5 ) embryos indicates that the cell division timing and pattern for all lineages is indistinguishable from controls ( Figure 4 ) . We also found that the correct number of cells underwent programmed cell death with normal timing , beginning one to two divisions after the temperature-sensitive period ( data not shown ) . These data indicate that pmr-1 embryos are not dying because of premature cell death or general defects in cell division timing or patterns . Although lineages appeared normal in pmr-1 ( ru5 ) embryos , we wanted to ascertain whether the cell fate was also normal . To address this question , we examined the expression patterns of GFP reporter fusion constructs that are expressed with specific timing , in a position-dependent manner within the embryo , or in specific tissues . Reporters included those expressed in the nervous system , hypodermis , muscle , intestine , and pharynx , some of which are derived from multiple lineages . We also included reporters , such as ceh-13 , ceh-16 and vab-7 , which are expressed in a position-dependent manner in multiple tissues and lineages [54]–[57] . In comparisons between control and pmr-1 ( ru5 ) embryos , the expression patterns of all of these reporter constructs were identical with only a few exceptions ( Table 3; Figure S3 , S4 ) . These exceptions included differences in the position , but not in the number of cells expressing epidermal , neuroblast , and muscle markers . The other exception involved the expression of vab-7/even-skipped [54] , which showed modest , but statistically significant delays in gene expression in C-lineage muscle and epidermal cells ( Table 3; Figure S3 ) . While we cannot rule out subtle effects on cell fate or changes in the fate of a small subset of cells , this analysis indicates that all major lineages and tissues are differentiating properly in pmr-1 ( ru5 ) embryos grown under restrictive conditions . The observed differences in the position and timing of the vab-7 reporter , as well as the Nomarski imaging analysis ( Figure 3; Table 3 ) , suggested that pmr-1 mutants may have defects in cell migration . Since cells that express vab-7 are derived from the C lineage [54] , we looked more closely at cell migration in C-derived blastomeres . In control embryos , the C-lineage blastomeres migrate from a dorsal position to a more ventral one along the posterior surface of the embryo . Muscle precursors , which have a more ventral starting position and migrate ahead of the hypodermal precursors , eventually ingress into the embryo at a posterior/ventral position ( Figure 5 ) [58] , [59] . When we compared control and pmr-1 ( ru5 ) embryos , we saw no differences in the timing of C-blastomere ingressions . Similarly , the timing of E , P4 , MS , and D-derived blastomere ingressions were similar in control and pmr-1 ( ru5ts ) embryos ( n = 6 for each strain; t-test; p>0 . 05 ) . This analysis shows that pmr-1 does not appear to be playing an essential role in cell ingression during C . elegans gastrulation . In contrast , pmr-1 does have a critical role in the migration of cells along the surface of the embryo . Using 4-dimensional representations of embryos created with Acetree , we measured the migration rates of C blastomeres along the posterior surface of the embryo ( Materials and methods ) . In pmr-1 ( ru5 ) embryos , the rates of migration of both muscle and hypodermal cell precursors were significantly slower , at 55% and 40% of controls , respectively ( Figure 5; t-test; p<0 . 05 ) . Given that these cell migrations occur during the temperature sensitive period for pmr-1 gene activity , this analysis is consistent with the interpretation that pmr-1 plays an important role in cell migration during C . elegans embryonic development . Given the variability of the pmr-1 terminal phenotypes , we examined several other lineages to determine if other cells exhibited cell migration defects . Terminal phenotype and gene expression analysis suggested that many cells are properly positioned in the embryo , so we focused our attention on cells that migrate during the temperature-sensitive period and whose migration defects might lead to enclosure failures and Pun phenotypes . In control embryos , the ventral neuroblasts derived from ABprp and ABplp lineages migrate from lateral to central positions along the ventral surface of the embryo , closing the gastrulation cleft ( Figure 6A; [47] , [58] . Hypodermal cells then crawl over these cells during enclosure [47] , [48] . Analysis of pmr-1 ( ru5 ) embryos showed that these ventral cells are properly positioned at the beginning of the temperature-sensitive period , but many move a shorter distance than controls ( Figure 6A ) , in some cases failing to close the gastrulation cleft , leading to later enclosure failures . The migration distances of lineages examined varied widely , and some cells migrated distances similar to controls . However , the majority of cells examined showed significant reduction in cell migration , on average 44% of controls in pmr-1 ( ru5 ) embryos ( Figure 6B; T-test; p<0 . 05 ) . These results show that pmr-1 is playing a key role in cell migration in cells that give rise to ventral neuroblasts , as well as epithelial cells . In the anterior of control embryos , cells derived from the ABala and ABalp lineages , many of which ultimately differentiate into epithelial or nervous system cells , undergo dynamic migration patterns , with individual blastomeres crossing the dorsal-ventral or left-right midlines during embryogenesis . Some anterior blastomeres also make specific turns and ingress to form the anterior pharynx ( Figure 7; [47] , [58] , [60] . In pmr-1 ( ru5 ) embryos , many anterior cells show migration defects . While some cells migrated normally in a subset of the pmr-1 ( ru5 ) embryos examined , many cells migrated shorter distances , in the wrong direction , or failed to make key turns , resulting in cells that were improperly positioned ( Figure 7; t-test; p<0 . 05 ) . Since these cells give rise to arcade cells , hypodermis , and ring ganglia , their failure to migrate properly during the temperature-sensitive period can readily account for the terminal phenotypes we observe in pmr-1 ( ru5 ) embryos . Taken together , these results show that pmr-1 plays a key role in cell migration and positioning during gastrulation . In Hailey-Hailey disease , cells show defects in cell attachment structures [23] , [24] . The defects in cell migration we observe in pmr-1 mutants could be due to changes in cell adhesion , as well as in other cellular processes including cell polarity or establishment of the basement membrane . Our analysis of gene expression suggests that each of these processes is relatively normal in pmr-1 ( ru5ts ) embryos . PAR-6 is an apically localized protein that plays a key role in cell polarity in a number of developmental processes [61] . SMA-1 is an actin-binding protein that localizes to the apical membrane of polarized epithelial cells [62] . NID-1 is a basement membrane protein expressed during embryonic development and nid-1 ( cg119 ) mutants have Pun phenotypes [63] , [64] . AJM-1 , MEL-11 , and VAB-9 localize to apical adhesion complexes [65]–[68] . Each of these proteins is expressed in the correct cells and shows normal subcellular localization in pmr-1 ( ru5ts ) embryos ( Table 3; Figure S4 ) . The localization is normal even in hypodermal cells that show positioning defects ( Figure S4M-S4R ) . Similarly , KAL-1 and PLX-2 , proteins that are expressed in ventral neuroblasts and play key roles in hypodermal cell migration and enclosure [69] , [70] , appear to be properly expressed in pmr-1 ( ru5 ) embryos ( Figure S3I–S3L ) . While we cannot rule out subtle changes in protein localization and gene expression , these data indicate that the defects we observe in pmr-1 mutant embryos are not due to general defects in cell polarity , cell adhesion , or expression of basement membrane proteins . Given the conserved molecular roles of PMR-1/SPCA , we asked whether perturbations in Ca2+ signaling were chiefly responsible for the pmr-1 mutant phenotypes . To address this question , we altered the activity of either the ITR-1/IP3R or UNC-68/RyR calcium channels , reasoning that if calcium levels were critical for normal cell migration , changes in the activities of these receptors would suppress or enhance the pmr-1 embryonic lethal phenotypes . We made strains carrying pmr-1 mutations and itr-1 ( sy327 ) , a gain-of-function allele that is thought to increase the affinity for IP3 [28] , [71] . At 20C , the itr-1 ( sy327 ) strain has slightly reduced embryonic viability ( 91% ) compared to controls while the pmr-1 ( tm1840 ) strain has 9% embryonic viability . The pmr-1 ( tm1840 ) ;itr-1 ( sy327 ) double mutant shows significant improvements in viability compared to the pmr-1 ( tm1840 ) single mutant ( 42%; Table 4 ) . This affect is not allele specific , as we observed similar suppression of phenotypes for the pmr-1 ( jc10 ) , pmr-1 ( ru5 ) , and pmr-1 ( tm1750 ) lines when placed in an itr-1 ( sy327 ) background ( Table 4 ) . These data show itr-1 phenotypes are epistatic to pmr-1 . We also made double mutants carrying pmr-1 alleles and itr-1 ( jc5 ) , a loss-of-function allele [27] . While we were unable to generate double mutants with pmr-1 ( tm1840 ) or pmr-1 ( tm1750 ) , suggesting the combination is lethal ( data not shown , V . P . , R . W . and R . M . ) , the pmr-1 ( ru5 ) ; itr-1 ( jc5 ) strain showed enhanced lethality compared to the pmr-1 ( ru5 ) strain alone ( Table 4 ) , suggesting the two genes act in parallel pathways during this stage of development . In contrast , depletion of unc-68 using RNAi significantly suppressed the embryonic lethality of pmr-1 ( ru5 ) strains ( Table 4 ) , indicating unc-68 acts in opposition to pmr-1 in a role very different from that of itr-1 . Taken together , these data show that the embryonic lethality we observe in pmr-1 mutant embryos can be affected by changes in the activity of calcium channels . While this type of analysis does not allow us to rule out other possibilities , it provides strong support for the hypothesis that pmr-1 is acting through its role in Ca2+ signaling to affect cell migration .
Our analysis of PMR-1/SPCA1 in C . elegans has revealed a novel and essential role in cell migration for this highly conserved gene [1] , [9] . Using a forward genetic screen designed to look for genes required for normal epidermal cell migration and adhesion , we identified mutations in the C . elegans SPCA1 gene pmr-1 . Strains homozygous for pmr-1 show temperature-sensitive embryonic lethality , with terminal phenotypes that include ventral enclosure failures , head ruptures , and morphogenetic phenotypes including a detached pharynx ( Pun ) phenotype . Phenotype and temperature shift analysis indicates that the pmr-1 gene plays a crucial role in a specific period of embryonic development , during gastrulation . Cell fate specification , cell division patterns , apoptosis , and cell ingression during gastrulation are all similar to control embryos . However , the migrations of specific subsets of cells , along the posterior , anterior , and ventral surfaces of the embryo , are defective in pmr-1 mutant embryos . The specificity of the phenotype , coupled with the temperature shift analysis , indicate that PMR-1/SPCA plays an essential role in a specific set of cell migrations during gastrulation in the developing C . elegans embryo . Given the broad expression pattern for pmr-1 in all cells , coupled with the proposed housekeeping functions one would expect for a calcium transporter [1] , [9] , the specific timing and proposed role for PMR-1 in ectodermal cell migration is unexpected . Studies examining the expression of pmr-1 indicate that PMR-1 protein localizes to the Golgi and that the gene product is present in the germ line and widely expressed in cells throughout embryogenesis , including during gastrulation [33]–[38] . This expression pattern , which is consistent with our genetic analysis that maternally or zygotically provided gene product is sufficient to rescue the Pmr-1 mutant phenotype , would suggest a broad role for the gene product . Yet our temperature shift and pulse analysis indicates requirements for PMR-1 can be bypassed during other stages of embryogenesis . The phenotype analysis identifies a very specific role in the migration of anterior , C-lineage , and ventral cells along the surface of the embryo during gastrulation , but with no discernable effects on other cell migration events such as ingression or ventral closure , nor on any other developmental pathway that might alter cell fate , division patterns , or cell polarity . The best explanation for these data is that PMR-1 may act redundantly in some aspects of embryogenesis , but plays a more critical or specific role in the cell migrations along the embryo surface that follow cell ingression , a markedly different developmental process [58] , [60] . PMR-1/SPCA is part of a network of proteins that regulate calcium levels in the cell cytoplasm and calcium stores in the secretory pathway . Models for modulating cellular calcium levels start with a signal transduced through PLC and IP3 triggering release of Ca2+ into the cytoplasm from secretory stores through ITR-1/IP3 and UNC-68/RyR . The calcium ATPases SCA-1/SERCA and PMR-1/SPCA , located predominantly in the ER and Golgi , respectively , then pump Ca2+ back out of the cytoplasm into these organelles [7] , [26] . A simple model would predict that reduced activity of a transporter acting in one direction is suppressed by reduced activity of channel acting in the opposite direction . This is the result we observe for interactions between pmr-1 and unc-68 , where pmr-1 ( ru5 ) strains showed reduced lethality in an unc-68 ( RNAi ) background ( Table 4 ) . In contrast , strains carrying pmr-1 and itr-1 ( jc5 ) loss-of-function alleles showed enhanced lethality at semi-permissive conditions while strains carrying the itr-1 ( sy327 ) gain-of-function allele showed significant suppression of pmr-1 ( ru5 ) embryonic lethality . This gene interaction analysis suggests that itr-1 and pmr-1 act in parallel pathways during embryonic development and that enhanced sensitivity of itr-1 to IP3 signaling in the itr-1 ( sy32y7 ) mutant [72] can help bypass the requirement for pmr-1 . While these interpretations reflect the genetic analysis , they are puzzling because ITR-1 and PMR-1 act in opposite directions with respect to cytosolic calcium . Similarly , UNC-68 does not appear to have any developmental roles nor does it appear to be expressed at high levels in the embryo until later stages [7] , [29]–[32] . A speculative model to explain the genetic interaction data we observe is that these proteins may play some redundant and some non-redundant roles in regulating calcium levels in the cell , due in part to the polarity of the secretory pathway . One would predict that when PMR-1 activity is reduced , the pool of calcium in secretory stores is also modestly reduced and the calcium levels in the cytoplasm might be elevated [34] . Reduction of UNC-68 or ITR-1 calcium release would , in theory , restore overall secretory pool levels . However , the calcium flicker model for cell migration predicts that calcium must be supplied in localized bursts at the leading edge of the cell [5] , [6] . In a pmr-1 mutant , the pool of available calcium would be disproportionately reduced in the Golgi , the secretory organelle closest to the leading edge . If release is , as predicted , strongly dependent on ITR-1 [5] , [6] , an itr-1 loss-of-function mutant would further diminish calcium release , making the phenotype worse , while the itr-1 ( sy327 ) gain-of-function mutation would release calcium at lower levels of signal [28] , [71] , suppressing the phenotype . A second possibility to explain the suppressor data reflects the distinct mechanisms by which the channels are activated . While both the UNC-68/RYR and ITR-1/IP3R channels are sensitive to changes in cytosolic calcium , only the ITR-1/IP3R channels responds to an IP3 ligand [73] , [74] . Reduced activity of the ITR-1/IP3R channel would make the secretory pathway insensitive to IP3 signaling , which when coupled with reduced stores would make the phenotype worse . Increasing IP3R sensitivity might compensate for altered cytosolic calcium , making the situation better , as observed . Finally , since the regulation of calcium signaling and transport is highly interdependent , with channels and targets acutely sensitive to minor changes in calcium concentration [1] , [72]–[74] , single loss-of-function mutations may result in some compensatory changes in the signaling system that cannot occur if two systems are defective . While it is tempting to assume that disruption of pmr-1 has a direct effect on calcium-mediated cell signaling , disruptions in the gene might instead effect other calcium-dependent processes required for cell migration . The embryonic lethal phenotypes in Pmr-1 strains were observed at all the temperatures assayed , but they were more severe at high temperatures for all four alleles tested . Although our experiments were in the normal temperature range for C . elegans growth [75] and we saw no evidence of necrosis in pmr-1 mutant embryos , recent research indicates that pmr-1 is important for survival of heat shock , acting to prevent necrosis [42] . It is possible that the migrating cells of the embryo are particularly sensitive to changes in the activity of this stress resistance pathway . Alternatively , the cytoskeletal and membrane dynamics of migrating cells may simply be more sensitive to temperature than other developmental processes , or the timing is more precise , such that changes in pmr-1 activity have a stronger impact . Another possibility is that the changes in pmr-1 activity alter calcium-dependent processing of specific proteins in the Golgi . No matter the precise mechanism by which changes in calcium dynamics alter cell migration , our identification of a role for PMR-1 , acting with ITR-1/IP3R and UNC-68/RyR , confirms earlier work that ITR-1 plays an important role in cell migration [27] , [28] and identifies new genes important for normal cell migration in the C . elegans embryo . The phenotypes we observe in C . elegans pmr-1 mutants have at least some superficial similarities to those observed in vertebrates with defects in the SPCA1 gene , including humans with Hailey-Hailey disease . In both C . elegans and in vertebrates , we observe semi-dominant phenotypes and the severity of the observed phenotypes are influenced by environmental stresses . The terminal phenotypes , including apparent loss of cell adhesion in the epidermis of humans with Hailey-Hailey or failures to close the rostral neural tube in mice embryos homozygous for SPCA loss-of-function alleles [12]–[16] , [19] , [23] , [24] also look similar to the enclosure failures and head ruptures we observe in C . elegans embryos . One major difference between the worm pmr-1 mutant phenotype and Hailey-Hailey disease patients is the nature of defects in cell adhesion . In Hailey-Hailey disease , symptoms include acantholysis of keratinocytes , marked by keratin dissociation from desmosomes [23] , [24] . Our analysis of pmr-1 mutants suggests that the gene does not play a direct role in cell adhesion . Proteins that localize to adhesion complexes , such as MEL-11 , AJM-1 , and VAB-9 , look normal in pmr-1 mutant embryos and epistasis analysis indicates that vab-9 and pmr-1 are in different pathways ( unpublished data , J . Simske ) . While intermediate filament and desmosomal structures have not yet been carefully examined in pmr-1 mutants , we do not suspect an association . While pmr-1 is essential during gastrulation , analysis by others indicates genes that encode intermediate filament and attachment structures in C . elegans , such as IFA-1 , IFA-2/MUA-6 , IFA-3 , IFA-4 , IFB-1 , VAB-10 , MUP-4 , and LET-805 , assemble later in development , playing key roles in the embryonic elongation that follows ventral closure [76]–[83] . This analysis strongly suggests intermediate filaments are unlikely to be directly involved in the PMR-1-mediated cell migration , although it is possible PMR-1 may play a later , non-essential role in maintaining these attachment structures . Similarly , we did not see elevated levels of cell death in pmr-1 mutants , as reported for the mice knockouts [19] . These data suggest that there may be differences between SPCA function in C . elegans and in vertebrate systems , perhaps due to fundamental differences in ectodermal cell structure . Alternatively , the terminal phenotypes observed in vertebrates could reflect earlier defects . Given our results in C . elegans , as well as the strong sequence conservation of the gene , it is tempting to speculate that the SPCA's have cell migration roles in vertebrates . The relative simplicity of C . elegans as model system , as well as powerful new lineaging tools [50]–[53] , has allowed us to distinguish between the terminal phenotype and the actual defects in cell migration found in pmr-1 mutants . In humans , keratinocytes differentiate in response to changes in extracellular calcium , migrating from the stratum basale into upper layers of the epidermis as they differentiate [84] . It is conceivable that in Hailey-Hailey patients , stress-damaged cells are replaced inefficiently because of a defect in cell migration , caused by altered calcium levels , leading to the observed lesions and keratinocyte defects [23] , [24] . Similarly , the rostral neural tube closure failings in SPCA mouse knockouts [19] could be due to defects in migration of specific subsets of ectodermal cells during this stage of development . In summary , we have shown that the PMR-1/SPCA1 Ca2+/Mn2+ ATPase plays a key role in cell migration during C . elegans embryonic development . Strains carrying mutant pmr-1 show defects in migration of cells along the surface of the embryo following the ingression of cells during gastrulation . The cell migration defects are likely caused by changes in intracellular calcium levels , as they can be enhanced or suppressed by changes in activity of the ITR-1/IP3R or UNC-68/RyR calcium channels . This analysis reveals a new role for calcium signaling in the migration of specific blastomeres during C . elegans development . It also reveals a new role in cell migration for the SPCA1 family of genes .
Strains were grown and maintained under standard conditions [85] . Wild type strain N2 was used as a control . The ru5 and jc10 alleles of pmr-1 were identified in two independent embryonic conditional embryonic lethal screens using the mutagen EMS [85]; J . Simske ) . The tm1840 and tm1750 alleles were produced in the National Bioresource Project ( kindly provided by S . Mitani; [43] . All pmr-1 mutant strains , which were maintained at 15°C , were backcrossed to the N2 strain at least 3 times . The deletions in tm1840 and tm1750 were confirmed using PCR analysis . PCR mapping lines CB4856 , RW700 , deficiency lines CB2775 eDf9/eDf24 I , CB2770 eDf4/eDf24 I , CB2769 eDf3/eDf24 I , KR2838 hDf17/hIn1[unc-54 ( h1040 ) ] , JK1542 ces-1 ( n703 ) qDf7/dpy-5 ( e61 ) srf-2 ( yj262 ) unc-75 ( e950 ) I , SL536 dxDf2/spe-9 ( eb19 ) unc-101 ( m1 ) I , and conventional mapping lines MT465 dpy-5 ( e61 ) I , bli-2 ( e768 ) II , unc-32 ( e189 ) III , MT464 unc-5 ( e53 ) IV , dpy-11 ( e224 ) V , lon-2 ( e678 ) X; DR210 dpy-5 ( e61 ) , daf-16 ( m26 ) , unc-75 ( e950 ) I , CB2010 unc-54 ( e675 ) , dpy-5 ( e61 ) I , JK228 glp-4 ( bn2 ) , unc-54 ( e675 ) I were used in mapping experiments . SU180 itr-1 ( jc5 ) jcIs1 IV , PS2368 itr-1 ( sy327 ) unc-24 ( e138 ) IV and QQ101 vab-9 ( ju6 ) II strains were used for double mutant analysis . For analysis using GFP fusions , F2 progeny exhibiting the Pmr-1 phenotype and carrying the appropriate markers were selected from crosses between pmr-1 ( ru5 ) males and the following strains: RW10026 unc-119 ( ed3 ) ; stIs10026 , SM467 pIs7; rol-6 ( su1006 ) , SM469 pIs6 , SU93 jcIs1 , PD7963 ccIs7963 , SU324 jcIs26 , FR317 swIs1 , OH904 otIs33 , SU272 jcIs1 , evIs136 , and JJ1579 zuIs77 . Strains expressing vab-7::GFP were generated by microparticle bombardment of pJA64 and pDPMM016b into a pmr-1 ( ru5 ) I; unc-119 ( ed3 ) III strain [86] and backcrossed to wild type to generate controls . The ru5 and jc10 alleles were mapped using conventional and deficiency mapping strategies [87] . The position of the ru5 allele was further refined using PCR-based and snp-snp mapping techniques [88]–[90] . Genomic DNA was extracted from the ru5 strain using standard methods and the whole genome was sequenced using Illumina sequencing and MAQGene , as described [91] . Alterations in candidate genes were confirmed in ru5 using PCR amplification followed by subcloning ( Strataclone , Agilent Technologies ) and conventional sequencing ( U . of Iowa Carver Center , U of Wisconsin , and Fred Hutchinson DNA sequencing facilities ) . The coding regions of these candidate genes were then sequenced in jc10 . In complementation analysis , ru5 males were crossed to jc10 , tm1840 , or tm1750 hermaphrodites and the F1 progeny scored for embryonic lethality at 25°C . For complementation rescue analysis , ru5 young adults were transformed by injection of 25–100 µg/ml of a fosmid or cosmid covering each of the candidate genes , along with the co-transformation marker pRF4 [92] , [93] . Rol Progeny were assayed for survivorship at 25°C for each injected construct . To determine rates of viability , L4 hermaphrodites were shifted from 15°C to the indicated temperature . Eggs from single hermaphrodites were counted and the number of resultant late larva or adult progeny scored . In brood size experiments , single L4 hermaphrodites were shifted to the indicated temperature and total broods quantified as previously described [94] . Both brood size and viability data were compared using t-tests . Terminal phenotypes were scored during mid-to-late embryogenesis . Embryos that failed to enclose were classified as “severe” , those with head or later body ruptures as “moderate” , and those with cell position or pharynx unattached ( Pun ) phenotypes as “mild” . We used chi-square analysis to compare the frequency of each class for a given strain or condition . For RNA interference experiments with pmr-1 , young adult hermaphrodites were injected with a pmr-1 dsRNA construct mv ZK256 . 1a at 0 . 9 mg/ml using standard methods [95] . Injected worms were allowed to recover and lay eggs . The embryos of injected progeny were examined for levels of embryonic lethality and terminal phenotypes . For RNAi with unc-68 , we used an RNAi feeding protocol [95] . Embryos were extracted from gravid pmr-1 ( ru5 ) adults grown at the indicated temperature from the early-to-mid-L4 stage , shifted to the second temperature at the indicated time point , and then maintained at that second temperature for the duration of embryogenesis . Each embryo was scored for hatching and terminal phenotypes . Experiments at the two temperatures were calibrated based on comparisons of the timing of key developmental events . For lineaging and cell migration analysis , pmr-1 ( ru5 ) embryos carrying nuclear-localized GFP were imaged at 25°C according to the protocol outlined in [50] , [52] , [53] . Four-dimensional representations of embryos were generated using the Acetree program [51] . pmr-1 ( ru5 ) lineages ( n = 6 ) were compared to controls ( n = 6 to 8 ) ( kindly provided by M . Boeck , P . Weisdepp , and R . Waterston ) . Files for each embryo were synchronized to the same developmental time points based on cell division patterns in several lineages . Because of the different migration patterns for the lineages tested , several techniques were used for measuring cell movement . To analyze ingression , the timing of the ingressions of P4 , D , MS , and C-lineage daughters in N2 was compared to pmr-1 ( ru5 ) embryos . C-lineage cell migration was calculated by measuring the distance of the migrating cell nucleus relative to the ventral-most cell in the embryo . The average slopes of distance over time were used for comparisons . Ventral cell migration was calculated by measuring the distance of the migrating cell nucleus relative to the left or right edge of the embryo and total migration distances over time were compared . The position of the anterior–most cells was measured using polar or Cartesian coordinates , the positions plotted over time , and positions compared in each dimension . In all cases , t-tests were used to determine if the differences between the controls and pmr-1 ( ru5 ) embryos were statistically significant . For analysis of the expression patterns of GFP-fusion proteins , control and pmr-1 ( ru5 ) embryos were grown at 25°C and compared at specific developmental stages . In some experiments , the position and number of nuclei expressing markers were analyzed using t-tests . For analysis using in situ immunofluorescence , control and pmr-1 strains were collected and fixed using a freeze cracking technique and stained with anti-SMA-1 , anti-NID-1 , anti-AJM-1 , anti-SQV-8 , or anti-PMR-1 antisera [34] , [62] , [63] , [66] , [96] . Nomarski and fluorescent images of embryos were collected using an Olympus spinning disc microscope using Slidebook , Nikon E600 microscopes using a coolsnap CCD camera controlled by Metaview software , and Olympus DeltaVision deconvolution microscope ( Applied Precision , Issaquah , WA ) ; projected images were created using 4D macros within NIH Image and Image J .
|
During growth or regeneration after damage , skin cells migrate from basal to superficial layers , forming tight attachments that protect an individual from environmental assaults . Proteins that remove calcium from the cell cytosol into secretory stores , where it is available for future release , play a key role in skin cell integrity . Defects in these secretory pathway calcium ATPase ( SPCA ) channels in humans cause Hailey-Hailey disease , a chronic disorder marked by skin lesions in areas of high-stress . Our study of the SPCA gene pmr-1 in Caenorhabditis elegans indicates the gene is essential for viability . Embryos with defective PMR-1 die with cell attachment defects superficially similar to those of Hailey-Hailey disease patients . To better understand this phenotype , we tracked the position of individual cells during development of pmr-1 mutant embryos . This analysis revealed that the cell attachment defects are caused by primary failures in cell migration . We also identified other calcium channel proteins involved in this process , indicating proper regulation of calcium is crucial for cell migration in C . elegans . If SPCA proteins act similarly in humans , this research will lead to better understanding of the molecules important for skin cell regeneration , as well as help to explain the defects observed in Hailey-Hailey disease patients .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"cellular",
"structures",
"subcellular",
"organelles",
"gene",
"function",
"animal",
"models",
"developmental",
"biology",
"caenorhabditis",
"elegans",
"model",
"organisms",
"molecular",
"genetics",
"morphogenesis",
"biology",
"calcium",
"signaling",
"cascade",
"cytoplasm",
"signal",
"transduction",
"genetic",
"screens",
"gene",
"identification",
"and",
"analysis",
"cell",
"migration",
"genetics",
"molecular",
"cell",
"biology",
"genetics",
"of",
"disease",
"cell",
"fate",
"determination",
"signaling",
"cascades"
] |
2013
|
The Secretory Pathway Calcium ATPase PMR-1/SPCA1 Has Essential Roles in Cell Migration during Caenorhabditis elegans Embryonic Development
|
Polycomb proteins are epigenetic regulators that localize to developmental loci in the early embryo where they mediate lineage-specific gene repression . In Drosophila , these repressors are recruited to sequence elements by DNA binding proteins associated with Polycomb repressive complex 2 ( PRC2 ) . However , the sequences that recruit PRC2 in mammalian cells have remained obscure . To address this , we integrated a series of engineered bacterial artificial chromosomes into embryonic stem ( ES ) cells and examined their chromatin . We found that a 44 kb region corresponding to the Zfpm2 locus initiates de novo recruitment of PRC2 . We then pinpointed a CpG island within this locus as both necessary and sufficient for PRC2 recruitment . Based on this causal demonstration and prior genomic analyses , we hypothesized that large GC-rich elements depleted of activating transcription factor motifs mediate PRC2 recruitment in mammals . We validated this model in two ways . First , we showed that a constitutively active CpG island is able to recruit PRC2 after excision of a cluster of activating motifs . Second , we showed that two 1 kb sequence intervals from the Escherichia coli genome with GC-contents comparable to a mammalian CpG island are both capable of recruiting PRC2 when integrated into the ES cell genome . Our findings demonstrate a causal role for GC-rich sequences in PRC2 recruitment and implicate a specific subset of CpG islands depleted of activating motifs as instrumental for the initial localization of this key regulator in mammalian genomes .
Polycomb proteins are epigenetic regulators required for proper gene expression patterning in metazoans . The proteins reside in two main complexes , termed Polycomb repressive complex 1 and 2 ( PRC1 and PRC2 ) . PRC2 catalyzes histone H3 lysine 27 tri-methylation ( K27me3 ) , while PRC1 catalyzes histone H2A ubiquitination and mediates chromatin compaction [1] , [2] . PRC1 and PRC2 are initially recruited to target loci in the early embryo where they subsequently mediate lineage-specific gene repression . In embryonic stem ( ES ) cells , the complexes localize to thousands of genomic sites , including many developmental loci [3]–[5] . These target loci are not yet stably repressed , but instead maintain a “bivalent” chromatin state , with their chromatin enriched for the activating histone mark , H3 lysine 4 tri-methylation ( K4me3 ) , together with the repressive K27me3 [6] , [7] . In the absence of transcriptional induction , PRC1 and PRC2 remain at target loci and mediate repression through differentiation . The mechanisms that underlie stable association of the complexes remain poorly understood , but likely involve interactions with the modified histones [8]–[12] . Proper localization of PRC1 and PRC2 in the pluripotent genome is central to the complex developmental regulation orchestrated by these factors . However , the sequence determinants that underlie this initial landscape remain obscure . Polycomb recruitment is best understood in Drosophila , where sequence elements termed Polycomb response elements ( PREs ) are able to direct these repressors to exogenous locations [13] . PREs contain clusters of motifs recognized by DNA binding proteins such as Pho , Zeste and GAGA , which in turn recruit PRC2 [14]–[17] . Despite extensive study , neither PRE sequence motifs nor binding profiles of PRC2-associated DNA binding proteins are sufficient to fully predict PRC2 localization in the Drosophila genome [1] , [16] , [18] , [19] . While protein homologs of PRC1 and PRC2 are conserved in mammals , DNA sequence homologs of Drosophila PREs appear to be lacking in mammalian genomes [13] . Moreover , it remains controversial whether the DNA binding proteins associated with PRC2 in Drosophila have functional homologs in mammals . The most compelling candidate has been YY1 , a Pho homolog that rescues gene silencing when introduced into Pho-deficient Drosophila embryos [20] . YY1 has been implicated in PRC2-dependent silencing of tumor suppressor genes in human cancer cells [21] . However , this transcription factor has also been linked to numerous other functions , including imprinting , DNA methylation , B-cell development and ribosomal protein gene transcription [22]–[26] . Recently , researchers identified two DNA sequence elements able to confer Polycomb repression in mammalian cells . Sing and colleagues identified a murine PRE-like element that regulates the MafB gene during neural development [27] . These investigators defined a critical 1 . 5 kb sequence element that is able to recruit PRC1 , but not PRC2 in a transgenic cell assay . Woo and colleagues identified a 1 . 8 kb region of the human HoxD cluster that recruits both PRC1 and PRC2 and represses a reporter construct in mesenchymal tissues [28] . Both groups note that their respective PRE regions contain YY1 motifs . Mutation of the YY1 sites in the HoxD PRE resulted in loss of PRC1 binding and partial loss of repression , while comparatively , deletion of a separate highly conserved region from this element completely abrogated PRC1 and PRC2 binding as well as repression [28] . In addition to these locus-specific investigations , genomic studies have sought to define PRC2 targets and determinants in a systematic fashion . The Ezh2 and Suz12 subunits have been mapped in mouse and human ES cells by chromatin immunoprecipitation and microarrays ( ChIP-chip ) or high-throughput sequencing ( ChIP-Seq ) [3]–[5] , [29] . Such studies have highlighted global correlations between PRC2 targets and CpG islands [5] , [30] as well as highly-conserved genomic loci [4] , [7] , [31] . Recently , Jarid2 has been shown to associate with PRC2 and to be required for proper genome-wide localization of the complex [32]–[35] . Intriguingly , Jarid2 contains an ARID and a Zinc-finger DNA-binding domain . However , it is unclear how Jarid2 could account for PRC2 targeting given the lack of sequence specificity and the low affinity of its DNA binding domains [33] , [36] . In summary , a variety of sequence elements including CpG islands , conserved elements and YY1 motifs have been implicated in Polycomb targeting in mammalian cells . Causality has only been demonstrated in two specific instances and a unifying view of the determinants of Polycomb recruitment remains elusive . Here we present the identification of multiple sequence elements capable of recruiting PRC2 in mammalian ES cells . This was achieved through an experimental approach in which engineered bacterial artificial chromosomes ( BACs ) were stably integrated into the ES cell genome . Evaluation of a series of modified BACs specifically identified a 1 . 7 kb DNA fragment that is both necessary and sufficient for PRC2 recruitment . The fragment does not share sequence characteristics of Drosophila PREs and lacks YY1 binding sites , but rather corresponds to an annotated CpG island . Based on this result and a genome-wide analysis of PRC2 target sequences we hypothesized that large GC-rich sequence elements lacking transcriptional activation signals represent general PRC2 recruitment elements . We tested this model by assaying the following DNA sequences: ( i ) a ‘housekeeping’ CpG island which was re-engineered by removal of a cluster of activating motifs; and ( ii ) two large GC-rich intervals from the E . coli genome that satisfy the criteria of mammalian CpG islands . We found that all three GC-rich elements robustly recruit PRC2 in ES cells . We propose that a class of CpG islands distinguished by a lack of activating motifs play causal roles in the initial localization of PRC2 and the subsequent coordination of epigenetic controls during mammalian development .
To identify DNA sequences capable of recruiting Polycomb repressors in mammalian cells , we engineered human BACs that correspond to genomic regions bound by these proteins in human ES cells . We initially targeted a region of the human Zfpm2 ( hZfpm2 ) locus , which encodes a developmental transcription factor involved in heart and gonad development [37] . In ES cells , the endogenous locus recruits PRC1 and PRC2 , and is enriched for the bivalent histone modifications , K4me3 and K27me3 ( Figure 1A ) . We used recombineering to engineer a 44 kb BAC containing this locus and a neomycin selection marker . The modified BAC was electroporated into mouse ES cells , and individual transgenic ES cell colonies containing the full length BAC were expanded ( Figure S1 ) . Fluorescent in situ hybridization ( FISH ) confirmed integration at a single genomic location ( Figure S2 ) . We used ChIP and quantitative PCR ( ChIP-qPCR ) with human specific primers to examine the chromatin state of the newly incorporated hZfpm2 locus . This analysis revealed strong enrichment for K27me3 and K4me3 ( Figure 1B ) . In addition , we explicitly tested for direct binding of the Polycomb repressive complexes using antibody against the PRC1 subunit , Ring1B , or the PRC2 subunit , Ezh2 . We detected robust enrichment for both complexes in the vicinity of the hZfpm2 gene promoter ( Figure 1B ) . To confirm this result and eliminate the possibility of integration site effects , we tested two additional transgenic hZfpm2 ES cell clones with unique integration sites as well as a fourth transgenic ES cell line containing a distinct Polycomb target locus , Pax5 . In each case , we observed a bivalent chromatin state analogous to the endogenous loci ( Figure S3 ) . Similar to endogenous bivalent CpG islands , we found the Zfpm2 CpG island was DNA hypomethylated ( Figure S4 ) . These results suggest that DNA sequence is sufficient to initiate de novo recruitment of Polycomb in ES cells . A key function of Polycomb repressors is to maintain a repressive chromatin state through cellular differentiation . To determine if the integrated BAC is capable of maintaining K27me3 , the hZfpm2 transgenic ES cells were differentiated to neural progenitor ( NP ) cells in vitro [38] . ChIP-qPCR analysis revealed continued enrichment of K27me3 but loss of K4me3 ( Figure 1C ) , a pattern frequently observed at endogenous loci that are not activated during differentiation [39] . This indicates that DNA sequence at the hZfpm2 locus is sufficient to initiate K27me3 chromatin modifications in ES cells , and maintain the repressive chromatin state through neural differentiation . We next sought to define the sequences within the hZfpm2 BAC required for recruitment of Polycomb repressors . First , we re-engineered the 44 kb hZfpm2 BAC to remove 20 kb of flanking sequences that contained distal non-coding conserved sequence elements ( Figure 1A ) . When we integrated the resulting 22 kb construct into ES cells we found that it robustly enriches for PRC1 , PRC2 , K4me3 and K27me3 ( Figure 1B ) . Hence , these particular distal elements do not appear to be required for the recruitment of the complexes . Next , we considered the necessity of the CpG island which corresponds to the peak of Ezh2 enrichment in ChIP-Seq profiles ( Figure 1A ) . We excised a 1 . 7 kb fragment containing the CpG island , and integrated the resulting BAC ( ΔCGI ) into ES cells . The ΔCGI BAC failed to recruit PRC1 or PRC2 , and showed significantly reduced K27me3 levels relative to the other constructs ( Figure 1B ) . This suggests that the CpG island is essential for recruitment of Polycomb proteins to the hZfpm2 locus . We next asked whether the hZfpm2 CpG island is sufficient to recruit Polycomb repressors to an exogenous locus . To test this , we selected an unremarkable gene desert region on human chromosome 1 that shows no enrichment for PRC1 , PRC2 or K27me3 in ES cells ( Figure 2A ) . We also verified that the gene desert BAC alone does not show any enrichment for K27me3 or Ezh2 when integrated into ES cells ( Figure 2B ) . Using recombineering , we inserted the 1 . 7 kb sequence that corresponds to the hZfpm2 CpG island into the gene desert BAC . The resulting construct was integrated into mouse ES cells and three independent clones were evaluated . ChIP-qPCR analysis revealed strong enrichment for K27me3 , K4me3 and PRC2 over the inserted CpG island ( Figure 2C , Figure S5 ) . In contrast , we observed relatively little enrichment for the PRC1 subunit Ring1B ( Figure 2C ) . We confirmed the specificity of these enrichments with primers that span the boundary between the insertion and adjacent gene desert sequence . Notably , K27me3 enrichment was detected across the gene desert locus up to 2 . 5 kb from the inserted CpG island ( Figure 2C ) . This indicates that the localized CpG island can initiate K27me3 that then spreads into adjacent sequence . Lastly we found no YY1 enrichment across the CpG island by ChIP-qPCR ( Figure S5 ) . Together , these data suggest that the hZfpm2 CpG island contains the necessary signals for PRC2 recruitment but is insufficient to confer robust PRC1 association . The functionality of a CpG island in PRC2 recruitment is consistent with prior observations that a majority of PRC2 sites in ES cells correspond to CpG islands [4] , [5] and with the striking correlation between intensity of PRC2 binding and the GC-richness of the underlying sequence ( Figure 2D ) . We therefore considered whether specific signals within the Zfpm2 CpG island might underlie its capacity to recruit PRC2 . First , we searched for sequence motifs analogous to the PREs that recruit PRC2 in Drosophila . We focused on motifs recognized by YY1 , the nearest mammalian homolog of the Drosophila recruitment proteins . Notably , both of the recently described mammalian PREs contain YY1 motifs [27] , [28] . The 44 kb hZfpm2 BAC contains 11 instances of the consensus YY1 motif . However , none of these reside within the CpG island ( Figure S6 ) ( see Methods ) . We also examined YY1 binding directly in ES cells and NS cells using ChIP-Seq . Consistent with prior reports , YY1 binding is evident at the 5′ ends of many highly expressed genes , including those encoding ribosomal proteins , and is also seen at the imprinted Peg3 locus ( Figure 2E , Table S1 ) [26] . However , no YY1 enrichment is evident at the Zfpm2 locus . Moreover , at a global level , YY1 shows almost no overlap with PRC2 or PRC1 , but instead co-localizes with genomic sites marked exclusively by K4me3 ( Figure 2F , Figure S6 , and Table S1 ) . Thus , although YY1 may contribute to Polycomb-mediated repression through distal interactions or in trans , it does not appear to be directly involved in PRC2 recruitment in ES cells . We previously reported that CpG islands bound by PRC2 in ES cells could be predicted based on a relative absence of activating transcription factor motifs ( AMs ) in their DNA sequence [5] . We reasoned that transcriptional inactivity afforded by this absence of AMs is a requisite for PRC2 association [40] , [41] . This could explain why PRC2 is absent from a majority of CpG islands , many of which are found at highly active promoters . Consistent with this model , when we examined a recently published RNA-Seq dataset for poly-adenylated transcripts in ES cells , we found that virtually all of the high-CpG promoters ( HCPs ) lacking Ezh2 are detectably transcribed ( Figure S7 ) . The small proportion of HCPs that are neither Ezh2-bound nor transcribed may reflect false-negatives in the ChIP-Seq or RNA-Seq data . Alternatively , these HCPs tend to correspond to CpG islands with relatively low GC-contents and lengths and may therefore have insufficient GC-richness to promote PRC2 binding ( Figure S7 ) . Thus , correlative analyses implicate large GC-rich elements that lack transcriptional activation signals as general PRC2 recruitment elements in mammals . To obtain direct experimental support for the general sufficiency of large GC-rich elements lacking AMs in PRC2 recruitment , we carried out the following experiments . First , we tested whether a K4me3-only CpG island could be turned into a PRC2 recruitment element by removing activating motifs . We targeted a 1 . 3 kb CpG island that overlaps the promoters of two ubiquitously expressed genes – Arl3 and Sfxn2 . Neither gene carries K27me3 in ES cells , or in any other cell type tested ( Figure S8 , and data not shown ) . This CpG island was selected as it has many conserved AMs clustered in one half of the island ( Figure 3A ) . We hypothesized that the portion of the Arl3/Sfxn2 CpG island lacking AMs would , in isolation , lack active transcription and recruit PRC2 . In contrast , we predicted that the half containing multiple AMs would lack Polycomb . To test this , we generated two additional BAC constructs containing the respective portions of the Arl3/Sfxn2 CpG island positioned within the gene desert , and integrated these constructs into ES cells ( Figure 3A ) . ChIP-qPCR shows that the portion of the CpG island lacking AMs is able to recruit PRC2 and becomes enriched for K27me3 ( Figure 3B ) . In contrast , the AM-containing portion shows no enrichment for K27me3 or Ezh2 , but is instead marked exclusively by K4me3 , similar to the endogenous human locus ( Figure 3C , Figure S8 ) . Thus , a GC-rich sequence element with no known requirement for Polycomb regulation can recruit PRC2 when isolated from activating sequence features . Next , we tested whether even more generic GC-rich elements might also be capable of recruiting PRC2 in ES cells . Here , we focused on sequences derived from the genome of E . coli , reasoning that there would be no selection for PRC2 recruiting elements in this prokaryote given the complete lack of chromatin regulators . We arbitrarily selected three 1 kb segments of the E . coli genome . Two with GC contents above the threshold for a mammalian CpG island but that each contained few AMs , and one AT rich segment as a control ( Table S3 ) . We recombined each segment into the gene desert BAC and integrated the resulting constructs into ES cells . ChIP-qPCR confirmed that both GC-rich E . Coli segments recruit Ezh2 and form a bivalent chromatin state ( Figure 4A , 4B , Figure S9 ) . Notably , the GC-rich segment also enriches for Jarid2 , a PRC2 component with DNA binding activity ( Figure S10 ) . In contrast , the AT-rich segment did not recruit Ezh2 or enrich for either K4me3 or K27me3 ( Figure 4C , Figure S9 ) . Together , our findings suggest that GC-rich sequence elements that lack signals for transcriptional activation have an innate capacity to recruit PRC2 in mammalian ES cells .
Several lines of evidence suggest that the initial landscape of Polycomb complex binding is critical for proper patterning of gene expression in metazoan development [1] , [2] , [13] . Failure of these factors to engage their target loci in embryogenesis has been linked to a loss of epigenetic repression at later stages . Accordingly , the determinants that localize Polycomb complexes at the pluripotent stage are almost certainly essential to the global functions of these repressors through development . We find that DNA sequence is sufficient for proper localization of Polycomb repressive complexes in ES cells , and specifically identify a CpG island within the Zfpm2 locus as being critical for recruitment . We provide evidence that GC-rich elements lacking activating signals suffice in general to recruit PRC2 . This includes demonstrations ( i ) that a motif devoid segment of an active ‘housekeeping’ CpG island can recruit PRC2; and ( ii ) that arbitrarily selected GC-rich elements from the E . coli genome can themselves mediate PRC2 recruitment when integrated into the ES cell genome . Several possible mechanistic models could explain the causality of GC-rich DNA elements in PRC2 recruitment ( Figure 5 ) . First , we note that CpG islands have been shown to destabilize nucleosomes in mammalian cells [42] . At transcriptionally inactive loci , this property could increase their accessibility to PRC2-associated proteins with DNA affinity but low sequence specificity , such as Jarid2 or AEBP2 [32]–[35] , [43] ( Figure S10 ) . Although this association would be abrogated by transcriptional activity at most CpG islands , those lacking activation signals would remain permissive to PRC2 association ( Figure 5 ) . In support of this model , PRC2 targets in ES cells are also enriched for H2A . Z and H3 . 3 , histone variants linked to nucleosome exchange dynamics [44] , [45] . Alternatively or in addition , targeting could be supported by DNA binding proteins with affinity for low complexity GC-rich motifs or CpG dinucleotides , such as CXXC domain proteins [46] . Localization may also be promoted or stabilized by long and short non-coding RNAs [47]–[50] as well as by the demonstrated affinity of PRC2 for its product , H3K27me3 [11] , [12] . Notably , PRC2 recruitment in ES cells appears distinct from that in Drosophila , as we do not find evidence for involvement of PRE-like sequence motifs or mammalian homologues such as YY1 . It should be emphasized that PRC2 localization does not necessarily equate with epigenetic repression . Indeed virtually all PRC2 bound sites in ES cells , and all CpG islands tested here , are also enriched for K4me3 , and presumably poised for activation upon differentiation . Epigenetic repression during differentiation may require PRC1 and thus depend on additional binding determinants . YY1 remains an intriguing candidate in this regard , given prior evidence for physical and genetic interactions with PRC1 [51] , [52] . YY1 consensus motifs are present in the Polycomb-dependent silencing elements recently identified in the MafB and HoxD loci . Interestingly , the HoxD element combines a CpG island with a cluster of conserved YY1 motifs . Mutation of the motifs abrogated PRC1 binding but left PRC2 binding intact . Still , the fact that only a small fraction of documented PRC2 and PRC1 sites have YY1 motifs or binding suggests that this transcription factor may act indirectly and/or explain only a subset of cases . Nonetheless , it is likely that a fully functional epigenetic silencer would require a combination of features , including a GC-rich PRC2 element as well as appropriate elements to recruit PRC1 . Further study is needed to expand the rules for PRC2 binding to include a global definition of PRC1 determinants and ultimately , to understand how the initial landscape facilitates the maintenance of gene expression programs in the developing organism .
BAC constructs CTD331719L ( ‘Zfpm2 44’ ) , CTD-2535J16 ( ‘Pax5’ ) and CTD-3219L19 ( ‘Gene Desert’ ) were obtained from Open Biosystems . Recombineering was done using the RedET system ( Open Biosystems ) in DH10B cells . Homology arms 200–500 bp in length were PCR amplified and cloned into a PGK; Neomycin cassette ( Gene Bridges ) . This cassette was used to recombineer all BACs to enable selection in mammalian cells . The 22 kb hZfpm2 BAC was created by restricting the hZfpm2 BAC at two sites using ClaI , and re-ligating the BAC lacking the intervening sequence . The CpG island was excised from the 22 kb hZfpm2 BAC by amplification of flanking homology arms , and cloned into a construct containing an adjacent ampicillin cassette ( Frt-amp-Frt; Gene Bridges ) . After recombination , the ampicillin cassette was removed using Flp-recombinase and selection for clones that lost ampicillin resistance ( Flp-706; Gene Bridges ) . PCR across the region confirmed excision of the CpG island . For the Gene Desert BACs , the Zfpm2 , Arl3 , Sfxn2 and E . coli CpG islands were amplified with primers containing XhoI sites and cloned into the Frt-amp-Frt vector that contains homology arms from the Gene Desert region . The final constructs were confirmed by sequencing across recombination junctions . All primers used for CpG islands and recombineering homology arms are listed in Table S2 . ES cells ( V6 . 5 ) were maintained in ES cell medium ( DMEM; Dulbecco's modified Eagle's medium ) supplemented with 15% fetal calf serum ( Hyclone ) , 0 . 1 mM ß-mercaptoethanol ( Sigma ) , 2 mM Glutamax , 0 . 1 mM non-essential amino acid ( NEAA; Gibco ) and 1000U/ml recombinant leukemia inhibitory factor ( ESGRO; Chemicon ) . Roughly 50 µg of linearized BAC was nucleofected using the mouse ES cell nucleofector kit ( Lonza ) into 106 mouse ES cells , and selected 7–10 days with 150 µg/ml Geneticin ( Invitrogen ) on Neomycin resistant MEFs ( Millipore ) . Individual resistant colonies were picked , expanded and tested for integration of the full length BAC by PCR . Differentiation of hZfpm2 ES cell clone 1 into a population of neural progenitor ( NP ) cells was done as previously described [53] . FISH analysis was done as described previously [54] . DNA methylation analysis was done as previously described [55] and primers used to amplify bisulfite treated DNA are listed in Table S2 . For each construct , between one and three ES cell clones were expanded and subjected to ChIP using antibody against K4me3 ( Abcam ab8580 or Upstate/Millipore 07-473 ) , K27me3 ( Upstate/Millipore 07-449 ) , Ezh2 ( Active Motif 39103 or 39639 ) , or Ring1B ( MBL International d139-3 ) as described previously [5] , [7] , [39] . ChIP DNA was quantified by Quant-iT Picogreen dsDNA Assay Kit ( Invitrogen ) . ChIP enrichments were assessed by quantitative PCR analysis on an ABI 7500 with 0 . 25 ng ChIP DNA and an equal mass of un-enriched input DNA . Enrichments were calculated from 2 or 3 biologically independent ChIP experiments . For K27me3 , and Ezh2 enrichment , background was subtracted by normalizing over a negative genomic control . Error bars represent standard error of the mean ( SEM ) . We confirmed that the human specific primers do not non-specifically amplify mouse genomic DNA . Primers used for qPCR are listed in Table S2 . Genomewide maps of YY1 binding sites were determined by ChIP-Seq as described previously [39] . Briefly , ChIP was carried out on 6×107 cells using antibody against YY1 ( Santa Cruz Biotechnology sc-1703 ) . ChIP DNA was used to prepare libraries which were sequenced on the Illumina Genome Analyzer . Density profiles were generated as described [39] . Promoters ( RefSeq; http://genome . ucsc . edu ) were classified as positive for YY1 , H3K4me3 or H3K27me3 if the read density was significantly enriched ( p<10−3 ) over a background distribution based on randomized reads generated separately for each dataset to account for the varying degrees of sequencing depth . ChIP-Seq data for YY1 are deposited to the NCBI GEO database under the following accession number GSE25197 ( http://www . ncbi . nlm . nih . gov/projects/geo/query/acc . cgi ? acc=GSE25197 ) . Sites of Ezh2 enrichment ( p<10−3 ) were calculated genomewide using sliding 1 kb windows , and enriched windows within 1 kb were merged . DNA methylation levels were calculated using previously published Reduced Representation Bisulphite Sequenced ( RRBS ) libraries [55] . Composite plots represent the mean methylation level in sliding 200 bp windows in the the 10 kb surrounding the TSSs of the indicated gene sets . YY1 motifs were identified using the MAST algorithm [56] where a match to the consensus motif was defined at significance level 5×10−5 . Candidate CpG islands for TF motif analysis were identified by scanning annotated CpG islands ( http://genome . ucsc . edu ) for asymmetric clustering of motifs related to transcriptional activation in ES cells [5] . Motifs shown in Figure 3A and Figure S6 are from UCSCs TFBS conserved track . GC-rich elements from the E . coli K12 genome were selected by calculating %GC and CpG O/E in sliding 1 kb windows . Sequences matching the criteria for mammalian CpG islands while simultaneously being depleted of motifs related to transcriptional activation [5] were chosen for insertion into mouse ES cells . Transcriptionally inactive HCPs were selected based on a lack of transcript enrichment by both expression arrays [39] and RNA-Seq data [57] . In the case of RNA-Seq , each gene was assigned the maximum read density within any 1 kb window of exonic sequence . To ease analysis of promoter CpG island statistics , only HCPs containing a single CpG island were considered .
|
Key developmental genes are precisely turned on or off during development , thus creating a complex , multi-tissue embryo . The mechanism that keeps genes off , or repressed , is crucial to proper development . In embryonic stem cells , Polycomb repressive complex 2 ( PRC2 ) is recruited to the promoters of these developmental genes and helps to maintain repression in the appropriate tissues through development . How PRC2 is initially recruited to these genes in the early embryo remains elusive . Here we experimentally demonstrate that stretches of GC-rich DNA , termed CpG islands , can initiate recruitment of PRC2 in embryonic stem cells when they are transcriptionally-inactive . Surprisingly , we find that GC-rich DNA from bacterial genomes can also initiate recruitment of PRC2 in embryonic stem cells . This supports a model where inactive GC-rich DNA can itself suffice to recruit PRC2 even in the absence of more complex DNA sequence motifs .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"molecular",
"biology/histone",
"modification",
"developmental",
"biology/stem",
"cells",
"genetics",
"and",
"genomics/gene",
"expression",
"developmental",
"biology/cell",
"differentiation",
"genetics",
"and",
"genomics/epigenetics",
"molecular",
"biology/chromatin",
"structure"
] |
2010
|
GC-Rich Sequence Elements Recruit PRC2 in Mammalian ES Cells
|
Experimental design attempts to maximise the information available for modelling tasks . An optimal experiment allows the inferred models or parameters to be chosen with the highest expected degree of confidence . If the true system is faithfully reproduced by one of the models , the merit of this approach is clear - we simply wish to identify it and the true parameters with the most certainty . However , in the more realistic situation where all models are incorrect or incomplete , the interpretation of model selection outcomes and the role of experimental design needs to be examined more carefully . Using a novel experimental design and model selection framework for stochastic state-space models , we perform high-throughput in-silico analyses on families of gene regulatory cascade models , to show that the selected model can depend on the experiment performed . We observe that experimental design thus makes confidence a criterion for model choice , but that this does not necessarily correlate with a model's predictive power or correctness . Finally , in the special case of linear ordinary differential equation ( ODE ) models , we explore how wrong a model has to be before it influences the conclusions of a model selection analysis .
Mathematical models provide a rich framework for biological investigation . Depending upon the questions posed , the relevant existing knowledge and alternative hypotheses may be combined and conveniently encoded , ready for analysis via a wealth of computational techniques . The consequences of each hypothesis can be understood through the model behaviour , and predictions made for experimental validation . Values may be inferred for unknown physical parameters and the actions of unobserved components can be predicted via model simulations . Furthermore , a well-designed modelling study allows conclusions to be probed for their sensitivity to uncertainties in any assumptions made , which themselves are necessarily made explicit . While the added value of a working model is clear , how to create one is decidedly not . Choosing an appropriate formulation ( e . g . mechanistic , phenomenological or empirical ) , identifying the important components to include ( and those that may be safely ignored ) , and defining the laws of interaction between them remains highly challenging , and requires a combination of experimentation , domain knowledge and , at times , a measure of luck . Even the most sophisticated models will still be subject to an unknown level of inaccuracy – how this affects the modelling process , and in particular experimental design for Bayesian inference , will be the focus of this study . Both the time and financial cost of generating data , and a growing understanding of the data dependancy of model and parameter identifiability [1] , [2] , has driven research into experimental design . In essence , experimental design seeks experiments that maximise the expected information content of the data with respect to some modelling task . Recent developments include the work of Liepe et . al [2] that builds upon existing methods [3]–[8] , by utilising a sequential approximate Bayesian computation framework to choose the experiment that maximises the expected mutual information between prior and posterior parameter distributions . In so doing , they are able to optimally narrow the resulting posterior parameter or predictive distributions , incorporate preliminary experimental data and provide sensitivity and robustness analyses . In a markedly different approach , Apgar et . al [8] use control theoretic principles to distinguish between competing models; here the favoured model is that which is best able to inform a controller to drive the experimental system through a target trajectory . In order to explore the effects of model inaccuracies we work with a computationally efficient experimental design framework . We build on the methods of Flassig and Sundmacher [9] where expected likelihoods are predicted using efficient Sigma-point approximations and leveraged for optimal experimental design , and Busetto et al . [10] where choosing the optimal measurement readouts and time points is undertaken in an iterative fashion , using Sigma-point approximations to update the posterior distributions . Here we show how mixtures distributions may be exploited to cope with non-Gaussian parameter and predictive distributions and further , derive an extension to the case of stochastic state space models . The intuition behind the approach ( described fully in Materials and Methods ) is shown in Figure 1 , where for identical inputs , two ODE models ( illustrated in blue and red respectively ) are simulated for a range of parameter values , with times and representing two possible choices of times at which the true system can be measured and data gathered . Time represents an uninformative experimental choice since the behaviour of the two models is very similar , while data obtained at time is more likely to favour one model over another , since the distributions of simulated trajectories completely separate . More formally , the key steps in the method are as follow: Firstly we define the limited range of experimental options to be explored and encode them as parameterised extensions of the competing models . Secondly , the so called unscented transform ( UT ) [11] is used to approximate the prior predictive distribution as a mixture of Gaussians , for each model and a given experiment . Finally , optimisation is performed over the experiment parameters in order to best 'separate' the prior predictive distributions of the competing models . Parameters obtained by this optimisation represent an experiment whose generated data is predicted to maximise the differences in the subsequent marginal likelihood values of the models . The contributions of this article are threefold; firstly , we extend a promising and computationally efficient experimental design framework for model selection to the stochastic setting , with non-Gaussian prior distributions; secondly , we utilise this efficiency to explore the robustness of model selection outcomes to experimental choices; and finally , we observe that experimental design can give rise to levels of confidence in selected models that may be misleading as a guide to their predictive power or correctness . The latter two points are undertaken via high-throughput in-silico analyses ( at a scale completely beyond the Monte Carlo based approaches mentioned above ) on families of gene regulatory cascade models and various existing models of the JAK STAT pathway .
We first illustrate the experimental design and model selection framework in the context of crosstalk identification . After observing how the choice of experiment can be crucial for a positive model selection outcomes , the example will be used to illustrate and explore the inconsistency of selection between misspecified models . We consider pairs of regulatory cascades , each consisting of four transcription factors , modelled by ordinary differential equations of the form , for , where is the rate at which protein degrades , represents the maximal rate of production of , is the amount of the transcription factor , , needed for half the maximal response , and is called the Hill-coefficient , and determines the steepness of the response . A range of crosstalk models are formed ( Figure 2 ) by inserting additional regulatory links between and with the same kinetics as above . A single model is chosen as the 'true' biological system to which we perform experiments , and six others with equal prior probabilities are proposed as models of the true system – our task will be to identify the most suitable one . An experiment is defined by the parameter , where denotes the strength of an external stimulus to the production of , which is modelled as a term , ( 1 ) ( 2 ) ( 3 ) added to the relevant ODE equations . The time delay between the two stimulus applications is given by , and is the time at which a single measurement of the system ( of species only ) is taken . Prior distributions for the model parameters are set as Gaussian with means of and covariances of for both the and respectively , with the Hill coefficient fixed at . The results of this round of experimental design are shown in the top left of Figure 3 , where a good choice of is found to be , with a corresponding score of . From the figure , it can be seen that this experiment is predicted to distinguish some pairs of models better than others . In particular , the distribution of scores suggests that while the marginal likelihoods of most pairs of models are separated as desired , there is no power to discriminate between models and , or models and . Indeed , data obtained by performing the experiment upon our 'true' system , leads to posterior probabilities for each model with the same pattern . As a sanity check , we first choose the true model from amongst the set of competing models ( ) , and as expected find that it is recovered by model selection with probability 1 . However if the true model is not represented by ( a far more realistic case ) but instead the crosstalk model with a single connection from to , then models and are found to have similar posterior probabilities of approximately . Likewise , and share a posterior probability of , while a clear difference exists between any other pair of models . To distinguish further between the pair of highest scoring models , a further round of experimental design was performed , with the resulting experiment and data providing strong evidence in favour of model . In an attempt to evaluate the added value of choosing rationally for this example , we calculate scores for a uniform sample of values of from the same range as explored above . The resulting score distribution shown in Figure 4a , peaks in the interval which corresponds to an average Hellinger distance of between the maximally separated marginal likelihoods of each pair of models . This is in contrast to the experiment found by our approach which lives in the tail of the distribution , with an average Hellinger distance of , and highlights how unlikely it is to find suitable experiments by chance alone . Experiments with even higher information content are found , which suggests that more care could be taken with the optimisation of , by for example , increasing the population size , or number of generations of the genetic algorithm used . Perhaps unnervingly , the evidence in the first experiment is found to contradict ( though not significantly in this case ) the decision in favour of model over , which is based on additional data from the second experiment . This suggests the possibility that the choice of experiment influences not only the amount of information available to select a particular model , but also the outcome of the model selection itself . Indeed the distribution of independently selected models from data generated by random experiments is surprisingly flat ( Figure 4b ) . Even at very low levels of assumed noise , the most frequently selected model is chosen for less than half the experiments undertaken . This has been , to our knowledge , completely overlooked by the experimental design literature , but has important implications that we will explore further below . To examine this last observation in more detail , we work with three of the crosstalk models described above , with connections between , , and respectively . The last of these is designated as the true model , and the others are considered as competing hypotheses about the location of the crosstalk connection . We perform 36100 experiments to collect data sets of size 1 , 2 , 4 and 8 equally spaced time points , each consisting of simulating the true model with different values of that correspond to changes in the delay between stimulus applications , and variation of the time at which the state of is first measured . An independent round of model selection is performed for each data set , and the posterior probabilities for each model are calculated . The results for data sets of size 1 and 8 are illustrated in Figure 4c and 4d as heatmaps of posterior probabilities of the first model , and show that the vast majority of the space of experiments is split into distinct regions of high , low and equal probability for each model . In the case of a single time point , most of the explored experiment subspace is found to be uninformative , with the data providing equal support for each model . Three other distinct regions are identified , of which two show decisive support ( on the Jeffreys scale ) for the first model , and one for which the second model is chosen decisively . In other words , by varying the experimental conditions an unequivocal choice ( in isolation ) for either model can be obtained . As more data points are considered , the uninformative region grows smaller , but regions of decisive support for each model remain . Interestingly , these regions are located in distinctly different places for single or multiple time points , although they remain similar for 2 or more time points . This reflects the added value of time series experiments – the marginal likelihoods now balance the ability of the models to reproduce each time point , with their ability to capture the autocorrelation of the time series . In order to establish whether the observed inconsistencies are an artefact of the UT approximations , we perform a similar but necessarily course grained study using MultiNest [12] , [13] , an implementation of nested sampling ( a Monte Carlo based technique with convergence rate [14] ) . Results obtained using MultiNest ( shown in the upper right of figure 5 ) are almost identical to those of figure 4c , displaying the same regions of decisive support for each model . Given how difficult it is to estimate marginal likelihoods in general , the excellent performance of the UT ( with only one Gaussian component ) may seem rather surprising , until one notes that for the models and experiments considered , the prior predictive distributions are approximately Gaussian themselves ( Figure 5 ) . We discuss how the framework can deal with non-Gaussian effects , such as those found in the next examples , in the appendix . In this section we undertake an analysis of three mass action models of varying degrees of resolution of the JAK-STAT signalling pathway [15] . Each model describes the initial pathway activity after receptor activation ( Figure 6 ) , but before any feedback occurs . In brief , the signalling process consists of a receptor binding to JAK to form a complex that can dimerise in the presence of interferon- ( IFN ) . This dimer is activated by phosphorylation by JAK , and in turn deactivated after being bound by tyrosine phosphatase ( SHP_2 ) . In its active state , the receptor complex phosphorylates cytoplasmic STAT1 , which is then able to dimerise and act as a transcription factor [16] . We take the most detailed model , , with 17 state variables and 25 parameters ( published by Yamada et al . [16] ) , as our true system to which in-silico experiments can be performed , and select between two of the other models proposed by Quaiser et al . The first of these competing models , , simplifies the true system , by neglecting a reaction – the re-association of phosphorylated STAT1 to the activated receptor – and thereby reducing the system to 16 states and 23 parameters . A series of five other 'biologically inspired' simplifications leads to our second model , , which has 9 states and 10 parameters ( these steps are summarised in Figure 6 ) . We set the parameter priors as a component mixture of Gaussians fit to a uniform sample from the hypercube , where is the parameter dimension , such that all the parameter values inferred for each model by Quaiser et al . are supported . We define and undertake two classes of experiment upon the true model ( with parameters fixed to the published values ) ; in the first , the IFN stimulus strength and the initial time point of a time series of 8 equally spaced measurements of the amount of JAK bound to the receptor are varied , and in the second , the species to be measured and the time at which this first measurement takes place are adjusted . Model selection outcomes for each experiment ( shown in Figure 7 ) show similar features to those for the crosstalk models , with distinct region of high posterior probability for each model . For the first class of experiments , selection between models and reveals strong support for the simpler model when data is gathered at earlier time points . The more complex model , , is generally favoured for later time series , and also for a very limited range of IFN stimuli strengths at early time series . For the second class of experiments , the model selection outcome is found to depend strongly upon which species is measured . The simpler model is chosen decisively and almost independently of the measurement times considered when cytoplasmic phosphorylated STAT1 , in monomeric or dimeric form , or two forms of the receptor complex ( IFN_R_JAKPhos_2 and IFN_R_JAK ) are measured . The same is true of the complex model for measurements of two other forms of the receptor complex ( IFN_R_JAK2 and IFN_R_JAKPhos_2_SHP_2 ) . Otherwise the model selection outcome is time dependant or the choice of species is found to be uninformative . Both these case studies make it clear that under the realistic assumption that all models are more or less incorrect , model selection outcomes can be sensitive to the choice of experiment . This observation has particular importance for studies that treat models as competing hypotheses that are decided between using experimental data; it is quite possible that if different experiments are undertaken , the conclusions drawn will also be different . In particular , the confidence calculated for such a conclusion ( using the Jeffreys scale or another measure ) can be misleading as a guide to how correct or predictive a model is ( Figure 8a ) ; in both the examples studied here , conditions exist such that any of the competing models can score a 'decisive' selection . The model selection outcome and associated confidence must therefore be strictly interpreted , as only increasing the odds of one model ( with respect to others ) for the data gathered under the specific experimental conditions . In light of this observation , the role of experimental design may need to be examined further . Since different models can be selected depending on the experiment undertaken , the use of experimental design will necessarily lead to choosing the model which , for some 'optimal' experiment , has the highest possible predicted level of confidence i . e . experimental design implicitly makes confidence a selection criterion . Is it misleading to claim high confidence in a model selection result when the models have been set up ( by extensions to mimic the optimal experiment ) for this purpose ? Is a bias introduced into the inference via experiment design ? In the context of experiment design for parameter estimation , MacKay suggests this is not a problem [17] , stating that Bayesian inference depends only on the data collected , and not on other data that could have been gathered but was not . Our situation here is different since we consider changes not only to the data collection procedure , but also the data generation process and in turn the competing models themselves . It seems plausible that some models will gain or lose more flexibility than others with regards to fitting data for a particular choice of experiment . Even if the actual model selection is not biased , the confidence we associate with it will scale with the optimality of the experiment . After performing the optimal experiment , should there be any surprise that the selected model seems to have high support from the data ? We feel these questions need further investigation . In practical terms , the important question seems to be: how wrong does the model structure ( or parameter values ) have to be before the less predictive model ( or that which captures less about the true system ) is chosen ? Clearly the answer is sensitive to the system and models under study , and moreover , the issue of how to compare the size of different structural inaccuracies is non trivial . Here , as a first attempt , we limit ourselves to considering the simple case of parameter inaccuracies in linear ODE models . We define a 'base' model as the linear ode system defined by its Jacobian matrix with entries , and 'extensions' to this model as an extra row and column , Biologically such an extension may represent the inclusion of an extra molecular species into the model , along with rules for how it interacts with components of the original system . Defining true base and extension models by and , we consider two models , and where and , are competing ( true and false ) hypotheses about the structure of the model extension , with a zero or indicating a belief that species does not directly affect the rate of increase of species . Parameters , are the unknown strengths of these interactions , over which we place a component mixture of Gaussians prior , fit to a uniform distribution over the interval for each parameter . We represent inaccuracies in modelling the base as additive perturbations and . Data was generated by simulating the state of the first variable of the true model at times , for initial condition . Model selection outcomes for different pairs of values for the perturbations , are shown in Figure 9 . Distinct regions for each possible outcome are found and colour coded in the figure , with red indicating that the true extension has been identified successfully , yellow representing a decision in favour of the false extension , orange that evidence for either model is not substantial on the Jeffreys scale , and finally blue indicating that the marginal likelihood for both models is found to be less than , for which any conclusion would be subject to numerical error . Increasing this threshold has the effect of replacing red areas with blue . In the majority of cases tested , the true extension is correctly identified despite inaccuracies in the base model . However , a set of perturbations are seen to confound the selection , and allow the false extension to obtain substantial support . Furthermore , the selection outcome is found to be more sensitive in some directions than others , with relatively small perturbations to base model entry causing a change in outcome and creating decision boundaries near the lines and . Prior to our analysis , it would be hard to predict these observations even when the true model is known and as simple as that explored here . In real applications , where the true model is unknown and more complex , it may not be possible to tell whether a conclusion is an artefact of model inaccuracies , even when the truth of the conclusion itself can be tested by direct experimental measurement . However , the type of analysis undertaken here at least gives a measure of robustness for the conclusion to a range of model inaccuracies . Unfortunately , this remains difficult to implement in a more general setting – for example , in climatology , where the accepted method of coping with structural uncertainty is through the use of large ensembles of similar models produced by various research groups [18] , a luxury that cannot be afforded on the scale of the most ambitious systems biology projects . While the practical challenges of dealing with large numbers of models is somewhat overcome by the model selection algorithm described above , a harder conceptual problem exists of how to define perturbations to more complicated classes of model , and to compare their strengths . Finally , the example also highlights the difficulty of testing a hypothesis that represents only part of a model . The study shows that the implicit assumption that the base model is accurate , is not necessarily benign , and can affect any conclusions drawn – a result that is borne out by the logical principle that from a false statement , anything is provable .
The scale of the analyses detailed above , comprising thousands of marginal likelihood computations , requires extreme computational efficiency . Indeed it is completely beyond Monte Carlo based methods such as that recently developed by Liepe et al . [2] , which are limited to exploring small sets of models and experiments . Here , the efficiency was obtained by using the unscented transform for propagating Gaussian mixture distributions through non-linear functions . Further computational savings can be made by exploiting the highly parallelizable nature of Flassig and Sundmacher's method [9] , which we have extended for use with mixture distributed priors and stochastic state space models . This efficiency has allowed us to explore model selection problems involving relatively large numbers of models and experiments , and investigate the robustness of model selection results to both changes in experimental conditions and inaccuracies in the models . Results from the latter two studies illustrate some common , but often ignored , pitfalls associated with modelling and inference . Firstly , we show that the conclusions of a model selection analysis can change depending on the experiment undertaken . Related to this , we observe that confidence in such a conclusion is not a good estimator of the predictive power of a model , or the correctness of the model structure . Further we note that the use of experimental design in this context maximises the expected discriminatory information available , and implicitly makes confidence in the outcome a criterion for model selection . In the future we intend to investigate the desirability of this property and how it affects the interpretation of the confidence associated with model selection outcomes . At the heart of these issues is a lack of understanding of the implications of model ( or parameter ) inaccuracies . Often improved fits to data or better model predictions are interpreted as evidence that more about the true system is being captured . This assumption underlines a guiding paradigm of systems biology [19] , where a modelling project is ideally meant to be a cycle of model prediction , experimental testing and subsequent data inspired model/parameter improvement . However , it is possible that improved data fitting and predictive power ( although desirable in their own right ) can be achieved by including more inaccuracies in the model . In the context of parameter estimation , this concept of local optima is widely known , and their avoidance is a challenge when performing any non-trivial inference . One simple method to do so is to include random perturbations in the inference , in order to 'kick' the search out of a local optimum . Perhaps a similar strategy might be included in the modelling paradigm; by performing random experiments , or adding or removing interactions in a model structure , data might be gathered or hypotheses generated that allows a leap to be made to a more optimal solution . While we have been concerned solely with the statistical setting , it is reasonable to expect similar results can be found for alternative model discrimination approaches e . g the use of Semidefinite programming to establish lower bounds on the discrepancy between candidate models and data [20] . Here the particular subset of models that are invalidated will be dependent upon the experiment undertaken . However , emphasis on invalidating wrong models instead of evaluating the relative support for each at least reduces the temptation for extrapolated and , perhaps , false conclusions . George E . P . Box famously stated that 'Essentially , all models are wrong , but some are useful' . Here we would add that if nothing else , models provide a natural setting for mathematicians , engineers and physicists to explore biological problems , exercise their own intuitions , apply theoretical techniques , and ultimately generate novel hypotheses . Whether the hypotheses are correct or not , the necessary experimental checking will reveal more about the biology .
The UT is a method that describes how the moments of a random variable , , are transformed by a non-linear function , . The algorithm begins by calculating a set of weighted particles ( called sigma-points ) with the same sample moments up to a desired order as the distribution . For the results shown here , we use a scaled sigma-point set that captures both means and covariances [21] , where is the dimension of , and are the mean and covariance of , represents the th column of a matrix , and The sigma-point weights are given by , and finally , the parameters , and may be chosen to control the positive definiteness of covariance matrices , spread of the sigma-points , and error in the kurtosis respectively . For the results in this article we take as is standard in the literature [22] , and which is optimal for Gaussian input distributions , while , controlling the spread of sigma-points is taken small as to avoid straddling non-local non-linear effects with a single Gaussian component [21] . The mean and covariance of the variable , can be estimated as the weighted mean and covariance of the propagated sigma-points , ( 4 ) ( 5 ) We denote the resulting approximate probability density function for , by . By matching terms in the Taylor expansions of the estimated and true values of these moments , it can be shown that the UT is accurate to second order in the expansion . More generally , if the sigma-point set approximates the moments of up to the order then the estimates of the mean and covariance of will be accurate up to the term [11] . Crucially , the number of points required ( for this scheme ) is much smaller than the number required to reach convergence with Monte-Carlo methods . We will consider discrete time state space models , , with state–transition ( ) and observation ( ) functions both parametrized by , ( 6 ) ( 7 ) where , is the time series of dimensional measurements that we are trying to model , is the dimensional true state of the system at time , and , and are independent , but not necessarily additive , Gaussian white-noise process and measurement terms . Bayesian model selection compares competing models , , by combining the a priori belief in each model , encoded by the model prior distribution , with the evidence for each model in the data , as quantified by the marginal likelihood , where is the parameter prior for model . In the Bayesian setting , the relative suitabilities of a pair of models are often compared using the ratio of posterior probabilities , known as the Bayes factor , with a Bayes factor of seen as substantial [23] . However , for complex or stochastic models , the marginal likelihood can be intractable , and so approximate likelihood free methods , such as Approximate Bayesian Computation are becoming increasingly important and popular within the biosciences [24] . A big drawback of such Monte-Carlo based algorithms is the large number of simulations – and associated computational cost – required to estimate the posterior distributions or Bayes factors . Even with GPU implementation [25] , applications are currently still limited to comparing pairs or handfuls of models . In order to address the issues raised above , a higher-throughput model selection algorithm is needed . Our approach will be to fit mixture of Gaussian models to the prior parameter distribution for each model , so that we can exploit the UT within the state-space framework to drastically reduce the number of simulations necessary to estimate the distribution of the output of the model . Gaussian mixture measurement and process noise can also be considered , as in the work on Gaussian sum filters [26] , [27] , although the number of mixture components required to model the output at each time point then increases exponentially , and in the case of long time series , component reduction schemes need to be implemented . With this approximation , the marginal likelihood may be expressed as the sum , ( 8 ) ( 9 ) ( 10 ) where the components , , can be determined using the UT as described below . Note that the accuracy of the approximation can be controlled by the number of components used . However , in the presence of nonlinearities , choosing the number and position of components solely to fit the prior distribution may not be adequate . This is because we need to have enough flexibility to also fit a complex and possibly multi-modal output . Indeed , except at the asymptotic limit of dense coverage by the mixture components , it is possible to construct badly behaved mappings that will lead to loss of performance . For the applications visited in this article , the models proved well behaved enough such that a single component and 10 components respectively for the crosstalk and JAK-STAT systems sufficed for sufficient agreement with the nested sampling and Monte Carlo results . An improvement to the method described here would be to update the number of components automatically with respect to the model behaviour in a manner similar to how Gaussian mixtures can be adaptively chosen in particle based simulation of Liouville-type equations [28] , [29] . For the deterministic case including the examples considered in this article , we have , and the state–space model simplifies to , where might represent the simulation of certain variables of a system of ODEs , parameterised by , with additive measurement error . In this case the marginal likelihood can then be expressed as , where each component is obtained simply through application of the UT with input distribution , and liklihood that is Gaussian with mean , , and variance , . To estimate the marginal likelihood in the stochastic case ( ) , we assume the observation function takes the form of a linear transformation of the true state and measurement noise at time with additive noise , ( 11 ) where is an matrix . In practice this might correspond to the common situation where observations are scaled measurements of the abundance of various homo- or heterogeneous groups of molecules . We may then write the mean of the observation , , in terms of the statistics of , ( 12 ) for any , and from the bilinearity of the covariance function , the covariance between any pair of observations , , as , ( 13 ) ( 14 ) since is independent of for all and . We now need to find expressions for the process state covariance terms in equation 14 . To do so we apply the UT iteratively for to transform the state-variable , through the state-transition function , with input distribution given by , The result is a Gaussian approximation to the joint distribution for each , and hence also to the conditional distributions . Given that is a Markov process and that the product of Gaussian functions is Gaussian , we also have a Gaussian expression for the joint distribution , , The covariance between any pair of observations and , may then be found by substituting relevant entries from the covariance matrix of the density of Equation into Equation 14 . The subsequent Gaussian approximation to the joint distribution of , given , constitutes one component in the mixture approximation of the marginal likelihood given in Equation 10 . We first introduce a vector of experiment parameters , , that describes how the dataset is created , specifying , for example , the times at which the system is stimulated , the strengths and targets of the stimuli , knockouts or knockdowns , along with the choice of observable to be measured at each time point . We can then model the system and experiments jointly , extending the to include terms describing the possible experimental perturbations , and the to capture the measurement options , ( 15 ) ( 16 ) We assume that there is overlap between the system observables appearing in each model so that experiments that allow model comparison can be designed . To illustrate how this might be done in practice , we consider a typical set of ordinary differential equations used to describe a gene regulatory mechanism , ( 17 ) ( 18 ) where are the parameters controlling the rates of production and degradation of an mRNA , , and a protein , , subject to the concentration of a repressor protein , . We define the state transition function as their solution evaluated at the next measurement time-point which is now dependant on the choice of , given the state at time , and subject to some additive noise . These equations have be extended as , ( 19 ) ( 20 ) to model a range of possible experimental perturbations , e . g . setting mimics a knockout of the gene producing mRNA , and an input stimulus to species . The observation function , as before can be some linear function of the states , however , the selection of variables and coefficients is now an experimental choice specified by , Given a particular set of experimental options , , the marginal likelihood of model for any possible data set ( the prior predictive distribution ) can be estimated efficiently from equation 10 , with the components calculated with respect to the extended system and experiment model . Comparisons between such prior predictive distributions for competing models provides a means to predict the discriminatory value of a proposed experiment . Intuitively , values of , for which the prior predictive distributions of two models are separated , correspond to experimental conditions under which the models make distinct predictions of the system behaviour . Data gathered under these conditions are thus more likely to yield a significant model selection outcome . More formally , we can quantify the value of an experiment , using the Hellinger distance between the prior predictive distributions , which takes the following closed form for multivariate Gaussian distributions , and , where , or for Gaussian mixtures , it can be evaluated using the method suggested in [30] . The experimental design problem may then be posed as an optimisation problem ( the results in this article used a genetic algorithm [31] of population size and generations ) over - we search for the set of experimental parameters , , for which the Hellinger distance between the competing models , , is maximal . will then specify the experiment that gives the greatest chance of distinguishing between and . In the case where more than two models are considered , the cost function is taken as where the sum of exponentials is introduced to encourage selection of experiments with a high chance of distinguishing between a subset of the model pairs , over experiments with less decisive information for any pair of models , but perhaps a larger average Hellinger distance over all model pairs .
|
Different models of the same process represent distinct hypotheses about reality . These can be decided between within the framework of model selection , where the evidence for each is given by their ability to reproduce a set of experimental data . Even if one of the models is correct , the chances of identifying it can be hindered by the quality of the data , both in terms of its signal to measurement error ratio and the intrinsic discriminatory potential of the experiment undertaken . This potential can be predicted in various ways , and maximising it is one aim of experimental design . In this work we present a computationally efficient method of experimental design for model selection . We exploit the efficiency to consider the implications of the realistic case where all models are more or less incorrect , showing that experiments can be chosen that , considered individually , lead to unequivocal support for opposed hypotheses .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"control",
"theory",
"applied",
"mathematics",
"signaling",
"networks",
"systems",
"science",
"mathematics",
"statistics",
"(mathematics)",
"network",
"analysis",
"complex",
"systems",
"computer",
"and",
"information",
"sciences",
"theoretical",
"biology",
"biophysics",
"systems",
"biology",
"biochemistry",
"biochemical",
"simulations",
"biology",
"and",
"life",
"sciences",
"physical",
"sciences",
"computational",
"biology",
"statistical",
"methods",
"biophysical",
"simulations"
] |
2014
|
Model Selection in Systems Biology Depends on Experimental Design
|
Ebola virus ( EBOV ) is a significant human pathogen that presents a public health concern as an emerging/re-emerging virus and as a potential biological weapon . Substantial progress has been made over the last decade in developing candidate preventive vaccines that can protect nonhuman primates against EBOV . Among these prospects , a vaccine based on recombinant vesicular stomatitis virus ( VSV ) is particularly robust , as it can also confer protection when administered as a postexposure treatment . A concern that has been raised regarding the replication-competent VSV vectors that express EBOV glycoproteins is how these vectors would be tolerated by individuals with altered or compromised immune systems such as patients infected with HIV . This is especially important as all EBOV outbreaks to date have occurred in areas of Central and Western Africa with high HIV incidence rates in the population . In order to address this concern , we evaluated the safety of the recombinant VSV vector expressing the Zaire ebolavirus glycoprotein ( VSVΔG/ZEBOVGP ) in six rhesus macaques infected with simian-human immunodeficiency virus ( SHIV ) . All six animals showed no evidence of illness associated with the VSVΔG/ZEBOVGP vaccine , suggesting that this vaccine may be safe in immunocompromised populations . While one goal of the study was to evaluate the safety of the candidate vaccine platform , it was also of interest to determine if altered immune status would affect vaccine efficacy . The vaccine protected 4 of 6 SHIV-infected macaques from death following ZEBOV challenge . Evaluation of CD4+ T cells in all animals showed that the animals that succumbed to lethal ZEBOV challenge had the lowest CD4+ counts , suggesting that CD4+ T cells may play a role in mediating protection against ZEBOV .
Ebola virus ( EBOV ) has been associated with sporadic episodes of hemorrhagic fever ( HF ) that produce severe disease in infected patients . Mortality rates in outbreaks have ranged from 50% for Sudan ebolavirus ( SEBOV ) to up to 90% for Zaire ebolavirus ( ZEBOV ) ( reviewed in [1] ) . A recent outbreak caused by an apparently new species of EBOV in Uganda appears to be less pathogenic than SEBOV or ZEBOV with a preliminary case fatality rate of about 25% [2] . EBOV is also considered to have potential as a biological weapon and is categorized as a Category A bioterrorism agent by the Centers for Disease Control and Prevention [3]–[5] . While there are no vaccines or postexposure treatment modalities available for preventing or managing EBOV infections there are at least four different vaccine systems that have shown promise in completely protecting nonhuman primates against a lethal EBOV challenge [6]–[12] . Of these prospective EBOV vaccines two systems , one based on a replication-defective adenovirus and the other based on a replication-competent vesicular stomatitis virus ( VSV ) , were shown to provide complete protection when administered as a single injection vaccine [7]–[9] . Most intriguingly , the VSV-based vaccine is the only vaccine which has shown any utility when administered as a postexposure treatment [13] , [14] . Of these two leading EBOV vaccine candidates that can confer protection as single injection vaccines each has advantages and disadvantages . Adenovirus vectors are highly immunogenic as documented by clinical trials evaluating gene transfer efficacy and immune responses . Because they are replication-defective adenovirus vectors are also perceived to be safer for human use than a replication-competent vaccine . The most significant challenge for the adenovirus-based vaccines is the concern that a significant portion of the global population has pre-existing antibodies against the adenovirus vector which may affect efficacy [15]–[17] and has performed poorly as a vaccine vector in recent clinical trials [18]–[19] . In contrast , pre-existing immunity against VSV in human populations is negligible [20] and efficacy is likely greater with replication-competent vectors . The main concern with the VSV vaccine vector is that replication-competent vectors may present more significant safety challenges in humans particularly those with altered immune status . Because EBOV outbreaks in man have occurred exclusively in Central and Western Africa , the populations in this region are among those that may benefit from the development and availability of an EBOV vaccine . However , populations in this region are among the most medically disadvantaged in the world . In particular , the prevalence of individuals with a compromised immune system is high and HIV infections rates range up to 10% or more in this area [21] . While the VSV vaccine vector has been enormously successful in protecting healthy immunocompetent animals against EBOV [7] , [13] , [14] , we are uncertain as to how these vectors would behave in individuals with altered or compromised immune systems . Therefore , we conducted a study to assess the pathogenicity and protective efficacy of the recombinant VSV-based ZEBOV vaccine vector in rhesus macaques that were infected with simian-human immunodeficiency virus ( SHIV ) which is known to deplete the populations of naive CD4+ T cells , naive CD8+ T cells , and memory CD4+ T cells in these animals [22] , [23] . In order to take into account the degree or severity of compromised immune function animals were selected with varying degrees of CD4+ T cell loss .
The recombinant VSV expressing the glycoprotein ( GP ) of ZEBOV ( strain Mayinga ) ( VSVΔG/ZEBOVGP ) was generated as described recently using the infectious clone for the VSV , Indiana serotype [24] . ZEBOV ( strain Kikwit ) was isolated from a patient of the ZEBOV outbreak in Kikwit in 1995 [25] . Nine filovirus-seronegative adult rhesus macaques ( Macaca mulatta ) ( 5–10 kg ) were used for these studies . The macaques were infected three months prior to the current study with SHIV162p3 ( kindly provided by Dr . Ranajit Pal , Advanced BioScience Laboratories , Inc . , Kensington , MD ) . These animals all had clinical laboratory evidence of SHIV infection as evidenced by reduced CD4+ T cell counts , decreased ratios of CD4+/CD8+ T cells ( Table 1 ) and the presence of SHIV in plasma of four out of nine animals ( Table 2 ) . Six of the nine SHIV-infected animals were vaccinated by i . m . injection with ∼1×10∧7 recombinant VSVΔG/ZEBOVGP . Three animals served as placebo controls and were injected in parallel with saline . All six VSVΔG/ZEBOVGP-vaccinated animals and two of the three control animals were challenged 31 days after the single dose immunization with 1000 pfu of ZEBOV ( strain Kikwit ) . The monkeys were challenged with the heterologous Kikwit strain of ZEBOV as our macaque models have been developed and characterized using this strain [1] , [26] . Animals were closely monitored for evidence of clinical illness ( e . g . , temperature , weight loss , changes in complete blood count , and blood chemistry ) during both the vaccination and ZEBOV challenge portions of the study . In addition , VSVΔG/ZEBOV and ZEBOV viremia and shedding were analyzed after vaccination and challenge , respectively . Animals were given physical exams and blood and swabs ( nasal , oral , rectal ) were collected at 2 , 4 , 7 , 10 , 14 , 21 , 28 , and 31 days after vaccination and on days 3 , 6 , 10 , 15 , and 28 after ZEBOV challenge . The vaccination portion of the study was conducted at BIOQUAL and was approved by NIAID , BIOQUAL , and USAMRIID Laboratory Animal Care and Use Committees . The ZEBOV challenge was performed in BSL-4 biocontainment at USAMRIID and was approved by the USAMRIID Laboratory Animal Use Committee . Animal research was conducted in compliance with the Animal Welfare Act and other Federal statues and regulations relating to animals and experiments involving animals and adheres to the principles stated in the Guide for the Care and Use of Laboratory Animals , National Research Council , 1996 . Both facilities used are fully accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International . Total white blood cell counts , white blood cell differentials , red blood cell counts , platelet counts , hematocrit values , total hemoglobin , mean cell volume , mean corpuscular volume , and mean corpuscular hemoglobin concentration were determined from blood samples collected in tubes containing EDTA , by using a laser-based hematologic Analyzer ( Coulter Electronics , Hialeah , FL , USA ) . The white blood cell differentials were performed manually on Wright-stained blood smears . Serum samples were tested for concentrations of albumin ( ALB ) , amylase ( AMY ) , alanine aminotransferase ( ALT ) , aspartate aminotransferase ( AST ) , alkaline phosphatase ( ALP ) , gamma-glutamyltransferase ( GGT ) , glucose ( GLU ) , cholesterol ( CHOL ) , total protein ( TP ) , total bilirubin ( TBIL ) , blood urea nitrogen ( BUN ) , and creatinine ( CRE ) by using a Piccolo Point-Of-Care Blood Analyzer ( Abaxis , Sunnyvale , CA , USA ) . 100 ul of whole blood was added to a 12×75 tube and incubate with the antibodies for 15 minutes at room temperature . The samples was then lysed and fixed in 1% paraformaldehyde and washed three times in PBS . Samples were analyzed on a Becton Dickinson FACS Calibur ( Becton Dickinson , San Jose , CA ) . All antibodies were purchased from Becton Dickinson; clones used were CD3 – SP34 , CD4 – L200 , CD8 – RPA-T8 and CD20 – 2H7 . For measurement of plasma SIV RNA levels , a quantitative TaqMan RNA reverse transcription-PCR ( RT-PCR ) assay ( Applied Biosystems , Foster City , CA ) was used , which targets a conserved region of SIV gag and has an accurate detection limit as low as 200 RNA copies/ml . Briefly , isolated plasma viral RNA was used to generate cDNA using One-Step RT-PCR Master Mix ( Applied Biosystems ) . The samples were then amplified as previously described [27] with the following PCR primer/probes: SIV-F 5′ AGTATGGGCAGCAAATGAAT 3′ ( forward primer ) , SIV-R 5′TTC TCTTCTGCGTG AATGC 3′ ( reverse primer ) , SIV-P 6FAM-AGATTTGGATTAGCAGAAAGCCTGTTG GA-TAMRA ( TaqMan probe ) in a 7700 Sequence Detection System ( 40 cycles of 95°C , 15 seconds , and 60°C , 1 minute ) . The signal was then compared to a standard curve of known concentrations to determine the viral copies present in each sample . The assay lower limit was 40 copies/ml . RNA was isolated from blood and swabs using Tripure Reagent ( INVITROGEN , Grand Island , New York ) . For the detection of VSV we used a Q-RT-PCR assay targeting the matrix gene ( nt position 2497–2556 , AM690337 ) . ZEBOV RNA was detected using a Q-RT-PCR assay targeting the L gene ( nt position 13874–13933 , AY354458 ) . The low detection limit for this ZEBOV assay is 0 . 1 pfu/ml of plasma . Virus titration was performed by plaque assay on Vero E6 cells from all blood and selected organ ( liver , spleen , lung , kidney , adrenal gland , pancreas , axillary lymph node , inguinal lymph node , mesenteric lymph node , ovary or testis , and brain ) and swab samples . Briefly , increasing 10-fold dilutions of the samples were adsorbed to Vero E6 monolayers in duplicate wells ( 0 . 2 ml per well ) ; thus , the limit for detection was 25 pfu/ml . IgG antibodies against ZEBOV were detected with an Enzyme-Linked Immunosorbent Assay ( ELISA ) using purified virus particles as an antigen source as previously described [7] , [9] . Necropsies were performed on each animal and selected tissues were collected for histological analysis . Histology and immunohistochemistry were performed as previously described for ZEBOV-infected monkeys [26] .
We employed nine SHIV-infected rhesus macaques , of which six animals were vaccinated by i . m . injection with a single dose of VSVΔG/ZEBOVGP ( Subjects #1–6 ) and the remaining three animals ( Controls #1–3 ) received sterile saline . The animals were monitored closely for clinical symptoms and shedding of recombinant VSVs . None of the animals vaccinated with VSVΔG/ZEBOVGP or treated with saline showed overt fever or any evidence of clinical illness during the 31 day vaccination period . Importantly , no evidence of reaction at the vaccine injection site was noted among any of the VSVΔG/ZEBOVGP-vaccinated animals nor was any change noted in activity or behavior during the vaccination phase of the study ( day 0 to day 31 after vaccination ) . In addition , no changes were detected in hematology or clinical chemistry following vaccination . A mild VSVΔG/ZEBOVGP viremia ( <103 pfu/ml ) was detected only on day 2 after vaccination by virus isolation ( Figure 1 ) and RT-PCR ( data not shown ) in four of the six VSVΔG/ZEBOVGP-immunized macaques ( Subjects #1 , 2 , 3 , 4 ) . Surprisingly , the two animals with the lowest CD4+ counts ( subjects #5 , 6 ) never showed any detectable level of VSV viremia . VSVΔG/ZEBOVGP was undetectable in all analyzed swab samples ( data not shown ) . Thus , vaccination led to a transient viremia from virus replication at as yet undetermined sites but no virus shedding of the vaccine virus . Following successful completion of the safety portion of the study all six of the VSVΔG/ZEBOVGP-vaccinated SHIV-infected monkeys and two of the three placebo control SHIV-infected monkeys were challenged 31 days after the single immunization by i . m . injection with 1000 pfu of ZEBOV ( strain Kikwit ) . Four of the six VSVΔG/ZEBOVGP-vaccinated SHIV-infected monkeys and both of the placebo control animals started to show clinical signs of disease on day 6 after challenge including fever ( Subject # 1 , 2 and Control #1 , 2 ) and lymphopenia and thrombocytopenia ( Subject #2 , 5 , 6 and Control #1 , 2 ) ( Table 3 ) . Disease progressed in two of the VSVΔG/ZEBOVGP-vaccinated SHIV-infected monkeys ( Subject #5 and 6 ) and both of the placebo control animals with the development of additional evidence of clinical illness including increased levels of serum enzymes associated with liver function , depression , anorexia , and the appearance of macular rashes ( Table 3 ) . All four of these animals succumbed to the ZEBOV challenge with the two VSVΔG/ZEBOVGP-vaccinated monkeys expiring on days 9 ( Subject #6 ) and 13 ( Subject #5 ) and the placebo controls succumbing on days 9 ( Control #1 ) and 10 ( Control #2 ) after ZEBOV challenge ( Figure 2 ) . Disease did not progress in the two VSVΔG/ZEBOVGP-vaccinated SHIV-infected monkeys that were febrile ( Subjects #1 , 2 ) and had changes in hematology values on day 6 ( Subjects #2 ) and both of these animals remained healthy and survived the ZEBOV challenge ( Figure 2 ) . The remaining VSVΔG/ZEBOVGP-vaccinated macaques ( Subject #3 , 4 ) never showed any evidence of clinical illness and survived ( Figure 2 ) . Interestingly , the VSVΔG/ZEBOVGP-vaccinated macaques that succumbed were the two animals with the most significant reduction in CD4+ T cells ( 84% , 96% ) ( Table 1 ) , the lowest total CD4+ T cell counts ( 83 , 42 ) ( Table 1 ) , the highest SHIV viremia ( Table 2 ) , and no evidence for VSV viremia ( Figure 1 ) suggesting that CD4+ T cells may play a role in protection . Blood samples were analyzed after challenge for evidence of ZEBOV replication by plaque assay and RT-PCR . By day 6 , both of the placebo control animals developed high ZEBOV titers in plasma as detected by plaque assay ( >104 . 5 log pfu/ml ) ( Table 4 ) . In comparison , only one of the VSVΔG/ZEBOVGP-vaccinated monkeys ( Subject #6 ) showed a ZEBOV viremia at day 6 by plaque assay ( ∼102 log pfu/ml ) ( Table 4 ) . ZEBOV was detected in a second VSVΔG/ZEBOVGP-vaccinated monkey ( Subject #5 ) by day 10 ( ∼104 . 2 log pfu/ml ) . RT-PCR was more sensitive and showed evidence of ZEBOV in plasma of this animal ( Subject #5 ) at day 6 . In addition , RT-PCR was more sensitive in detecting ZEBOV in swabs which were positive on a number of samples derived from Subject #5 at day 6 and day 10 ( Table 4 ) . In contrast , no ZEBOV was detected in the plasma by virus isolation or RT-PCR in the four VSVΔG/ZEBOVGP-vaccinated monkeys that survived ZEBOV challenge . Moreover , no evidence for reactivation of VSVΔG/ZEBOVGP was detected from any blood or swab sample from any animal after ZEBOV challenge ( data not shown ) . Although we failed to detect ZEBOV viremia in the two surviving animals that were clinically ill ( Subject #1 and 2 ) at days 3 , 6 , 10 , and 14 after ZEBOV challenge we cannot exclude the possibility that these animals had low levels of circulating ZEBOV at time points not evaluated . The four surviving VSVΔG/ZEBOVGP-vaccinated macaques ( Subjects #1 , 2 , 3 , 4 ) were euthanized 28 days after the ZEBOV challenge to perform a virological and pathological examination of tissues . Organ infectivity titration from these four animals showed no evidence of ZEBOV in any of the tissues examined . In comparison , ZEBOV was recovered from tissues of both VSVΔG/ZEBOVGP-vaccinated animals that succumbed ( Subject #5 , 6 ) and both SHIV-infected control animals . Organ titers of infectious ZEBOV were consistent with values previously reported for immunocompetent ZEBOV-infected rhesus macaques [27] , [28] . VSVΔG/ZEBOVGP was not recovered in any of the tissues examined from any animal on this study . Pathological and immunohistochemical evaluation of tissues from the four VSVΔG/ZEBOVGP-vaccinated animals ( Subjects #1 , 2 , 3 , 4 ) that survived ZEBOV challenge showed no evidence of ZEBOV antigen . In contrast , ZEBOV antigen was readily detected in typical target organs ( e . g . , liver , spleen , adrenal gland , lymph nodes ) of the two VSVΔG/ZEBOVGP-vaccinated animals that succumbed to ZEBOV challenge ( Subject #5 , 6 ) ( Figure 3 ) and the two placebo controls . Lesions and distribution of ZEBOV antigen in these macaques was consistent with results reported in other studies [27] , [29] . While cellular immune responses against ZEBOV GP in macaques vaccinated with VSVΔG/ZEBOVGP vectors have been difficult to detect before challenge in previous studies [7] , humoral immune responses have been more robust and consistent ( [7]; TW Geisbert , unpublished observations ) . Therefore , we measured the antibody responses of the rhesus macaques vaccinated with VSVΔG/ZEBOVGP before vaccination ( day −7 ) , after vaccination ( day 14 and day 31 ) , and after ZEBOV challenge ( day 46 and day 59 after vaccination ) by IgG ELISA . None of the six VSVΔG/ZEBOVGP-vaccinated macaques developed IgG antibody titers against the ZEBOV GP by the day of ZEBOV challenge ( Figure 4 ) . Two animals ( Subjects #1 , 2 ) developed modest IgG antibody titers against ZEBOV by day 15 after ZEBOV challenge ( day 46 after vaccination ) while a third animal developed a titer by day 28 after ZEBOV challenge ( day 59 after vaccination ) ( Figure 4 ) .
An often raised concern regarding the use of the recombinant VSV vaccine platform in humans is related to the fact that this is a replication-competent vaccine , and thus demonstration of safety is of paramount importance . Taking into account our previous work it is not surprising that the VSVΔG/ZEBOVGP was tolerated well in our SHIV-infected macaques . Specifically , we failed to observe evidence of any adverse events in a large cohort of over 90 macaques receiving VSV vectors expressing different GPs from viral HF agents ( 38 cynomolgus macaques and 3 rhesus macaques vaccinated with VSVΔG/ZEBOVGP; 12 cynomolgus macaques and 3 rhesus macaques vaccinated with VSV expressing SEBOV GP; 29 cynomolgus macaques and 3 rhesus macaques vaccinated with VSV expressing the Marburg virus GP; and 6 cynomolgus macaques vaccinated with VSV expressing the Lassa GP ) ( [7] , [30] , [31]; TW Geisbert , H Feldmann , and SM Jones unpublished observations ) . We have also failed to observe any adverse events in a variety of immunocompetent laboratory mice ( different inbred strains ) , outbred guinea pigs ( Hartley strain ) and goats vaccinated with the above mentioned VSV vectors at doses ranging from 2×100–2×105 pfu ( [24] , [32]; SM Jones and H Feldmann , unpublished observations ) . More recently we have also demonstrated that vaccination of severely immunocompromised SCID mice with 2×105 pfu of the VSV-based ZEBOV vaccine ( VSVΔG/ZEBOVGP ) resulted in no clinical symptoms [32] . While transient VSV viremia in this study was only observed in surviving macaques but not in animals that had succumbed to ZEBOV challenge ( Figure 1 ) , viremia data from previous studies [7] , [30] , [31] do not support any correlation between VSV viremia and survival . In addition , no evidence for vaccine vector shedding was detected in this study supporting previous results [7] , [30] , [31] with no compelling evidence to suggest that occasional virus shedding ( only detected by RT-PCR; negative on virus isolation ) would lead to vaccine vector transmission . The VSV glycoprotein exchange vector that we employed in this study has also shown promise as a preventive vaccine and postexposure treatment against Marburg HF [30] , [33] and as a preventive vaccine against Lassa fever in nonhuman primates [31] . Similar recombinant VSV vectors have been evaluated in animal models as vaccine candidates for a number of viruses that cause disease in humans including HIV-1 , influenza virus , respiratory syncytial virus , measles virus , herpes simplex virus type 2 , hepatitis C virus , and severe acute respiratory syndrome coronavirus [34]–[40] . Many of these studies have employed VSV vectors that maintained either the entire VSV glycoprotein ( G ) or the transmembrane and/or cytoplamic domains of this protein to facilitate more efficient incorporation of the foreign antigen . It is known that VSV G is an important VSV protein associated with pathogenicity [38] , [41] . It has been shown that truncation of the cytoplasmic tail has greatly reduced vector pathogenicity in mice following intranasal inoculation indicating the importance of this domain for pathogenicity [42] . In this regard , a VSV vector including portions of the VSV G and expressing HIV genes was found to be insufficiently attenuated for clinical evaluation when assessed for neurovirulence in nonhuman primates [43] . These investigators subsequently showed that safety and immunogenicity can be improved by genetic manipulation of the VSV genome but it remained unclear whether neurovirulence was associated with the VSV G or other genome manipulations [44] . Nevertheless , our ZEBOV vaccine is a G-deficient VSV vector [24] and thus lacks G-associated pathogenicity [41] as well as the target for VSV-specific neutralizing antibodies [45] . Aside from G , the VSV matrix ( M ) protein has been associated with cytopathic effects in vitro including the inhibition of host gene expression , induction of cell rounding and induction of apoptosis [46] , [47] . It is largely unclear to what extent M alone contributes to pathogenicity , but inoculation studies with the VSV-based vaccines in different animal species ( as described above ) do not suggest a major pathogenic effect of the M protein in vivo [7] , [13] , [32] . Currently , the mechanism by which any filovirus vaccine confers protection in nonhuman primates is not well understood . Nearly all studies have detected modest to good humoral immune responses . For the VSVΔG/ZEBOVGP vaccine a humoral response is detected in macaques by day 14 after vaccination ( [7]; TW Geisbert , unpublished observations ) . However , in the current study and consistent with an impaired immune system , our SHIV-infected macaques did not develop a humoral immune response by the time of ZEBOV challenge . Three animals developed modest anti-ZEBOV IgG titers 14 to 28 days after ZEBOV challenge . We are uncertain as to why four of the six VSVΔG/ZEBOVGP-vaccinated macaques survived ZEBOV challenge . Regardless of any humoral immune response elicited in these animals it is unlikely that antibody alone confers protection . Specifically , passive antibody studies in nonhuman primates using a variety of anti-ZEBOV immune reagents including polyclonal equine immune globulin [25] , a recombinant human monoclonal antibody [48] , and convalescent monkey blood [49] have uniformly failed to provide protection and more importantly have failed to provide any beneficial effect . A number of studies have evaluated the cellular immune response in nonhuman primates vaccinated against EBOV and the results have been mixed with some studies showing a modest cellular response and other studies showing weak and/or no cellular immune responses [7] , [9] , [10] . However , it is likely that the intracellular cytokine assays that have been employed in some of these studies are not sensitive or thorough enough to detect a cellular immune response against ZEBOV . Indeed , it has been reported that the inability to demonstrate a robust cellular response may illustrate the limitation of the evaluation of cellular immune responses using small numbers of functional measurements ( such as interferon-gamma ) [50] . One interesting finding in the current study may begin to shed some light on the mechanism of protection elicited by the VSVΔG/ZEBOVGP . Notably , the two rhesus macaques that grouped together with the most severe loss of CD4+ T cells were the only animals that failed to survive ZEBOV challenge . This suggests that CD4+ T cells may play a role in mediating protective immunity in EBOV infections . CD4+ T cells have been shown to be depleted in nonhuman primate following ZEBOV infections [27] , [51] and in vitro ZEBOV infection of human peripheral blood mononuclear cells causes massive bystander death of CD4+ T cells by apoptosis [52] . While rodents do not appear to faithfully reproduce ZEBOV infection of humans and nonhuman primates [53] studies have suggested that CD4+ T cells are required for protection of rodents against ZEBOV . Specifically , in a study using liposome-encapsulated ZEBOV antigens , Rao and colleagues showed that treatment of mice with anti-CD4 antibodies before or during vaccination abolished protection , while treatment with anti-CD8 antibodies had no effect , thus indicating a requirement for CD4+ T lymphocytes for successful immunization [54] . Similarly , depletion of CD8+ T cells did not compromise protection in mice indicating that CD8+ cytotoxic T cells are not a requirement for protection [32] . In conclusion , our results show that the VSV-based ZEBOV vaccine ( VSVΔG/ZEBOVGP ) did not cause any illness in immunocompromised SHIV-infected rhesus macaques and resulted in sufficient protective efficacy in all but the most severely compromised animals against a lethal ZEBOV challenge . Protection in the immunocompromised macaques appeared to be dependent on CD4+ T cells rather than the development of EBOV-specific antibodies . This provides strong support for the safety of the VSV-based vectors and further development of this promising vaccine platform for its use in humans . While these data are very encouraging , as the number of SHIV-infected macaques in the current study was small , additional safety studies will be needed in order to determine whether vaccines based on attenuated VSV will ultimately prove safe in immunocompromised humans .
|
Ebola virus is among the most lethal microbes known to man , with case fatality rates often exceeding 80% . Since its discovery in 1976 , outbreaks have been sporadic and geographically restricted , primarily to areas of Central Africa . However , concern about the natural or unnatural introduction of Ebola outside of the endemic areas has dramatically increased both research interest and public awareness . A number of candidate vaccines have been developed to combat Ebola virus , and these vaccines have shown varying degrees of success in nonhuman primate models . Safety is a significant concern for any vaccine and in particular for vaccines that replicate in the host . Here , we evaluated the safety of our replication-competent vesicular stomatitus virus ( VSV ) -based Ebola vaccine in SHIV-infected rhesus monkeys . We found that the vaccine caused no evidence of overt illness in any of these immunocompromised animals . We also demonstrated that this vaccine partially protected the SHIV-infected monkeys against a lethal Ebola challenge and that there appears to be an association with levels of CD4+ lymphocytes and survival . Our study suggests that the VSV-based Ebola vaccine will be safe in immunocompromised populations and supports further study and development of this promising vaccine platform for its use in humans .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"virology/vaccines",
"virology/animal",
"models",
"of",
"infection",
"virology"
] |
2008
|
Vesicular Stomatitis Virus-Based Ebola Vaccine Is Well-Tolerated and Protects Immunocompromised Nonhuman Primates
|
GATA transcription factors are highly conserved among eukaryotes and play roles in transcription of genes implicated in cancer progression and hematopoiesis . However , although their consensus binding sites have been well defined in vitro , the in vivo selectivity for recognition by GATA factors remains poorly characterized . Using ChIP-Seq , we identified the Dal80 GATA factor targets in yeast . Our data reveal Dal80 binding to a large set of promoters , sometimes independently of GATA sites , correlating with nitrogen- and/or Dal80-sensitive gene expression . Strikingly , Dal80 was also detected across the body of promoter-bound genes , correlating with high expression . Mechanistic single-gene experiments showed that Dal80 spreading across gene bodies requires active transcription . Consistently , Dal80 co-immunoprecipitated with the initiating and post-initiation forms of RNA Polymerase II . Our work suggests that GATA factors could play dual , synergistic roles during transcription initiation and post-initiation steps , promoting efficient remodeling of the gene expression program in response to environmental changes .
In eukaryotes , gene transcription by RNA polymerase II ( Pol II ) is initiated by the binding of specific transcription factors to double-stranded DNA . The yeast transcription factors target regulatory regions called UAS or URS ( for Upstream Activating/Repressing Sequences ) , generally directly adjacent to the core promoter . The generated regulatory signals converge at the core promoter where they permit the regulation of Pol II recruitment via the ‘TATA box-binding protein’ and associated general transcription factors [1 , 2] . The transcription factor binding sites are usually short sequences ranging from 8 to 20 bp [3] . They are most often similar but generally not identical , differing by some nucleotides from one another [3] , making it sometimes difficult to predict whether a given UAS will function as such in vivo . GATA factors constitute a family of transcription factors highly conserved among eukaryotes and characterized by the presence of one or two DNA binding domains which consists of four cysteines ( fitting the consensus sequence CX2CX17-18CX2C ) coordinating a zinc ion followed by a basic carboxy-terminal tail [4] . While vertebrate GATA factors possess two adjacent homologous zinc fingers , fungal ones contain only one single zinc finger , being most closely related to the C-terminal vertebrate zinc finger [5 , 6] , which is the one responsible for determining the binding specificity of GATA-1 , the founding member of the GATA factor family [7] . The specificity of GATA factor binding has been thoroughly characterized in yeast [8–10] and metazoans [11–18] . In addition , structure determinations of protein-DNA complexes , first for GATA-1 [4] , then for its fungal orthologue AreA [19] , allowed for the identification of the subtle determinants of DNA specificity for GATA factors . Notably , the conserved DNA binding domain of GATA factors was reported to bind to consensus sequences ( corresponding to GATAA ( G ) or GATTAG for the yeast GATA factors described hereafter ) , as shown in various organisms using direct or indirect methods [4 , 19–22] . These consensus sequences are accordingly referred to as GATA motifs . Since its discovery 40 years ago in chicken cells , the family of GATA factors was extended in human cells and represents master regulators of hematopoiesis and cancer [23] . However , although approximately 7 million GATA motifs can be found in the human genome , the GATA factors occupy only 0 . 1–1% of them . Conversely , other regions are occupied by GATA factors despite lacking the consensus motif [24 , 25] . Consistently , even if most GATA factors bind to core GATA sequences , peculiar specificities have been reported for the flanking bases as well as for the fourth base of the GATA core element [26–29] . These studies revealed an elevated flexibility in the recognition sites for vertebrate and fungal GATA factors , much greater than previously anticipated , making the search for GATA sites and their enrichment in GATA-regulated genes tedious and unproductive . In addition , GATA factors can swap among them for the same motif and switch from active or repressive transcriptional activity . All these observations developed the main paradigm shift of how GATA factors are recruited and reside on the chromatin [30 , 31] . In yeast , the family of GATA transcription factors contains over 10 members [32] . Four of them are implicated in the regulation of Nitrogen Catabolite Repression ( NCR ) -sensitive genes , the expression of which is repressed in the presence of a preferred nitrogen source ( glutamine , asparagine , ammonia ) and derepressed when only poor nitrogen sources ( e . g . proline , leucine , urea ) are available [10] . The key GATA factors involved in NCR signaling are two activators ( Gln3 and Gat1/Nil1 ) and two repressors ( Gzf3/Nil2/Deh1 and Dal80/Uga43 ) [33–38] . In a perfect feedback loop , the expression of DAL80 and GAT1 is also NCR-sensitive , which implies cross- and autogenous regulations of the GATA factors in the NCR mechanisms [38–41] . Under nitrogen limitation , expression of DAL80 is highly induced [35] , and Dal80 enters the nucleus where it competes with the two GATA activators for the same binding sites [20 , 39 , 42] . Although initially described as being active under nitrogen abundance [37 , 38] , the Gzf3 repressor also localizes to NCR-sensitive promoters in conditions of activation [40] . The sequence conservation among the four yeast NCR GATA factors is remarkable and the residues involved in contacts with the DNA , thus specificity determination , are 100% conserved . In this respect , the binding sites of Dal80 on target DNA are likely to be recognized also by Gln3 , Gat1 and Gzf3 [28] . In vitro , the Gln3 and Gat1 activators bind to single GATA sequences , presumably as monomers [43] , like their orthologous vertebrate counterparts , while Dal80 was found to bind to two GATA sequences , 15–35 bp apart , in a preferred tail-to-tail orientation or to a lower extent in a head-to-tail configuration [9 , 20 , 39 , 44] . In vivo , GATA factor binding site recognition also appears to require repeated GATA motifs within promoters , as shown for the NCR-sensitive DAL5 promoter [45–47] . This led to the actual fuzzy definition of UASNTR , consisting in two GATA sites located close to one another to present a binding platform for GATA factors [45–47] . Finally , in some cases , the existence of auxiliary promoter sequences was shown to compensate single GATA site , allowing for transcriptional activation [48] , although this was never as efficient as additional GATA sites [49] . The antagonistic role of Dal80 also requires multiple GATA sites [39 , 42] , and inactivation of one of the four GATA sites of the UGA4 promoter results in the loss of the Dal80-repressive activity while affecting moderately Gln3- and Gat1- activation capacity [20] . In summary , although NCR-sensitive genes are recognized to contain at least one GATA site , and often more , a precise definition of the minimal element required for binding and transcriptional regulation is still lacking . In yeast , genome-wide ChIP analyses have allowed gaining insights into the GATA factor gene network through the identification of direct targets [50–53] . However , these studies were not performed in activating conditions , when all GATA factors are expressed , localized in the nucleus and active , so that the current list of GATA factor targets are likely to be underestimated . On another hand , bioinformatic analyses have shown that , since GATA sequences are short , they can be found almost everywhere throughout the genome . Therefore , based on the sole criteria of the presence of repeated GATA sequences in yeast promoters , a third of the yeast genes could hypothetically be NCR regulator targets [54] . However , such GATA motif repetitions have been found in the promoter of 91 genes , inducible by GATA activators in absence of a good nitrogen source , supposed to be directly targeted by the GATA activators [55] . Nevertheless , the functionality of these hypothetical UAS still needs to be directly demonstrated in vivo [1] . Here , we provide the first genome-wide identification of Dal80 targets in yeast , in physiological conditions where Dal80 is fully expressed and active . Using a ChIP-Seq approach combined to a bioinformatic peak-calling procedure , we defined the exhaustive set of Dal80-bound promoters , which turned out to be much larger than anticipated . Our data indicate that at some promoters , Dal80 recruitment occurs independently of GATA sites . Strikingly , Dal80 was also detected across the body of a subset of genes bound at the promoter , globally correlating with high and Dal80-sensitive expression . Mechanistic single-gene experiments confirmed the Dal80 binding profiles , further indicating that Dal80 spreading across gene bodies requires active transcription . Finally , co-immunoprecipitation experiments revealed that Dal80 physically interacts with active form of Pol II .
In order to determine the genome-wide occupancy of a GATA factor in yeast , our rationale was to choose Dal80 as it is known to be highly expressed in derepressing conditions and forms chromosome foci when tagged by GFP [56] . We grew yeast cells in proline-containing medium and performed a ChIP-Seq analysis using a Dal80-Myc13-tagged strain and the isogenic untagged strain , as a control ( Fig 1A ) , after ensuring that the Myc13-tagged form of Dal80 was functional ( S1A Fig ) . Dal80-bound regions were then identified using a peak-calling algorithm ( see Material & Methods ) . A promoter was defined as bound by Dal80 on the basis of a >75% overlap of the -100 to -350 region ( relative to the downstream ORF start site ) by a peak ( Fig 1B ) . We chose to use as the reference coordinate the translation initiation codon rather than the transcription start site ( TSS ) since the latter has not been accurately defined for all genes . Then , our arbitrary definition of the promoter as the -350 to -100 region relative to the ATG codon was based on the distribution of the TSS-ATG distance for genes with an annotated TSS ( median and average distance = 58 and 107 bp , respectively; see S1B Fig ) . Strikingly , Dal80 was found to bind to 1269 gene promoters ( Fig 1C and 1D and S1 Table ) . This number , corresponding to 22% of all protein-coding gene promoters , is much higher than anticipated given the roughly hundred target genes generally cited for the GATA transcriptional activators Gat1 and Gln3 [55 , 57] , presumably sharing binding sites with Dal80 . However , we noted that some peaks ( 221 ) overlapped several promoters ( 471 ) , mainly of divergent genes ( 442 ) , as shown in Fig 1E for an illustrative example . Despite it is possible that in such cases , only one of the two divergent promoters is targeted by Dal80 , the number of in vivo Dal80 target sites we identified here has been extensively extended from what was acknowledged so far . Among the genes showing Dal80 binding at their promoter , we noticed a significant enrichment for cytoplasmic translation genes , as well as genes involved in small molecule biosyntheses , including amino acids ( S2 Table ) . Before our work , very few studies have investigated the transcriptional targets of Dal80 in vivo in conditions of nitrogen deprivation . One of them , based on mini-arrays [58] , identified 19 Dal80-regulated genes , all of which have been isolated in our ChIP-Seq analysis ( highlighted in orange in column B of S3 Table ) . As expected given the similarity between binding sites of Dal80 and the other nitrogen-regulated GATA factors , other genes related to previous nitrogen regulation screens [55 , 57–64] are also significantly enriched within our list: 103 of the 205 previously identified nitrogen-regulated genes have been identified in our ChIP-Seq analysis using Dal80 as the bait , which is much more than expected by chance ( P<0 . 001 , Chi-square test; S3 Table , column B ) . Surprisingly , analysis of GATA site occurrence over Dal80-bound and unbound promoters revealed no difference between the two classes , 48 . 2% and 51 . 3% of Dal80-bound and unbound promoters containing at least two GATA sites , respectively ( Fig 1F ) . Likewise , we observed no major difference between the Dal80-bound and unbound promoters in respect of the GATA sites spacing ( S1C Fig ) and orientation ( S1D Fig ) preferences defined in vitro for Dal80 binding [9] . Intriguingly , 20% of Dal80-bound promoters do not contain any GATA site ( Fig 1F ) , indicating that Dal80 recruitment can also occur independently of the presence of consensus GATA sites ( see S1B Fig for visualization of Dal80 recruitment to a GATA-less promoter ) . In summary , our ChIP-Seq analysis revealed that Dal80 binds to a set of promoters larger than previously expected , targeting biosynthetic functions and protein synthesis in addition to nitrogen catabolite repression . We asked whether Dal80-binding to promoters could be associated to regulation of gene expression by the nitrogen source and/or Dal80 . We therefore performed RNA-seq in wild-type cells grown in glutamine- and proline-containing medium , and in dal80Δ cells grown in proline-containing medium . Firstly , we identified 1682 ( 30% ) genes differentially expressed ( fold-change ≥2 or ≤0 . 5 , P ≤0 . 01 ) in wild-type cells according to the nitrogen source provided ( Fig 2A ) , including 754 genes upregulated ( NCR-sensitive ) and 928 downregulated ( revNCR-sensitive ) in proline-containing medium ( see lists in S4 Table ) . Consistent with previous reports , DAL80 was found in our set of NCR-sensitive genes ( S4 Table ) , showing very low expression in glutamine-containing medium and strong derepression in proline ( S2A Fig ) . More globally , 97 of the 205 genes previously identified as NCR-sensitive were also found in our list ( P<0 . 0001 , Chi-square test; S4 Table ) . In parallel , we identified 546 genes showing significantly altered expression ( fold-change ≥2 or ≤0 . 5 , P ≤0 . 01 ) in proline-grown dal80Δ cells compared to wild type ( Fig 2B; S5 Table ) . In agreement with the previously described repressive activity of Dal80 [35] , 232 genes are indeed negatively regulated by Dal80 ( up in dal80Δ; red dots in Fig 2B ) . Unexpectedly , 314 genes are positively regulated by Dal80 ( down in dal80Δ; blue dots in Fig 2B ) . This is the first in vivo global indication suggesting a positive function for Dal80 in gene expression . The Dal80-repressed group was enriched for genes involved in small molecule catabolic processes ( S6 Table ) , while the Dal80-activated genes were mostly involved in amino acid biosynthesis ( S7 Table ) . Again , we noticed an overlap between Dal80-regulated genes and nitrogen regulated genes that were identified in other screens: 86 of the 205 previously identified nitrogen-regulated genes have been identified as Dal80-regulated , which is much more than expected by chance ( P<0 . 0001 , Chi-square test; column D of S3 Table ) . Globally , we observed a significant correlation between Dal80-sensivity and regulation by the nitrogen source ( P<0 . 00001 , Chi-square test; Fig 2C; see also S2B Fig ) . Indeed , there are more NCR-sensitive Dal80-activated and Dal80–repressed genes than expected in case of independence ( Fig 2C; see also S2B Fig ) . Similarly , the number of revNCR-sensitive Dal80-repressed genes is also significantly higher than expected by chance ( Fig 2C; see also S2B Fig ) . In contrast , the number of revNCR-sensitive Dal80-activated genes is significantly lower than expected by chance ( Fig 2C; see also S2B Fig ) , indicating a negative correlation in this case . This observation is consistent with the DAL80 gene itself being NCR-sensitive , so that the Dal80-activated genes can only be activated when DAL80 is expressed . More importantly , Dal80 recruitment to promoters significantly correlated with nitrogen- and Dal80-sensitivity . In fact , nitrogen-regulated expression and Dal80-binding are not independent , as NCR-sensitive ( 212 ) and especially revNCR-sensitive ( 325 ) genes are significantly enriched in Dal80-bound genes ( P<0 . 00001 , Chi-square test; Fig 2D; see also S2C Fig ) . We also observed a significant correlation between Dal80-sensitive gene expression and Dal80 recruitment at the promoter: 211/546 of Dal80-regulated genes were bound by Dal80 , including 120/314 Dal80-activated and 91 Dal80-repressed genes , which again is much more than expected by chance ( P<0 . 00001 , Chi-square test; Fig 2E; see also S2D Fig ) . Fig 2F shows an illustrative example of an NCR-sensitive , Dal80-activated gene ( UGA3 ) , the promoter of which is bound by Dal80 ( Fig 1E ) . S3A Fig shows the RNA-Seq signals for another NCR-sensitive , Dal80-repressed and Dal80-bound gene ( MEP2 ) , correlating with Pol II occupancy levels ( S3B Fig ) . In summary , there is a significant correlation between Dal80 recruitment to the promoter of genes and a regulation by the nitrogen source and/or Dal80 at the RNA level , indicating that Dal80 recruitment to promoters is physiologically relevant . More specifically , we identified a subset of 211 Dal80-bound genes that are regulated by Dal80 ( S3 Table ) , and that are therefore a robust class of direct Dal80 targets . The metagene analysis described above revealed that the genes bound by Dal80 at the promoter also display a signal along the gene body , although this intragenic signal remains globally lower than in the promoter-proximal region ( Fig 1D ) . This observation prompted us to investigate the possibility that Dal80 also occupies the gene body , at least for a subset of genes . We identified 189 genes showing Dal80 intragenic occupancy , according to a >75% overlap of the ORF by a Dal80-Myc13 peak ( Fig 3A and 3B ) . Among them , 144 ( 76% ) were also bound at the promoter ( Fig 3B ) . On the other hand , 45 genes showing Dal80 intragenic binding were not bound at the promoter ( Fig 3B ) . Hence , we distinguished four classes of genes ( S8 Table ) : ( i ) those bound by Dal80 at the promoter only ( “P” class; Fig 3C; S8 Table , column C ) , ( ii ) those showing both promoter and intragenic binding ( “P&O” class; Fig 3D; S8 Table , column E ) , ( iii ) those bound across the ORF only ( “O” class; Fig 3E; S8 Table , column D ) , ( iv ) the unbound genes ( Fig 3F ) . Interestingly , we noted that the global Dal80-Myc13 signal at the promoter was higher for the “P&O” class in comparison to the “P” class ( Fig 3C and 3D ) . Most of the genes of the “O” class are not Dal80-sensitive ( 40/45; S8 Table , column J ) . Furthermore , a substantial fraction of them correspond to small dubious ORFs , close to or even overlapping an adjacent Dal80-bound gene promoter . In these cases , the limited resolution of the ChIP-Seq technique , combined to the small size of these genes , might have allowed them to pass the filters we used to identify Dal80 intragenic binding . Overall , these observations suggest that the existence of the “O” class is likely to be physiologically irrelevant . Therefore , this class will not be further considered in our study . In conclusion , we identified a subset of genes showing intragenic Dal80 occupancy , in most cases correlating with a strong Dal80 recruitment at the promoter . We asked whether Dal80 occupancy across gene bodies correlates with nitrogen-regulated gene expression and Dal80-sensitivity . We observed that nitrogen-regulated genes ( NCR and revNCR; Fig 4A; see also S4A Fig ) and Dal80-regulated genes ( Dal80-activated and -repressed; Fig 4B; see also S4B Fig ) were significantly more represented in the P&O class compared to the Dal80-unbound class . Strikingly , we also observed that the genes of the P&O class are more expressed than the unbound genes ( P < 2 . 2e-16 , Wilcoxon rank-sum test; Fig 4C ) but also than the P-bound genes ( P = 1 . 3e-14 , Wilcoxon rank-sum test; Fig 4C ) . However , it should be noted that a fraction of P-bound and unbound genes are expressed to higher levels than genes of the “P&O” class ( S4C and S4D Fig ) , indicating that high expression does not always imply intragenic Dal80 occupancy . Together with the observation that genes of the “P&O” class globally showed higher Dal80-Myc13 ChIP-Seq signal at the promoter than those of the “P” class ( Fig 3C and 3D ) , our results indicate that Dal80 occupancy across gene bodies correlates with a stronger recruitment at the promoter and higher expression in proline-containing medium . This raises the question of the specificity of the intragenic signal observed by ChIP-Seq . Indeed , for several proteins , unspecific ChIP signals have been detected across the body of a subset of highly expressed Pol II- and Pol III-dependent genes , referred to as ‘hyper-ChIPable’ loci [65–67] . We asked whether genes of our P&O class have been previously identified as ‘hyper-ChIPable’ ( S9 Table , column G ) . This comparison indicated that 48/1125 of the P-bound genes and 27/144 of the P&O genes match with hyper-ChIPable loci ( S4E and S4F Fig; see also S9 Table , columns H-I ) , suggesting that for a minority of cases , the intragenic Dal80 signal could be due to the ‘hyper-ChIPability’ of the locus and therefore be non-specific . However , since these ‘hyper-ChIPable’ loci were defined under growth conditions that are different from those used in our study ( growth in rich medium vs proline-containing synthetic medium ) , we aimed to get a more robust control for the specificity of Dal80 within gene bodies . Our rationale was to evaluate how similar and/or specific two close GATA factors could share/distinguish this “so called” artefactual hyper-ChIPability property . We performed a similar ChIP-Seq analysis using another GATA factor , the Gat1 activator [68] , using the same conditions and following the same experimental procedure as described above ( Figs 1A , 1B & 3A ) . Interestingly , 83 . 2% ( 936/1125 ) of the promoters bound by Dal80 were also bound by Gat1 ( S4G Fig; S9 Table , column E ) , reinforcing the accuracy of the extended list of novel GATA-bound genes in yeast . Strikingly , the proportion of common targets among the P&O class dramatically decreased , 55% ( 79/144 ) of the genes bound by Dal80 at the promoter and across the gene body also showing promoter and intragenic binding for Gat1 ( S4H Fig; S9 Table , column F ) . Importantly however , 65/144 P&O for Dal80 do not display intragenic binding for Gat1 ( S4H Fig; S9 Table , column F ) , although Gat1 is recruited to the promoter of 57 of them . Thus , we can define a subset of 57 genes showing a specific intragenic occupancy of Dal80 , while both Dal80 and Gat1 are recruited to their promoters similarly . As an illustrative striking example , Fig 4D shows a snapshot of the ChIP-Seq signals across MEP2 , a well-characterized NCR-sensitive gene , the promoter of which is bound by the two GATA factors , but only Dal80 is found within the gene body . To summarize , Dal80 occupancy across the gene body correlates with high expression levels . In a substantial proportion of cases , intragenic occupancy was found to be specific for Dal80 , as another GATA factor also recruited to the promoter in the same experimental conditions was not detected within the gene body . In order to validate our genome-wide observations and get additional mechanistic insights into the molecular bases of Dal80 occupancy across the body of highly expressed genes , we characterized the binding profile of Dal80 along the ammonium permease-coding gene MEP2 , an NCR-sensitive gene of the “P&O” class ( see Fig 4D ) . ChIP experiments followed by qPCR confirmed that Dal80 binds not only the promoter , but also across the coding region of MEP2 in proline-grown cells ( Fig 5A and 5B ) . No signal was observed in glutamine-grown cells ( Fig 5B ) , indicating that Dal80 recruitment only occurs when it is expressed ( S2A Fig ) . To determine whether Dal80 intragenic occupancy is mediated by nascent RNA binding during transcription , we performed a similar ChIP experiment on the MEP2 gene , treating the chromatin with RNase before the immunoprecipitation . Our results show no significant change of the Dal80-Myc13 signal across MEP2 upon RNAse treatment of the chromatin extracts before the immunoprecipitation ( Fig 5C ) , indicating that Dal80 occupancy across the gene body does not depend on RNA . Since genes of the Dal80 “P&O” class are globally highly expressed , we asked whether active transcription is a prerequisite for Dal80 binding across the ORF . Our strategy was to select an NCR gene for which Dal80 is bound at the promoter when repressed and then monitor Dal80 occupancy once the gene is activated . Our RNA- and ChIP-Seq data allowed us to isolate the UGA4 locus , another well-characterized NCR-sensitive gene , bound by Dal80 at the promoter ( Fig 6A; see snapshot in S5A Fig ) . UGA4 expression is induced by GABA ( γ-aminobutyric acid ) and is strongly repressed by Dal80 in the absence of the inducer [69] . To derepress UGA4 without inducer , a Dal80-specific deletion in the C-terminal leucine zipper domain was generated , impairing Dal80 repressive activity without affecting its binding capacity [34 , 44] . Indeed , in the Dal80ΔLZ-Myc13 strain ( Fig 6B ) , the steady-state level of UGA4 mRNA ( S5B Fig ) and Pol II occupancy ( S5C Fig ) both increased to derepressed levels in non-inducing conditions , like in a dal80Δ strain . Strikingly , in these conditions , full-length Dal80-Myc13 binding was restricted to the UGA4 promoter ( Fig 6A; see also S5A Fig ) , while Dal80ΔLZ-Myc13 binding was detected at the promoter and across the body of UGA4 ( Fig 6A ) . Interestingly , the leucine zipper of Dal80 and consequently , its dimerization , needed for UGA4 repression , were not required for its localization across the UGA4 gene body . Importantly , these results confirm that promoter binding is not sufficient to confer intragenic binding , but suggest that transcription activation is required . Altogether , these observations prompted the important mechanistic question of how Dal80 can be localized to gene bodies upon transcription activation . In order to test if the presence of an NCR-sensitive promoter could confer intragenic Dal80 binding across the body of a non-NCR-sensitive gene , we placed the URA3 ORF under the control of different promoters bound or not by Dal80: the MEP2 and TDH3 promoters as P&O representative , the ALD6 promoter for the P class and the VMA1 promoter , which is not bound by Dal80 ( Fig 7A ) . When driven by PMEP2 , the expression of URA3 becomes NCR-sensitive and followed wild-type MEP2 expression ( S6 Fig ) , correlating with Pol II recruitment over the URA3 ORF ( Fig 7B ) . In these conditions , we observed Dal80-Myc13 binding at the promoter of MEP2 and also across URA3 ( Fig 7C ) . Similarly for PTDH3-URA3 construct , Dal80 also was relocalized within the URA3 ORF , although to a lesser extent . Importantly , Dal80 binding was not detected across URA3 when it was expressed from its native locus , under the control of its promoter ( Fig 7C ) or under the control of the Dal80-bound PALD6 or unbound PVMA1 ( Fig 7C ) , reinforcing the idea that those promoters fail to carry sufficient information for Dal80 to occupy the URA3 ORF . Among the obvious characteristics , we noticed that Pol II occupancy is higher within those P&O URA3 genes than the P only , suggesting that transcription strength might be a key determinant for Dal80 localization across the ORF . Interestingly , among the P&O fusions ( MEP2 and TDH3 ) , we noted a difference in Dal80 binding levels to the adjacent URA3 ORF , while those of Pol II remain similar across the two coding regions , suggesting that Pol II level might not be the only factor that control Dal80 occupancy . In conclusion , these results show that for the same URA3 sequence , the Dal80 occupancy displays distinct features depending only on the promoter characteristics to be classified as P , P&O or unbound , reflecting transcriptional strength . We propose that Dal80 presence within the ORF could be attributed to a spreading mechanism , controlled by Pol II complex and Dal80-promoter recognition capacity . These results exclude strongly DNA motif ( s ) as a main determinant for Dal80 spreading into ORF but rather raise the question of the direct implication of Pol II itself . To test the hypothesis that the active Pol II complex could be responsible for Dal80 spreading beyond Dal80-bound promoters , we assessed the effect of rapid inactivation of Pol II using the thermosensitive rpb1-1 strain [70 , 71] . We analyzed Dal80-Myc13 binding along MEP2 in WT and rpb1-1 cells . When rpb1-1 cells were shifted at 37°C for 1h , MEP2 mRNA and Pol II levels showed a 2-fold ( S7A Fig ) and >10-fold decrease ( S7B Fig ) , respectively , reflecting the expected transcription shut-down when rpb1-1 cells are shifted in non-permissive conditions . In the same conditions , we observed a significant >5-fold reduction of Dal80-Myc13 levels across the MEP2 ORF , while the binding at the promoter was not affected ( Fig 8A ) . This result reinforces the idea that Dal80 spreading across the body of NCR-sensitive genes is strongly correlated to an active Pol II . To get insights into the mechanism by which Dal80 associates to actively transcribed gene bodies , we tested whether it physically interacts with the transcriptionally engaged form of Pol II ( Fig 8B ) . Total protein extracts from Dal80-Myc13 cells were immunoprecipitated with antibodies directed against the Pol II CTD and its phospho-forms Ser2P and Ser5P , respectively characteristic of elongating and initiating Pol II forms . All three antibodies enabled effective immunoprecipitation , whereas no antibody and nonspecific antibody controls generated a lower or no signal at all . Thus , Dal80 would physically interact with phosphoforms of the Pol III , suggesting a strong association with Pol II engaged in active transcription from initiating to elongating polymerase . Together , our data indicate that Dal80 spreading across the body of NCR-sensitive genes depends on active transcription and that Dal80 interacts with the transcriptionally active forms of Pol II , supporting a model where Dal80 spreading across the body of highly expressed , NCR-sensitive genes might be the result of Dal80-Pol II association at post-initiation transcription phases .
Eukaryotic GATA factors belong to an important family of DNA binding proteins involved in development and response to environmental changes in multicellular and unicellular organisms , respectively . In yeast , four GATA factors are involved in Nitrogen Catabolite Repression ( NCR ) , controlling gene expression in response to nitrogen source availability . One of them , the Dal80 repressor , itself NCR-sensitive , acts to modulate the intensity of NCR responses . Over the past decade , a number of studies have screened the genome aiming at gathering an inventory of genes regulated by the nitrogen source . Although >500 genes have been shown to be differentially expressed upon change of the nitrogen source [57 , 64] , the list of NCR-sensitive genes was reduced to about 100 , based on their sensitivity to GATA factors [55 , 57 , 60 , 63] , suggesting that the number of Dal80 targets would be situated in that range . Here , using ChIP-Seq , we identified 1269 Dal80-bound promoters , which considerably extends the list of potential Dal80 targets . In fact , the number of Dal80-bound promoters could even have been greater . Indeed , the GATA consensus binding site is rather simple and short , so that in yeast , a total number of 10 , 000 putative binding sites can be found in all protein-coding gene promoters , 2930 promoters having at least two GATA sites , which is thought to be a prerequisite for in vivo binding and function of the GATA factors . The difference between the number of promoters with ≥2 GATA sites and the number of Dal80-bound promoters suggests the existence of a selectivity for Dal80 recruitment . This selectivity could rely on promoter architecture and/or chromatin structure , conditioning the requirement for auxiliary DNA binding factors that would stabilize Dal80 at some promoters . Moreover , although we observed a significant correlation between Dal80 binding and regulation , the expression of most of the Dal80-bound genes was not affected in a dal80Δ mutant strain . Again , Dal80-dependence for transcribing these genes , as well as their NCR sensitivity , could require the presence of yet unknown cofactors which are not produced or inactive under the tested growth conditions . In mammals , GATA factors also display an extraordinary complexity in the relationships between binding and expression regulation . Like Dal80 , GATA-1 and GATA-2 only occupy a small subset of their abundant binding motif throughout the genome , and the presence of the conserved binding site is insufficient to cause GATA-dependent regulation in most instances [72] . GATA-1 binding kinetics , stoichiometry and heterogeneous complex formations , conditioned by composite promoter architecture , influence its transcriptional activity and hence diversify gene expression profiles [72] . Given the high conservation at the amino acid level between the DNA binding domains of the four yeast NCR GATA factors , it is likely that they all recognize identical sequences ( GATAA , GATAAG or GATTAG ) . This consensus has been largely validated in the past using gene reporter experiments , mutational analyses and in vitro binding experiments on naked DNA . Nonetheless , of the 1269 bound promoters , 48% contained at least two GATA sites , a proportion that is not different from that observed among unbound promoters , and the amount of GATA sites per promoter was not different between the two groups either . In addition , Dal80 recruitment was found to occur independently of the presence of GATA sites in 20% of Dal80-bound promoters , as also previously observed in mammalian cells [24 , 73] . Future experiments will be required to decipher how Dal80 can be recruited to these GATA-less promoters . Among the different possibilities is a recruitment of Dal80 by degenerated GATA motifs . In this regard , we identified 5 degenerated GATA motifs within a 70 bp window corresponding to the peak of Dal80 binding signal at the promoter of the GATA-less , Dal80-sensitive gene ALD6 ( see S1E Fig ) . However , it also has to be noted that upon tolerance of only one mismatch within the GATA consensus , multiple degenerated motifs are detected in every yeast promoter . Unexpectedly , although Dal80 has always been described as a repressor , we identified 314 genes that are positively regulated by Dal80 ( their expression is significantly decreased upon Dal80 deletion; S5 Table ) . These genes are significantly enriched in amino acid biosynthetic processes , resembling the amino acid starvation response mediated by the Gcn4 transcriptional activator . Interestingly , the promoter of 122/314 Dal80-activated genes contain Gcn4-binding sites ( S5 Table ) , and this group of 314 Dal80-activated genes is significantly enriched for genes regulated by the General Amino Acid Control ( GAAC; YeastMine Gene List , Publication Enrichment , P<1 . 6e-13 ) , through the Gcn4 activator . Interconnections between NCR and GAAC have already been demonstrated , mostly at the level of nitrogen catabolism control: 1-a large number of non-preferential nitrogen sources leads to increased transcription of GAAC targets [57]; and 2- Gcn4 contributes , with Gln3 , to the expression of some but not all NCR-sensitive genes [74 , 75] . However , this is the first time that evidence are provided indicating a positive role for Dal80 at the level biosynthetic gene expression . The most striking and unexpected finding of this work is the observation that Dal80 also occupied the body of a subset of genes . Dal80 binding at the promoter and spreading across the body of the 144 genes of the “P&O” class correlated with high expression levels and sensitivity to Dal80 . It has been previously reported that at some loci , referred to as ‘hyper-ChIPable’ , high expression levels might induce artefactual detection of DNA-binding factors across gene bodies [65] . However , in the context of this work , several observations argue for a specific association of Dal80 with gene bodies , at least for a subset of genes . Firstly , a considerable fraction of genes of the “P” class show similar or even higher expression levels than genes of the “P&O” class ( S4C and S4D Fig ) , indicating that high expression does not always induce spreading of Dal80 across the gene body . Secondly , only 27 of the genes of our “P&O” class have been previously defined as ‘hyper-ChIPable’ ( S9 Table , column I ) , even if the conclusion should be taken with caution as the two sets of experiments were performed upon very distinct physiological conditions . Thirdly , and more importantly , a similar ChIP-Seq analysis performed under the same experimental conditions using another GATA factor ( the Gat1 activator ) allowed us to define a subset of 57 genes that are specifically and only bound by Dal80 across their body , while both Dal80 and Gat1 are recruited to their promoter ( see Fig 4D and S4H Fig ) . Thus , although we cannot exclude that in few cases , the signals for Dal80 across the intragenic region could still depend on the hyper-ChIPability of the locus , we propose that for the majority of “P&O” genes , the intragenic association of Dal80 is specific and biologically relevant . This is further supported by the observation that Dal80-sensitive ( -activated and–repressed ) genes are statistically more enriched within the “P&O” class , compared to the “P” class ( Fig 4B ) . However , the causality relationship between Dal80 intragenic binding and high expression levels in derepressing conditions ( proline ) remains unclear to date . The observations we made at the genome-wide level were experimentally confirmed using ChIP experiments , at the level of single well-characterized NCR-sensitive genes . Promoter binding appears to be required but not sufficient . Indeed , the inactivation of Pol II-dependent transcription correlates with decreased intragenic binding ( and vice versa ) , further indicating that Dal80 spreading across gene bodies depends on active transcription . Consistently , we detected a physical interaction between Dal80 and transcriptionally active forms of Pol II . Together , our data lead us to propose a model where Dal80 could travel from the promoter of highly expressed , NCR-sensitive genes through the gene body by accompanying the elongating Pol II complex ( Fig 9 ) . However , it is also possible that Dal80 spreading across gene bodies is determined , but yet temporally distinct , from the passage of the elongating Pol II . For instance , chromatin marks deposited upon Pol II passage could favor Dal80 intragenic binding afterwards . Additional investigations will be required to define which domain of Dal80 is responsible for the interaction with the transcription machinery , to determine whether there is any causal relationship between Dal80 intragenic binding and high expression levels , and to decipher the potential role of Dal80 during active transcription . In this respect , we propose that the leucine zipper domain is not involved . Whereas the binding of elongation factors across gene bodies has been thoroughly documented [76] , it has also been described for some specific transcription factors . For example , Gal4 was reported to bind to its consensus DNA target within the ACC1 ORF , but the authors concluded that the observed transcriptional repression of the ACC1 gene was most likely resulting from random GAL4 binding “noise” over the genome , thus having no physiological explanation for this ORF-bound transcription factor [77] . Likewise , Gcn4 was detected across the PHO8 ORF , with concomitant recruitment of the SAGA complex , but without any impact on gene expression [78] . More recently , binding of the Gcn4 transcription factor to its consensus site at some ORFs , when located in proximity of the transcriptional start site , was found to play a consistent role in controlling embedded cryptic promoters in yeast , thereby affecting Gcn4-dependent transcription of some genes [79] . A recent study has identified CTD phosphorylation of Pol II as a hub that optimizes transcriptome changes to adequately balance optimal growth and stress tolerance responses [80] . The addition of nitrogen to nitrogen-limited cells rapidly results in the transient overproduction of transcripts required for protein translation ( stimulated growth ) whereas accelerated mRNA degradation favours rapid clearing of the most abundant transcripts , like those involved in high affinity permease production , that are highly expressed NCR-sensitive genes , for example [64] . The involvement of the Nrd1-Nab3-Sen1 ( NNS ) and TRAMP complexes in these regulatory responses has been envisioned very recently [81 , 82]; deadenylation , decapping and exonuclease mutants display impaired GAP1 mRNA clearance upon nitrogen upshift [83] . Thus , a possible role of Dal80 ( and possibly of the other GATA factors ) binding along highly expressed genes could be to transmit nutritional signals to elongation-related processes , like histone modification , chromatin remodelling [84 , 85] , mRNA export/processing [86] or roadblock termination [87] . Interestingly , in human cells , GATA factors are also reported to occupy non-canonical sites within the genome , further reinforcing that they can be recruited to the chromatin independently of their motif [24 , 73] . In addition , 43% of the GATA1 peaks were collected among exon , introns and 3’UTR of coding genes in human erythroleukemia cells [73] . It is tempting to hypothesize that GATA factors could have a dual or synergistic role during transcription , i . e . recruiting/stabilizing the PIC complex as for any classical transcription factor in the promoter/enhancer regions and promoting competent transcription at a post initiation step interacting with the RNAPII .
Experiments were conducted using S . cerevisiae strains of the FY genetic background . The strains used are listed in S10 Table . Dal80 and Gat1 were tagged with 13 copies of the c-myc epitope ( Myc13 ) as described [88] using primers listed in S10 and S11 Tables . The PMEP2-URA3 allele in strains FV806-808 , and PTDH3-URA3 , PVMA1-URA3 , PALD6-URA3 alleles in strains FV1105-1107 , respectively , were created by amplification of the URA3 gene using the same strategy , with primers listed in S10 and S11 Tables . Cultures were grown at 29°C to mid-log phase ( A660nm = 0 . 5 ) in YNB ( without amino acids or ammonia ) minimal medium containing the indicated nitrogen source at a 0 . 1% final concentration , glucose ( 3% ) and the appropriate supplements ( 20 μg/ml uracil , histidine and tryptophan ) to cover auxotrophic requirements . Cell extracts and chromatin immunoprecipitations were conducted as described [40] using primers listed in S11 Table . The cells ( 100 ml cultures grown to an absorbance ( A660 nm = 0 . 6 ) corresponding to 6 × 106 cells/ml ) were treated with 1% formaldehyde for 30 min at 25°C and mixed by orbital shaking . Glycine was then added to a final concentration of 500 mM and incubation continued for 5 min . The cells were collected , washed once with cold 10 mM Tris-HCl , pH 8 , washed once with cold FA-SDS buffer ( 50 mM HEPES-KOH , pH 7 . 5 , 150 mM NaCl , 1 mM EDTA , 1% Triton X-100 , 0 . 1% sodium deoxycholate , 0 . 1% SDS , 1 mM phenylmethylsulfonyl fluoride ) , and resuspended in 1 ml of cold FA-SDS buffer . An equal volume of glass beads ( 0 . 5 mm in diameter ) was added , and the cells were disrupted by vortexing for 30 min in a cold room . The lysate was diluted into 4 ml of FA-SDS buffer , and the glass beads were discarded . The cross-linked chromatin was then pelleted by centrifugation ( 17 , 000 × g for 35 min ) , washed for 60 min with FA-SDS buffer , resuspended in 1 . 6 ml of FA-SDS buffer for 15 min at 4°C , and sonicated three times for 30 s . each ( Bioruptor , Diagenode ) , giving fragments with an average size of 250–300 bp . Finally , the sample was clarified by centrifugation at 14 , 000 × g for 30 min and diluted 4-fold in FA-SDS buffer , and aliquots of the resultant chromatin containing solution were stored at –80°C . Pol II and Myc13-tagged proteins were immunoprecipitated by incubating 100 μl of the chromatin containing solution for 180 min at 4°C with 2 μl of mouse anti-Pol II and anti-Myc antibodies , respectively ( SCBT CTD4H8 or SC-40 , respectively ) prebound to 10 μl of Dynabeads Pan Mouse IgG ( Dynal ) according to the manufacturer's instructions . Immune complexes were washed six times in FA-SDS buffer and recovered by treating with 50 μl of Pronase Buffer ( 25 mM Tris , pH 7 . 5 , 5 mM EDTA , 0 . 5% SDS ) at 65°C with agitation . Input ( IN ) and immunoprecipitated ( IP ) fractions were then subjected to Pronase treatment ( 0 . 5 mg/ml; Roche Applied Science ) for 60 min at 37°C , and formaldehyde cross-links were reversed by incubating the eluates overnight at 65°C . Finally , the samples were treated with RNase ( 50 μg/ml ) for 60 min at 37°C . DNA from the IP fractions was purified using the High Pure PCR Product Purification Kit ( Roche Applied Science ) and eluted in 50 μl of 20 mM Tris buffer , pH 8 . IN fractions were boiled 10 min and diluted 500-fold with no further purification prior to quantitative PCR analysis . Quantitative RT-PCR was performed as described previously [40] using primers listed in S11 Table . Total RNA was extracted from 4-ml cultures and cDNA was generated from 100 to 500 ng of total RNA using a RevertAid H Minus first-strand cDNA synthesis kit with oligo ( dT ) 18 primers from Fermentas using the manufacturer's recommended protocol . cDNAs were subsequently quantified by RT-PCR using the Maxima SYBR green qPCR master mix from Fermentas . Cultures ( 100 ml ) were harvested , washed once in 50 mM Tris , pH 8 , and resuspended in 1ml of buffer ( 50 mM Tris , pH 8 , 150 mM NaCl , 5 mM EDTA , 0 . 05% NP-40 , 1 mM phenylmethylsulfonyl fluoride , and complete protease inhibitor cocktail tablets [Roche] ) . Lysis was performed by shaking with 425–600 μm acid-washed glass beads ( Sigma ) on an IKA Vibrax VXR orbital shaker at maximum speed for 30 min at 4°C . Cell debris and glass beads were removed by centrifugation . Immunoprecipitation was performed by incubating 200 μl of total cell extracts with 20 μl of Dynabeads PAN mouse immunoglobulin G ( Invitrogen ) that were preincubated with anti-HA ( SCBT , SC-7392 ) , anti-CTD ( SCBT , CTD4H8 ) , anti-Ser2P ( BioLegend , H5 ) or anti-Ser5P ( BioLegend , H14 ) antibodies and 20 μl of 1% phosphate-buffered saline-bovine serum albumin for 2 h under orbital shaking ( 800 rpm ) at 30°C . Immune complexes were washed three times in lysis buffer , eluted by boiling in sodium dodecyl sulfate ( SDS ) sample buffer , and loaded on SDS-polyacrylamide gel for anti-Myc Western blotting . ChIP-Seq analysis was performed from two biological replicates of proline-grown 25T0b ( no tag ) , FV078 ( DAL80-MYC13 ) and FV034 ( GAT1-MYC13 ) cells . Lysis and chromatin extraction was as described above . The average fragment length of sonicated fragment was 300–350 bp . For each condition , libraries were prepared from 10 ng of “input” or “IP” DNA using the TruSeq ChIP Sample Preparation Kit ( Illumina ) . Single-read sequencing ( 50 nt ) of the libraries was performed on a HiSeq 2500 sequencer . Reads were uniquely mapped to the S . cerevisiae S288C reference genome using Bowtie2 v2 . 1 . 0 [89] , with a tolerance of 1 mismatch in seed alignment . Tags densities were normalized on the total number of uniquely reads mapped . Dal80- and Gat1-bound regions were identified through a peak-calling procedure using version 2 . 0 . 9 of MACS [90] , with a minimum false discovery rate ( FDR ) of 0 . 001 . For each strain and condition , total RNA was extracted from two biological replicates using standard hot phenol procedure , ethanol-precipitated , resuspended in nuclease-free H2O ( Ambion ) and quantified using a NanoDrop 2000c spectrophotometer . Ribosomal RNAs were depleted from 1 μg of total RNA using the RiboMinus Eukaryote v2 Kit ( Life Technologies ) . After concentration using the Ribominus Concentration Module ( Life Technologies ) , rRNA-depleted RNA was quantified using the Qubit RNA HS Assay kit ( Life Technologies ) . In parallel , rRNA depletion efficiency and integrity of both total and rRNA-depleted RNA were checked by analysis in a RNA 6000 Pico chip , in a 2100 bioanalyzer ( Agilent ) . Strand-specific total RNA-Seq libraries were prepared from 125 ng of rRNA-depleted RNA using the TruSeq Stranded Total RNA Sample Preparation Kit ( Illumina ) , following manufacturer’s instructions . Paired-end sequencing ( 2 x 50 nt ) of the libraries was performed on a HiSeq 2500 sequencer . Sequenced reads were mapped to the reference genome using version 2 . 0 . 6 of TopHat [91] , as described [92] . Tags densities were normalized on the total number of reads uniquely mapped on ORFs . Differential expression analysis was performed using DESeq [93] . Differentially expressed genes were identified on the basis of a fold-change ≥2 and a P-value ≤0 . 01 . Statistical details can be found in the corresponding figure legends . Error bars correspond to standard error . Statistical significance tests were carried out using the Student’s t test when indicated . Sequence data can be accessed at the NCBI Gene Expression Omnibus using accession numbers GSE86307 and GSE86325 . Genome browsers for visualization of processed ChIP-Seq and RNA-Seq data are accessible at http://vm-gb . curie . fr/dal80 . Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact , Isabelle Georis ( igeoris@ulb . ac . be ) . Bioinformatics and genome wide dataset requests could also be addressed to antonin . morillon@curie . fr for rapid processing .
|
GATA transcription factors are highly conserved among eukaryotes and play key roles in cancer progression and hematopoiesis . In budding yeast , four GATA transcription factors are involved in the response to the quality of nitrogen supply . Here , we have determined the whole genome binding profile of the Dal80 GATA factor , and revealed that it also associates with the body of promoter-bound genes . The observation that intragenic spreading correlates with high expression levels and exquisite Dal80 sensitivity suggests that GATA factors could play other , unexpected roles at post-initiation stages in eukaryotes .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"statistics",
"gene",
"regulation",
"regulatory",
"proteins",
"dna-binding",
"proteins",
"dna",
"transcription",
"chi",
"square",
"tests",
"mathematics",
"genome",
"analysis",
"transcription",
"factors",
"sequence",
"motif",
"analysis",
"research",
"and",
"analysis",
"methods",
"sequence",
"analysis",
"bioinformatics",
"proteins",
"mathematical",
"and",
"statistical",
"techniques",
"gene",
"expression",
"statistical",
"methods",
"genetic",
"loci",
"biochemistry",
"statistical",
"hypothesis",
"testing",
"database",
"and",
"informatics",
"methods",
"genetics",
"biology",
"and",
"life",
"sciences",
"physical",
"sciences",
"genomics",
"gene",
"prediction",
"computational",
"biology"
] |
2019
|
Transcription-dependent spreading of the Dal80 yeast GATA factor across the body of highly expressed genes
|
Streptococcus pneumoniae ( pneumococcus ) is an opportunistic pathogen that causes otitis media , sinusitis , pneumonia , meningitis and sepsis . The progression to this pathogenic lifestyle is preceded by asymptomatic colonization of the nasopharynx . This colonization is associated with biofilm formation; the competence pathway influences the structure and stability of biofilms . However , the molecules that link the competence pathway to biofilm formation are unknown . Here , we describe a new competence-induced gene , called briC , and demonstrate that its product promotes biofilm development and stimulates colonization in a murine model . We show that expression of briC is induced by the master regulator of competence , ComE . Whereas briC does not substantially influence early biofilm development on abiotic surfaces , it significantly impacts later stages of biofilm development . Specifically , briC expression leads to increases in biofilm biomass and thickness at 72h . Consistent with the role of biofilms in colonization , briC promotes nasopharyngeal colonization in the murine model . The function of BriC appears to be conserved across pneumococci , as comparative genomics reveal that briC is widespread across isolates . Surprisingly , many isolates , including strains from clinically important PMEN1 and PMEN14 lineages , which are widely associated with colonization , encode a long briC promoter . This long form captures an instance of genomic plasticity and functions as a competence-independent expression enhancer that may serve as a precocious point of entry into this otherwise competence-regulated pathway . Moreover , overexpression of briC by the long promoter fully rescues the comE-deletion induced biofilm defect in vitro , and partially in vivo . These findings indicate that BriC may bypass the influence of competence in biofilm development and that such a pathway may be active in a subset of pneumococcal lineages . In conclusion , BriC is a part of the complex molecular network that connects signaling of the competence pathway to biofilm development and colonization .
Bacteria form sessile communities termed biofilms , where they interact with each other to engage in collaborative and/or competitive behaviors [1] . In Streptococcus pneumoniae ( pneumococcus ) , these cell-cell interactions are commonly mediated by secreted peptides that interact with both producing and neighboring cells of the same species , and induce changes in gene regulation that result in altered phenotypes [2] . These dynamic pneumococcal biofilms occur in chronic otitis media , chronic rhinosinusitis and nasopharyngeal colonization [3–8] . The ability to form biofilms is a critical component of pneumococcal disease [9] . Biofilms serve as reservoirs for acute infections [10] . In the middle ear , cells released from a biofilm are thought to be responsible for recurrent episodes of infection [4] . Bacterial cells released from nasopharyngeal biofilms can seed pneumococcal transmission between individuals by being incorporated into nasal shedding . Alternatively , these cells can disseminate to tissues causing mild to severe diseases , such as otitis media , pneumonia , and sepsis [10] . Pneumococcal cells released from biofilms display increased virulence relative to their planktonic or biofilm counterparts , suggesting that chronic biofilms set the stage for the stimulation of a virulence program activated upon the dispersal of cells [11] . Moreover , pneumococci in a biofilm display decreased susceptibility to antibiotics , and are recalcitrant to treatment [6] . Thus , biofilms are an important component of pneumococcal epidemiology in transmission , maintenance of asymptomatic colonization , and development of disease . The transcriptional program required for the initiation and the growth of pneumococcal biofilms has been the subject of numerous investigations . It is clear that at least two quorum sensing ( QS ) signal transduction pathways are critical for biofilm development: competence and Lux [7] , [12–15] . The competence pathway has been the subject of intense investigation for decades [16–28] . Competence is activated by a classic two-component system ( TCS ) where the extracellular competence stimulating peptide ( CSP , encoded by comC ) binds to the surface exposed ComD histidine kinase receptor , inducing its autophosphorylation and the subsequent transfer of the phosphate group to its cognate regulator , ComE [23] , [29] . Activation of the competence pathway leads to increased expression of 5–10% of the pneumococcal genome in two main waves of gene expression [21] , [26] . The first wave of induction is carried out directly by ComE; it upregulates a subset of competence genes ( early genes ) that include comAB , comCDE , as well as the alternative sigma factor , comX . The second wave of competence induction is regulated by ComX; it leads to an increase in the levels of at least 80 genes ( late genes ) , that subsequently modulate important phenotypes such as transformation , metabolism , fratricide and biofilm formation [16] , [26] , [30] . This competence program is upregulated during biofilm mode of growth in vitro , during interactions with human epithelial cells , and in lungs and brain after intranasal and intracranial challenges respectively in murine infection models [7] , [12] , [31] . Importantly , in cell culture models , comC is required for biofilm development [12] , [14] . Thus , activation of the competence pathway is important for productive biofilm formation and critical for pneumococcal infection and adaptation . The Lux QS system also plays a role in biofilm formation . In this system , Lux QS is controlled by the AI-2 autoinducer , which is secreted and sensed by both Gram-positive and Gram-negative species . LuxS is a node in the regulation of competence , fratricide , and biofilm development [15] , [32] . Lux upregulates competence via ComD and ComX [13] . It contributes to bactericidal activity via upregulation of the choline binding murein hydrolase ( CbpD ) . Through lysis , this bactericidal activity increases the levels of extracellular DNA , which is a key ingredient in the extracellular polymeric substance ( EPS ) that makes up the biofilm . Thus , the competence and Lux systems provide the molecular framework to coordinate multi-cellular bacterial communities to form and develop robust biofilms during infection . Whereas the role of competence signaling in biofilm development is well established , the molecules that connect competence to biofilms are poorly understood [3] , [7] , [15] , [33] . In this study , we identify one such molecule that links competence and biofilms . We characterize the gene encoding BriC ( Biofilm regulating peptide induced by Competence ) , a novel colonization factor in the competence pathway . Levels of briC are regulated by ComE , and briC promotes biofilm development and nasopharyngeal colonization . Further , genomic analysis of briC reveals polymorphisms in its promoter , where a subset of strains encode a RUP ( for repeat unit of pneumococcus ) sequence , which leads to additional and CSP-independent expression of briC .
We have identified the gene encoding a putative secreted peptide that is co-regulated with competence ( spd_0391 ( D39 ) ; spr_0388 ( R6 ) ; sp_0429 ( TIGR4 ) ) . Based on the results presented in this manuscript , we have termed it Biofilm-regulating peptide induced by Competence ( BriC ) . BriC was identified in our previously described in silico screen designed to capture cell-cell communication peptides in the pneumococcal genome [34] . The known double glycine ( Gly-Gly ) streptococcal peptides are exported and proteolytically processed by dedicated ABC transporters that recognize N-terminal sequences with the Gly-Gly leader peptide ( LSXXELXXIXGG ) [23] . In our previous work , we identified novel secreted pneumococcal peptides using computational analysis to search for proteins with N-termini that contain a Gly-Gly leader . Our input set consisted of the alleles of two exported Gly-Gly peptides , the signaling molecule CSP and the bacteriocin BIP [23] , [35] . Our output consisted of a position dependent probability matrix that captures the length and positional variability at each residue of the Gly-Gly motif . When we searched for this motif in a database of sixty streptococcal genomes , we defined a predicted secretome consisting of twenty-five sequence clusters , one of which corresponds to BriC [34] . To identify genes co-regulated with briC , we performed transcriptional studies using a NanoString probe set that reports on the abundance of the briC transcript as well as transcripts encoding a subset of pneumococcal regulators and cell wall proteins . We assessed the levels of briC transcript in vitro and in vivo . In vitro expression was measured by screening RNA extracted from mid-log planktonic cultures of a laboratory strain ( R6-derivative ( R6D ) ) . In vivo expression was evaluated by analysis of middle-ear effusions recovered from chinchillas infected with a clinical PMEN1 strain . The mRNA levels of briC were positively associated with comC and comE in vitro ( strain R6D: R2 = 0 . 61 and 0 . 79 , respectively ) and in vivo ( strain PN4595-T23: R2 = 0 . 92 and 0 . 88 , respectively ) . It is noteworthy that changes in the expression of genes in this locus were observed in the studies that first documented the competence regulon [21] , [26] . Specifically , Peterson and colleagues observed changes in briC levels , however the association between briC and CSP was below the statistical threshold [26] . Further , Dagkessamanskaia and colleagues observed an upregulation in the gene downstream of briC , predicted to be in the same operon [21] . Given that in strains R6 , R6D , and D39 , this downstream gene is truncated , this study does not explore the function of the downstream gene . In summary , our gene expression analysis suggests that briC is induced by competence . To directly test whether briC is a competence-regulated peptide , we employed a fusion of the briC promoter to the lacZ reporter ( R6 PbriC-lacZ ) . Stimulation of the signal transduction system that initiates competence by addition of CSP1 led to an induction of the β-galactosidase activity by about twenty-five-fold ( Fig 1 ) . Induction of the briC promoter was specific to the CSP pherotype encoded by strain R6 . The β-galactosidase activity was observed upon addition of CSP1 , the CSP pherotype from strain R6 , but not upon addition of the non-cognate CSP2 pherotype ( Fig 1 ) . Thus , we conclude that briC is a competence-responsive gene . Our in silico analysis of the briC promoter in strains R6 and R6D revealed the presence of a ComE-binding site . ComE binds a well-defined sequence consisting of two imperfect direct repeats of nine nucleotides separated by a gap of twelve or thirteen base pairs [36] . Our analysis of the putative briC promoter across pneumococcal strains revealed an excellent match to the ComE-binding box ( Fig 2A ) . To further investigate the association between ComE and briC , we tested whether CSP-induction of briC requires ComE . We compared the CSP-induction of briC in a wild-type ( R6D WT ) strain to that of an isogenic comE-deletion mutant ( R6DΔcomE ) , using qRT-PCR analysis . In WT cells , the addition of CSP triggered a significant increase in levels of briC at 10 minutes post-addition , with levels slowly decreasing by 15 minutes ( Fig 2B ) . This trend follows the temporal pattern observed for the levels of comE that has been associated with genes under direct controls of ComE [21] , [26] . In contrast , the transcript levels of briC were unaffected by CSP addition in the ΔcomE strain , indicating that the expression of briC requires ComE ( Fig 2B ) . These results strongly suggest that briC is directly regulated by ComE . In addition , ComE plays a critical role in controlling transformation , thus we investigated whether briC impacts transformation efficiency ( S1 Result and S1 Fig ) . We find that absence of briC leads to only a minor decrease in transformation efficiency . To investigate whether expression of briC plays a role in biofilm development , we compared biofilm formation across WT ( R6D WT ) , briC deletion mutant ( R6DΔbriC ) , and briC complemented ( R6DΔbriC::briC ) strains grown on an abiotic surface at 24h and 72h post-seeding . No difference was observed in biofilm biomass and thickness at 24h post-seeding , suggesting that expression of briC does not contribute to early stages of biofilm formation ( Fig 3A and 3B ) . In contrast , at 72h post-seeding , ΔbriC biofilms displayed significantly reduced biomass and thickness when compared to WT ( Fig 4A and 4B ) . Further , biofilms with ΔbriC::briC cells restored the WT phenotype at this time-point ( Fig 4A and 4B ) . The indistinguishable biofilm parameters of WT and ΔbriC cells at 24h post-seeding suggests that there is no fitness-related growth difference between the strains and indicates that the biofilm defect is biologically relevant . These findings suggest that briC contributes to late biofilm development . To investigate the prevalence of briC , we investigated its distribution across the genomes of pneumococcus and related streptococci . To place the distribution in the context of phylogeny , we used a published species tree generated from a set of fifty-five genomes [34] , [38] ( S1 Table ) . The genomes encompass thirty-five pneumococcal genomes that span twenty-nine multi-locus sequence types as well as eighteen serotypes and nontypeable strains; eighteen genomes from related streptococcal species that also colonize the human upper respiratory tract , namely S . pseudopneumoniae , S . mitis , and S . oralis; and finally , two distantly related S . infantis strains as an outgroup . In this set , all the pneumococcal genomes encode briC , and there are two highly similar alleles ( labeled allele 1A and 1B , Figs 5 and 6 ) . Further , we identified four additional alleles in the related streptococci ( Fig 5 , S1 File ) . Next , we extended this analysis to a set of 4 , 034 pneumococcal genomes available in the pubMLST database ( these correspond to all the genomes with at least 2Mb of sequence ) [39] . In total , 98 . 5% ( 3 , 976 out of 4 , 034 ) of these genomes encode a briC allele , suggesting briC is highly prevalent across pneumococcal strains . We find that alleles 1A and 1B are prominent in this larger set , with 1 , 824 and 1 , 187 representatives respectively . After manual curation , we retrieved nineteen distinct briC alleles across these pneumococcal genomes ( Fig 6 ) . Six of the polymorphic residues are located in the putative leader sequence . The conserved region of the leader sequences corresponds to the amino acids preceding the Gly-Gly [23] , [34] , thus the polymorphisms at the N-terminal end of the BriC sequence are not predicted to influence export . One polymorphic residue replaces the glycine in the Gly-Gly motif with a serine; it seems probable that this variation may influence processing and/or export . In addition , position -2 from the C-terminus encodes either an alanine or a threonine . This variation is at the C-terminus , predicted to be the functional end of the molecule , such that the difference in size and polarity at this residue is likely to impact structure or binding of BriC to its targets . The briC gene is induced by competence , so we investigated whether there is a correlation between CSP pherotypes and alleles encoding BriC . However , we did not find these to be associated ( S2 Table ) . Finally , to investigate the distribution of briC in other species and genera , we used BLASTp to search the non-redundant database [40] . We find that BriC homologs are present in strains of related streptococci , S . pseudopneumoniae , S . mitis and S . oralis , but we did not identify homologues in more distant species . The phylogenetic distribution of briC supports a conserved role across pneumococci and a subset of related streptococcal species . Our analysis of the promoter region of briC in our curated set of genomes reveals that a subset of strains encode for a 107bp insertion within the region upstream of briC ( Fig 5 ) . The additional nucleotides are located after the ComE-binding site and before the transcriptional start site , and correspond to a repeat unit in pneumococcus ( RUP ) sequence [41] , [42] . RUP is an insertion sequence derivative with two clear variants , which may still be mobile [42] . The RUP sequence upstream of briC corresponds to RUPB1 . In our curated genomes , the long RUP-encoding promoter is present in multiple strains , including those from the clinically important PMEN1 and PMEN14 lineages ( Fig 5 ) . Using our expanded database of 4 , 034 pneumococcal genomes , we determined that the vast majority of the PMEN1 and PMEN14 genomes encode a long promoter . Specifically , of the manually curated subset , 99 . 4% of PMEN1 strains and 100% of PMEN14 strains encode the long promoter . The high prevalence of the long promoter in these lineages suggests that this form was present in the ancestral strains from these lineages and/or provides a fitness advantage in these genomic backgrounds . To investigate how this genomic difference influences briC expression , we generated a LacZ reporter strain . The 263bp upstream of briC from the PMEN1 strain , PN4595-T23 , were fused to lacZ to produce the PbriClong-lacZ reporter , and its reporter activity was compared to that of the PbriC-lacZ generated with the fusion of 159bp upstream of briC obtained from strain R6 . To control for the possibility that the influence of the RUP sequence might be strain-dependent , we tested these reporter constructs in the R6 and the PMEN1 backgrounds , in both the absence and presence of CSP treatment ( Fig 7A and 7B ) . The presence of RUP dramatically increased the basal levels of briC in the absence of CSP , and this increase was observed in both R6 and PMEN1 . Furthermore , both constructs were induced upon addition of the cognate CSP . These findings suggest that the RUP sequence serves as an expression enhancer; it increases the levels of briC transcripts and this increase is CSP-independent . Thus , in some lineages , briC appears to be under the control of both CSP-dependent and CSP-independent regulation . Next , we investigated the biological impact of the natural variations in the briC promoter on biofilm development . It has been well established that competence promotes biofilm development . Specifically , deletion of the comC ( encodes CSP ) and comD ( encodes histidine kinase of competence TCS ) genes lead to a reduction in in vitro biofilms in strains D39 and TIGR4 [7] , [12] . In this study , we have established that briC also promotes biofilm development ( Fig 4A and 4B ) , and that the RUP-containing long promoter serves as an expression enhancer , increasing basal expression of briC to levels higher than the CSP-dependent induction observed with the short promoter ( Fig 7 ) . Thus , we hypothesized that expression of briC from the long promoter may bypass the impact of competence in biofilm development . In concurrence with previous work , we observed that a strain with a comE deletion ( R6DΔcomE ) displays a reduction in biofilm biomass and thickness relative to the WT strain ( Fig 8A and 8B ) . ComE influences the expression of numerous genes . To determine whether the biofilm defects were primarily due to its impact on briC induction , we tested a construct with a disruption of the ComE-binding box in the briC promoter ( R6DΔbriC::PbriCShuffled ComE-box-briC ) . This strain displays a significant reduction in biofilm biomass and thickness relative to the WT strain ( Fig 8A and 8B ) . Moreover , no difference was observed in the biofilm parameters for both of these mutants , suggesting that the absence of briC expression is a contributor to the in vitro biofilm defect in the comE deletion mutant . Next , we determined that a strain with increased basal levels of briC driven by the RUP-containing long promoter ( R6DΔcomE::PbriClong-briC ) fully rescued the biofilm defect observed in R6DΔcomE ( Fig 8A and 8B ) . In addition , increased expression of briC in the wild type background ( R6D WT::PbriClong-briC ) did not lead to a significant increase in biofilm biomass and thickness relative to the wild type ( Fig 8A and 8B ) . Together , these data strongly suggest that briC is a critical molecular link between competence and biofilm formation , and that natural variations in the briC promoter are physiologically relevant . Since BriC is associated with the competence pathway and is able to rescue the biofilm defects associated with competence signaling , we investigated whether competence associated transporters play a role in exporting BriC . In pneumococcus , the ComAB and BlpAB C39-peptidase transporters export peptides with a Gly-Gly leader [43–45] . These transporters recognize the N-terminal leader of target sequences , and cleave these sequences at the Gly-Gly motif [45] , [46] . In strains R6 and R6D , BlpAB is not functional due to a frameshift mutation that leads to an early stop codon [47] . Thus , we hypothesized that as a Gly-Gly peptide co-expressed with genes of the competence pathway , BriC may be exported via the ComAB transporter . We tested this hypothesis in two ways . First , we measured whether deletion of comAB influences the ability of a strain with competence-independent expression of briC to rescue the ΔcomE-biofilm defect . Second , we compared secretion of a BriC reporter construct in a WT strain with that in a comAB deletion mutant . Our biofilm data suggests that ComAB plays a role in transporting BriC . At 72h post-seeding , a comE/comAB-double deletion mutant strain expressing briC from the RUP-encoding long promoter ( R6DΔcomEΔcomAB::PbriClong-briC ) displayed a biofilm with reduced biomass and thickness when compared to the equivalent construct in only a comE-deletion background ( R6DΔcomE::PbriClong-briC ) ( Fig 9A and 9B ) . However , the biofilm levels were not reduced to the levels observed in the ΔcomE strain . These results suggest that under these conditions , ComAB may not be the only transporter that contributes to the export of BriC . To further elucidate the role of ComAB in the export of BriC , we employed the HiBiT tag detection system , which was recently used to detect secretion of BlpI [44] . The HiBiT tag corresponds to an 11-residue peptide . The assay works by addition of an inactive form of luciferase ( LgBit ) to the extracellular milieu . When both LgBit and HiBiT combine , they generate bioluminescence [48] , [49] . To study BriC transport , we fused the putative BriC leader sequence to the HiBiT tag and expressed this reporter under the control of the native ( short ) briC promoter in WT and ΔcomAB R6-strains . We measured the extracellular bioluminescence produced by this reporter both in the presence and absence of CSP ( Fig 9C , S3 Table ) . In the absence of CSP , the levels of secreted HiBiT resembled that of background ( WT cells without HiBiT ) , consistent with very low expression of N-terminal BriC-HiBiT as well as low expression of the ComAB transporter . In the WT background , upon addition of CSP , N-terminal BriC-HiBiT is induced and the extracellular level of HiBiT is significantly increased , consistent with HiBiT export . In the ΔcomAB background , upon addition of CSP , N-terminal BriC-HiBiT is induced and the extracellular levels of HiBiT also increase . However , the increase in the level of extracellular HiBiT observed in ΔcomAB strain is significantly less than that in the WT strain , consistent with a severe reduction in HiBiT export in the absence of comAB . Combined , these results strongly suggest that ComAB serves as a major transporter for BriC . During nasopharyngeal colonization , pneumococci form biofilms and upregulate the competence pathway . Thus , we investigated the role of briC in nasopharyngeal colonization using an experimental murine colonization model . Our in vitro investigations have been performed using strain R6D , which is defective in colonization due to the absence of a capsule [50] . Thus , we performed colonization experiments with the serotype 2 strain , D39 , which is the ancestor of strain R6 [51] . Mice were colonized with D39 WT , the briC-deletion mutant ( D39ΔbriC ) or the briC-complemented ( D39ΔbriC::briC ) strains . Comparison of the number of bacteria in nasal lavages immediately after inoculation revealed that mice in the three cohorts received the same number of bacteria . In contrast , nasal lavages at three and seven days post-inoculation revealed decreased levels of D39ΔbriC relative to WT in the nasal wash ( Fig 10A ) . Furthermore , the WT levels were restored in the complemented strain ( Fig 10A ) . These findings indicate that briC encodes a novel colonization factor . In in vitro biofilms , briC links competence to biofilms . First , disruption of the ComE- binding box in the briC promoter led to a biofilm defect similar to that observed in a ΔcomE strain . Second , overexpression of briC driven by the long version of the promoter was found to restore the competence-dependent defect in in vitro biofilm development . Thus , we investigated the behavior of these strains in the pneumococcus colonization model . We found that the strain with a disruption of the ComE-binding box ( D39ΔbriC::PbriCShuffled ComE-box-briC ) within the briC promoter was defective for colonization , the decreased bacterial counts resembled those in the D39ΔcomE strain ( Fig 10B ) . These findings suggest that briC is a substantial mediator of the role of ComE on colonization . Further , addition of this long briC promoter to ΔcomE cells ( D39ΔcomE::PbriClong-briC ) partially rescues the colonization defect of the D39ΔcomE strain . That is , the numbers of bacterial cells of strain D39ΔcomE::PbriClong-briC recovered from the nasal lavages at both three and seven days post-inoculation were significantly higher than the numbers of D39ΔcomE cells recovered , but less than that of the D39 WT ( Fig 10B ) . Finally , the overexpression of briC in the WT background ( D39 WT::PbriClong-briC ) does not impact colonization . Thus , we conclude that BriC is a contributor to the competence-induced stimulation of nasopharyngeal colonization observed in strain D39 . Further , natural variations leading to a long briC promoter appear to dampen the impact of competence in colonization .
An important component of pneumococcal pathogenesis is its ability to form complex biofilm structures . Pneumococci in a biofilm mode of growth display decreased sensitivity to antibiotics and increased resistance to host immune responses [6] . These properties make the bacteria recalcitrant to treatment and highlight the need to better understand the molecular mechanisms that drive biofilm development . Activation of the competence pathway is critical for biofilm development . Previous in vitro studies have demonstrated that while cell-adherence and early biofilm formation are competence-independent , an intact competence system is required in the later stages of biofilm development . It was shown that the competence pathway positively influences structure and stability of late stage biofilms [12] . However , the molecules downstream of competence activation by ComDE that regulate biofilm development remain poorly understood . In this study , we present BriC , a previously uncharacterized peptide , that we show is regulated by competence and plays a role in promoting biofilm development and nasopharyngeal colonization . We have presented extensive evidence that briC is a competence regulated gene . We have shown that induction of briC is triggered by addition of CSP and requires ComE . Further , we have also shown that the briC promoter encodes the consensus ComE-binding box , and that briC expression follows the temporal pattern described for genes directly regulated by ComE . Previous studies have used microarray analysis to identify pneumococcal genes differentially regulated upon CSP stimulation [21] , [26] and have categorized these genes into two main groups—early genes regulated by ComE or late genes regulated by the alternative sigma factor , ComX . In the study by Peterson and colleagues , briC was found to be upregulated in a pattern consistent with early genes [26] . However , the upregulation was not found to be statistically significant , and this study is the first validation of briC as a competence-regulated peptide . We have provided evidence that briC stimulates biofilm development on abiotic surfaces and promotes nasopharyngeal colonization in a murine model . These findings are consistent with studies that show that pneumococcal biofilms contribute to colonization . Colonization of the upper respiratory tract is a requisite for pneumococcal dissemination to distant anatomical sites and subsequent disease [10] . These sessile communities serve as a source of pneumococcal cells with an activated virulence-associated transcription program . That is , when compared to cells originating from a planktonic mode of growth , those originating from a biofilm mode of growth are more likely to cause disease upon infecting other tissues [11] . In this manner , increased biofilm development likely heightens the risk of disease . Biofilms and competence are also associated with transformation efficiency . We have observed a mild but significant decrease in the transformation efficiency of briC-deletion mutants relative to WT R6D cells ( S1 Fig ) . Finally , colonization of the upper respiratory tract is also a reservoir for pneumococcal transmission . Transmission occurs when cells migrate from the nasopharynx of one host to that of another . Thus , BriC’s contribution to colonization may influence both disease severity and transmission . While it has been established that CSP contributes to biofilm development , the competence-dependent genes that regulate biofilm development are not well understood [7] , [12] . Our finding that increased levels of briC can fully rescue biofilm defects from a comE deletion mutant in vitro , and partially rescue its colonization defects in vivo suggests that briC expression may bypass the requirement for competence in biofilm development . ComE is a key regulator of competence whose activity is required to regulate approximately 5–10% of the genome , and as such deletion of comE is expected to have substantial global consequences [21] , [26] . In this context , it is noteworthy that overexpression of one gene ( briC ) in the comE-deletion mutant was able to improve colonization in the murine model . These findings strongly suggested that BriC is a molecular link between competence , biofilm development , and colonization . Our data suggests that many strains have multiple inputs to the regulation of briC . Shared across all strains is the regulation by ComE , the key regulator of the competence pathway . Competence is responsive to environmental cues , such as changes in cell density , pH , mutational burden in cells , and exposure to antibiotics [16] , [52–54] . Conversely , competence is inhibited by the degradation of CSP via the activity of the CiaHR TCS and the serine protease , HtrA [55] , [56] . Factors altering competence will also alter briC levels due to its competence-dependent induction . Our comparative genomics suggest that a subset of pneumococcal lineages may encode an additional briC-regulatory element . Specifically , the briC promoter differs across strains , in that a subset of lineages encodes a long promoter with a RUP sequence ( PbriClong ) and has higher basal levels of briC expression . This long promoter is constitutively active , even when competence is off . The long promoter is encoded in the vast majority of strains from the PMEN1 lineage ( Spanish-USA ) and the PMEN14 ( Taiwan-19F ) lineages . These lineages are prominent in the clinical setting; they are multi-drug resistant and pandemic [57–59] . This additional competence-independent regulation of the long promoter may provide promoter-binding sites for additional regulators or reflect consequences of positional differences for the existing promoter binding sites . Our biofilm and colonization experiments suggest that encoding the long briC promoter has functional consequences . We conclude that the response of briC to competence is ubiquitous , but that additional lineage-specific factors influence briC regulation and downstream phenotypic consequences . We propose a model where briC encodes a signaling molecule with a role in biofilm development and colonization . First , the transcription of briC is induced by ComE through competence signal transduction pathway in all lineages , and possibly by additional regulator ( s ) in a subset of lineages . Once this Gly-Gly peptide is produced , we propose that it is exported through ABC transporters , a process in which ComAB plays a prominent role . Based on a bioinformatic comparison with other Gly-Gly peptides we suggest that BriC is cleaved into its active form ( BRIC ) during export . It is tempting to speculate that BRIC is a new member of the expanding set of pneumococcal secreted peptides that signal to neighboring cells promoting population-level behaviors . In this era of emerging antibiotic resistance , it is imperative that we test the potential of alternative strategies to inhibit bacterial carriage and disease . One such strategy is to specifically target bacterial communities and population-level behaviors . In that regard , molecules such as BriC present promising alternatives to be used as targets for discovery of novel drugs and therapeutic interventions .
Four wild-type ( WT ) Streptococcus pneumoniae strains were used for this experimental work . The majority of studies were performed on a penicillin-resistant derivative of R6 ( R6D ) ; this strain was generated from a cross where parental strain R6 was recombined with Hungary19A and the recombinant was selected for penicillin resistance [60] . The briC allele in R6D is identical to the allele present in the parental R6 . This laboratory strain is non-encapsulated and does not colonize mice , thus mice colonization experiments were performed with the serotype 2 D39 strains ( GenBank CP000410 ) [61] . The D39 strain contains the same briC allele as is present in the R6D strain , which has been used for most of the work in this study . Finally , for a representative of PMEN1 , we used the carriage isolate , PN4595-T23 ( GenBank ABXO01 ) graciously provided by Drs . Alexander Tomasz and Herminia deLancastre [62] . Colonies were grown from frozen stocks by streaking on TSA-II agar plates supplemented with 5% sheep blood ( BD BBL , New Jersey , USA ) . Colonies were then used to inoculate fresh Columbia broth ( Remel Microbiology Products , Thermo Fisher Scientific , USA ) and incubated at 37°C and 5% CO2 without shaking . When noted , colonies were inoculated into acidic Columbia broth prepared by adjusting the pH of Columbia broth to 6 . 6 using 1M HCl . Acidic pH was used to inhibit the endogenous activation of competence . The mutant strains ( R6DΔbriC and PN4595ΔbriC ) were constructed by using site-directed homologous recombination to replace the region of interest with erythromycin-resistance gene ( ermB ) or kanamycin-resistance gene ( kan ) , respectively ( S4 Table ) . The kan and spectinomycin-resistance gene ( aad9 ) were used to construct ΔcomE strains in R6D and PN4595-T23 respectively . Briefly , the transformation construct was generated by assembling the amplified flanking regions and antibiotic resistance cassettes . ~2kb of flanking regions upstream and downstream of the gene of interest was amplified from parental strains by PCR using Q5 2x Master Mix ( New England Biolabs , USA ) . The antibiotic resistance genes , kan and aad9 were amplified from kan-rpsL Janus Cassette and pR412 , respectively ( provided by Dr . Donald A . Morrison ) , and ermB was amplified from S . pneumoniae SV35-T23 . SV35-T23 is resistant to erythromycin because of the insertion of a mobile element containing ermB [62] . These PCR fragments were then assembled together by sticky-end ligation of restriction enzyme-cut PCR products . The deletion mutant in R6D is an overexpressor of the downstream peptide ( spr_0389 ) . The briC complement and overexpressor strains were generated using constructs containing the CDS of briC along with either its entire native promoter region or overexpressing promoter respectively , ligated at its 3’ end with a kanamycin resistance cassette . The promoters used to overexpress briC included either the constitutive amiA promoter , or PbriClong . These were assembled with the amplified flanking regions by Gibson Assembly using NEBuilder HiFi DNA Assembly Cloning Kit . The construct was introduced in the genome of R6D downstream of the bga region ( without modifying bga ) , a commonly employed site [63] . Primers used to generate the constructs are listed in S5 Table . Like R6DΔbriC , R6DΔbriC::briC is also an overexpressor of the downstream peptide ( spr_0389 ) , which is annotated as a pseudogene in strains R6D , R6 and D39 ( Fig 2A ) . The expression of spr_0389 remains unchanged in the mutant and the complement ( expression of spr_0389 was induced by 5-6-fold in both the mutant and the complement relative to WT ) . The R6DΔcomE::PbriClong-briC strain was constructed by replacing comE with spectinomycin resistant cassette in the R6D PbriClong-briC strain . The comAB-deletion mutant in a briC overexpressor R6D genomic background strain ( R6DΔcomAB::PbriClong-briC ) was constructed by transforming the R6D::PbriClong-briC strain with the genomic DNA of ADP226 . ADP226 is a strain from the D39 genomic background with comAB replaced by erythromycin resistance cassette . To make the construct , the flanking regions and erythromycin resistance cassette were amplified , and then assembled together by sticky-end ligation of restriction enzyme-cut PCR products . The construct was then transformed into D39 ADP225 and selected on Columbia blood agar supplemented with 0 . 25 μg mL−1 erythromycin ( S4 Table ) . The briC promoter region was modified by shuffling the ComE-binding box ( R6DΔbriC::PbriCShuffled ComE-box-briC ) . The ComE-binding box was shuffled using PCR by amplifying from R6DΔbriC::briC and introducing the shuffled sequence ( CAGACCAGTTAGTCTAGGATAGAGCTTAAG ) into the primers . The resulting amplicons were assembled using Gibson Assembly . The modified construct was transformed into R6DΔbriC strain in the region downstream of the bgaA gene . The D39 briC deletion mutant ( D39ΔbriC ) , briC complemented ( D39ΔbriC::briC ) , comE deletion mutant ( D39ΔcomE ) , briC overexpressor in comE deletion background ( D39ΔcomE::PbriClong-briC ) , and briC expressed from a promoter with a shuffled ComE-binding box ( R6DΔbriC::PbriCShuffled ComE-box-briC ) strains were generated by transformation with the corresponding constructs amplified from R6D into strain D39 . Chromosomal transcriptional lacZ-fusions to the target promoters were generated to assay promoter activity . These lacZ-fusions were generated via double crossover homologous recombination event in the bgaA gene using modified integration plasmid pPP2 . pPP2 was modified by introducing kan in the multiple cloning site , in a direction opposite to lacZ . The modified pPP2 was transformed into E . coli TOP10 . The putative briC promoter regions were amplified from R6 and PN4595-T23 strains , and modified to contain KpnI and XbaI restriction sites , which were then assembled in the modified pPP2 plasmid by sticky-end ligation of the enzyme digested products . These plasmids were transformed into E . coli TOP10 strain , and selected on LB ( Miller’s modification , Alfa Aesar , USA ) plates , supplemented with ampicillin ( 100μg/ml ) . These plasmids were then purified by using E . Z . N . A . Plasmid DNA Mini Kit II ( OMEGA bio-tek , USA ) , and transformed into pneumococcal strains R6 and PN4595-T23 and selected on Columbia agar plates supplemented with kanamycin ( 150μg/ml ) . For all bacterial transformations to generate mutants , target strains ( R6D or D39 ) were grown in acidic Columbia broth , and 1μg of transforming DNA along with 125μg/mL of CSP1 ( sequence: EMRLSKFFRDFILQRKK; purchased from GenScript , NJ , USA ) was added to them when the cultures reached an OD600 of 0 . 05 , followed by incubation at 37°C . After 2 hours , the treated cultures were plated on Columbia agar plates containing the appropriate antibiotic; erythromycin ( 2μg/ml ) , or kanamycin ( 150μg/ml ) . Resistant colonies were cultured in selective media , and the colonies confirmed using PCR . Bacterial strains generated in this study are listed in S4 Table . For transformation efficiency experiments , R6D strain was grown in acidic Columbia broth until it reached an OD600 of 0 . 05 . At this point , number of viable cells was counted by plating serial dilutions on TSA-blood agar plates . Transformations were carried out by adding either 100ng or 500ng of transforming DNA in the media supplemented with 125μg/mL of CSP1 and incubated at 37°C for 30mins . For transforming DNA , we used either genomic DNA or PCR products . The donor DNA contained spectinomycin-resistance gene ( aad9 ) in the inert genomic region between spr_0515 and spr_0516 . This construct was generated in PN4595-T23 , specR , followed by its amplification and transformation into R6D and Taiwan-19F strains ( Sp3063-00 ) . The genomic DNA was extracted from Taiwan-19F , specR strain . The purified linear DNA was an amplimer of the region from R6D . After 30 minutes , the cultures were plated on Columbia agar plates containing spectinomycin ( 100μg/ml ) , incubated overnight , and colonies were counted the next day . RNA extraction consists of sample collection , pneumococcal cell lysis , and purification of RNA . For qRT-PCR analysis , the strains ( R6D and R6DΔcomE ) were grown to an OD600 of 0 . 3 in acidic Columbia broth , followed by CSP1 treatment for 0 , 10 , or 15 minutes . For in vitro transcriptomic analysis using NanoString Technology , the R6D strain was grown to an OD600 of 0 . 1 in Columbia broth ( in one experimental set , the samples were grown in sub-lethal concentration of penicillin ( 0 . 8μg/ml ) for an hour ) . RNA was collected in RNALater ( Thermo Fisher Scientific , USA ) to preserve RNA quality and pelleted . For the in vivo experiments , the RNA was extracted from middle-ear chinchilla effusions infected with PN4595-T23 and PN4595-T23ΔcomE strains and preserved by flash freezing the effusion . In all the samples , the pneumococcal cell lysis was performed by re-suspending the cell pellet in an enzyme cocktail ( 2mg/ml proteinase K , 10mg/ml lysozyme , and 20μg/ml mutanolysin ) , followed by bead beating with glass beads ( 0 . 5mm Zirconia/Silica ) in FastPrep-24 Instrument ( MP Biomedicals , USA ) . Finally , RNA was isolated using the RNeasy kit ( Genesee Scientific , USA ) following manufacturer’s instructions . For analysis with the NanoString , which does not require pure DNA , the output from the RNeasy kit was loaded on the machine without further processing . For analysis using qRT-PCR , contaminant DNA was removed by treating with DNase ( 2U/μL ) at 37°C for at least 45 mins . The RNA concentration was measured by NanoDrop 2000c spectrophotometer ( Thermo Fisher Scientific , USA ) and its integrity was confirmed on gel electrophoresis . The purity of the RNA samples was confirmed by the absence of a DNA band on an agarose gel obtained upon running the PCR products for the samples amplified for gapdh . nCounter Analysis System from NanoString Technology provides a highly sensitive platform to measure gene expression both in vitro and in vivo , as previously described [64] . Probes used in this study were custom-designed by NanoString Technology , and included housekeeping genes gyrB and metG , which were used as normalization controls . 5μL of extracted RNA samples were hybridized onto the nCounter chip following manufacturer’s instructions . RNA concentration ranged from 80-200ng/μL for in vivo samples , and between 60-70ng total RNA for in vitro samples . A freely available software from manufacturers , nSolver , was used for quality assessment of the data , and normalization . The RNA counts were normalized against the geometric mean of gyrB and metG [65] , [66] . The 16S rRNA gene is not optimal for normalization in the NanoString , as the high abundance of this transcript packs the field of view . Pearson’s Correlation Coefficient was used to estimate correlation in the expression levels of different genes . High quality RNA was used as a template for first-strand cDNA synthesis SuperScript VILO synthesis kit ( Invitrogen ) . The resulting product was then directly used for qRT-PCR using PerfeCTa SYBR Green SuperMix ( Quantabio , USA ) in an Applied Biosystems 7300 Instrument ( Applied Biosystems , USA ) . 16S rRNA counts were used for normalization . The raw data was then run through LinregPCR for expression data analysis , where the output expression data is displayed in arbitrary fluorescence units ( N0 ) that represent the starting RNA amount for the test gene in that sample [67] , [68] . Fold-change relative to WT was then calculated for each individual experiment . β-galactosidase assays were performed as previously described [69] using cells that were grown in acidic Columbia broth to exponential phase . Cells were either left untreated , or independently treated with CSP1 ( EMRLSKFFRDFILQRKK ) or CSP2 ( EMRISRIILDFLFLRKK ) ( Genscript , USA ) for 30 minutes and processed for analysis . Pneumococcal cultures grown in Columbia broth were used to seed biofilms on abiotic surfaces . When the cultures reached an OD600 of 0 . 05 , each bacterial strain was seeded on 35MM glass bottom culture dishes ( MatTek Corporation , USA ) . To promote biofilm growth , the plates were incubated at 37°C and 5% CO2 . Every 24 hours , the supernatant was carefully aspirated , followed by addition of the same volume of pre-warmed Columbia broth at one-fifth concentration . The biofilm samples were fixed at two time-points: 24 and 72h . For fixing , the supernatants were carefully aspirated , and biofilms were washed thrice with PBS to remove non-adherent and/or weakly adherent bacteria . Subsequently , biofilms were fixed with 4% PFA ( Electron Microscopy Sciences ) , washed three times with PBS , and prepared for confocal microscopy . Fixed biofilms were stained with SYTO59 Nucleic Acid Stain ( Life Technologies , USA ) for 30 minutes , washed three times , and preserved in PBS buffer for imaging . Confocal microscopy was performed on the stage of Carl Zeiss LSM-880 META FCS , using 561nm laser line for SYTO59 dye . Stack were captured every 0 . 46 μm , imaged from the bottom to the top of the stack until cells were visible , and reconstructed in Carl Zeiss black edition and ImageJ . The different biofilm parameters ( biomass , maximum thickness , and average thickness over biomass ) were quantified using COMSTAT2 plug-in available for ImageJ [70] . For depiction of representative reconstructed Z-stacks , empty slices were added to the images so the total number of slices across all the samples were the same . These reconstructed stacks were pseudo-colored according to depth using Carl Zeiss black edition . The color levels of the images being used for representation purposes were adjusted using GNU Image Manipulation Program ( GIMP ) . HiBiT constructs were designed by fusing the C-terminus of the region of interest with the 11-amino acid HiBiT peptide using a 10-amino acid linker . The region of interest was the putative secretion signal ( until the double glycine ) of the briC gene . The expression of these constructs was designed to be controlled by the briC promoter region . The construct was introduced in the genome of R6D and R6DΔcomAB strains downstream of the bgaA gene ( without modifying bgaA ) . R6D strains containing HiBiT constructs were started from overnight blood agar plates into acidic Columbia broth ( pH 6 . 6 ) and incubated at 37°C and 5% CO2 without shaking . Cultures were grown to an OD600 of ~0 . 2 . Cultures were either left untreated or treated with 125μg/mL of CSP1 for 30 minutes , followed by measuring optical density at 600nm . Cells were pelleted by centrifuging the cultures for 5 minutes at 3700×g . The resulting supernatants were removed and filtered through 0 . 2μm syringe filters . The cell pellets were resuspended in equal volume of PBS . To obtain cell lysate , triton X-100 was added to 1ml of the resuspended cells to a final concentration of 1% . Additionally , to minimize non-specific binding , triton X-100 was also added to 1ml of the filtered supernatant to a final concentration of 1% . 75μl of the supernatant , whole cells , lysates , were added to a Costar96 well flat white tissue culture treated plates and mixed with an equal volume of the Nano-Glo Extracellular Detection System reagent as specified in the manufacturer’s instructions . Additionally , media and PBS samples were used as controls . Reactions were incubated at room temperature for 10 minutes followed by measuring luminescence on a Tecan Spark with an integration time of 2000 milliseconds . All chinchilla experiments were conducted with the approval of Allegheny-Singer Research Institute ( ASRI ) Institutional Animal Care and Use Committee ( IACUC ) A3693-01/1000 . Research grade young adult chinchillas ( Chinchilla lanigera ) weighing 400-600g were acquired from R and R Chinchilla Inc . , Ohio . Chinchillas were maintained in BSL2 facilities and experiments were done under subcutaneously injected ketamine-xylazine anesthesia ( 1 . 7mg/kg animal weight for each ) . Chinchillas were infected with 100 CFUs in 100μL of S . pneumoniae PN4595-T23 by transbullar inoculation within each middle ear . For RNA extraction , chinchillas were euthanized 48h post-inoculation of pneumococcus , and a small opening was generated through the bulla to access the middle ear cavity . Effusions were siphoned out from the middle ear and flash frozen in liquid nitrogen to preserve the bacterial RNA . Animals were euthanized by administering an intra-cardiac injection of 1mL potassium chloride after regular sedation . The role of briC in experimental pneumococcal colonization was assessed as previously described [71] , [72] . For this , 10 weeks old female CD1 mice ( Charles River ) , weighing approximately 30-35g were anesthetized with 2 . 5% isoflurane over oxygen ( 1 . 5 to 2 liter/min ) , and administered intranasally with approximately 1X105 CFU/mouse in 20μl PBS . At predetermined time intervals , a group of 5 mice were euthanized by cervical dislocation , and the nasopharyngeal lavage of each animal was obtained using 500μl PBS . The pneumococci in nasopharyngeal wash were enumerated by plating the serial dilutions onto blood agar plates . The statistical differences among different groups were calculated by performing ANOVA followed by Tukey’s post-test , unless stated otherwise . p-values of less than 0 . 05 were considered to be statistically significant . To identify briC homologs we used tBLASTn with default parameters on the RAST database to search the genome sequences of all fifty-five strains . Predicted protein sequences were downloaded as well as nucleotide sequences for the briC homolog and 1500-bp flanking regions surrounding the briC homolog . Predicted protein sequences for BriC were aligned using NCBI Cobalt [73] and visualized using Jalview [74] . One sequences ( CDC3059-06 ) appeared to have a frame-shift after a string of guanines . Given that sequencing technologies are often inaccurate after a string of identical bases , we curated this sequence in the dataset . The sequences were translated in Jalview and organized based on polymorphisms in the translated sequences . The briC alleles were then organized in the context of the species tree . For this we used a published phylogenetic tree [34] , [38] . As previously described , the whole genome sequence ( WGS ) for these strains were aligned using MAUVE [75] , [76] , the core region was extracted and aligned using MAFFT ( FFT-NS-2 ) [77] . Model selection was performed using MODELTEST [78] , and the phylogenetic tree was built with PhyML 3 . 0 [79] , model GTR+I ( 0 . 63 ) using maximum likelihood and 100 bootstrap replicates . On the visualization , each allelic type is shape-coded , and the visualization was generated using the Interactive Tree of Life ( iTOL ) [80] . Next , we expanded the search to a set of 4 , 034 genomes . These correspond to the genomes within pubMLST , with at least 2Mb of genomic data ( Genome IDs are listed in S6 Table ) . We used BLASTn to search for genomes that encode sequences that are at least 70% identical over 70% of their length to briC alleles 1A or 1B . The 3 , 976 hits were organized to parse out and enumerate the unique sequences using Python . Next , the hits were visualized and further annotated using Jalview [74] . As in the smaller genomic set , one allele representative appeared to have a frame-shift after a string of guanines and was curated in the dataset . Next the DNA sequences were translated , and the predicted protein sequences were organized to display the unique alleles . The resulting 19 coding sequence were colored in Jalview based on percent identity to highlight the variability ( Fig 6 ) . To search for briC in related species , we performed a BLASTp analysis in NCBI . We used alleles 1A and 1B as query sequences , default parameters , and the non-redundant database excluding Streptococcus pneumoniae ( taxid: 1313 ) . In order to examine the structure of the promoter region upstream of the briC gene , a 1500-bp flanking region on both sides of the briC gene was pulled from the RAST database [81] . Sequences were aligned using Kalign [82] and then visualized with Jalview [74] . The alignment revealed two clear groups within the dataset: those with the RUP insertion and those without . We also noted that CDC1087-00 may have an additional mobile element inserted within the RUP . However , given that the RUP and this mobile element exist in multiple places in the genome , we cannot determine whether this is real or an artifact of assembly without the isolate . Thus , we opted not to use the promoter sequence for the consensus in Fig 2A , and we did not mark this genome as having a long promoter in Fig 5 . We marked the species tree with allelic variants that contain the RUP insertion . We observed that RUP was present in the representative isolates from two clinically important lineages PMEN1 and PMEN14 . To check the distribution of the long promoter in a larger set strains , we used PubMLST [39] to inspect 4 , 034 sequences with complete genomes ( sequence IDs for these 4 , 034 sequences are listed in S6 Table ) . This set includes 198 ST81 ( PMEN1 ) , as well as 104 ST236 ( PMEN14 ) and 15 ST320 ( PMEN14 ) strains . In the PMEN1 strains ( ST81 ) , one genome encodes the short promoter and 178 genomes encode the long promoter . In the PMEN14 strains , none of the strains encode the short promoter , and 71 out of 104 ST236 strains and 14 out of 15 ST320 strains encode the long promoter . Manual curation of selected genomes from the remaining set did not reveal a third promoter . Instead , we captured contig breaks and likely issues with assembly that we deduce are linked to repetitive nature of the RUP . Thus , we conclude that the long promoter is present in the vast majority of the PMEN1 and PMEN14 isolates . For analysis of the ComE-binding box , the ComE consensus sequence was extracted from the promoter regions of the pneumococcal strains and aligned with Jalview . The logo was generated using WebLogo [83] . Mouse experiments were performed at the University of Leicester under appropriate project ( permit no . P7B01C07A ) and personal licenses according to the United Kingdom Home Office guidelines under the Animals Scientific Procedures Act 1986 , and the University of Leicester ethics committee approval . The protocol was agreed by both the U . K . Home Office and the University of Leicester ethics committee . Where specified , the procedures were carried out under anesthetic with isoflurane . Animals were housed in individually ventilated cages in a controlled environment and were frequently monitored after infection to minimize suffering . Chinchilla experiments were performed at the Allegheny-Singer Research Institute ( ASRI ) under the Institutional Animal Care and Use Committee ( IACUC ) permit A3693-01/1000 . Chinchillas were maintained in BSL2 facilities , and all experiments with chinchillas were done under subcutaneously injected ketamine-xylazine anesthesia ( 1 . 7mg/kg animal weight for each ) . All chinchillas were maintained in accordance with the applicable portions of the Animal Welfare Act , and the guidelines published in the DHHS publication , Guide for the Care and Use of Laboratory Animals .
|
Pneumococcal biofilms occur in chronic otitis media , chronic rhinosinusitis , and nasopharyngeal colonization . These biofilms are an important component of pneumococcal epidemiology , particularly in influencing transmission , maintenance of asymptomatic colonization , and development of disease . The transcriptional program initiated via signaling of the competence pathway is critical for productive biofilm formation and is a strong contributor of pneumococcal infection and adaptation . In this study , we have identified BriC , a previously uncharacterized peptide that serves as a bridge between the competence pathway and biofilm development . We show that briC is induced by ComE , the master regulator of competence , and promotes biofilm development . Moreover , our studies in the murine model demonstrate that BriC is a novel colonization enhancer . Our studies of briC regulation capture an instance of genomic plasticity , where natural variation in the briC promoter sequence reveals the existence of an additional competence-independent regulatory unit . This natural variation may be able to modify the extent to which competence contributes to biofilm development and to nasopharyngeal colonization across different pneumococcal lineages . In summary , this study introduces a colonization factor and reveals a molecular link between competence and biofilm development .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"&",
"methods"
] |
[
"bacteriology",
"biofilms",
"chinchillas",
"medicine",
"and",
"health",
"sciences",
"pathology",
"and",
"laboratory",
"medicine",
"pneumococcus",
"pathogens",
"microbiology",
"vertebrates",
"animals",
"mammals",
"genome",
"analysis",
"sequence",
"motif",
"analysis",
"bacteria",
"bacterial",
"pathogens",
"research",
"and",
"analysis",
"methods",
"sequence",
"analysis",
"genomics",
"sequence",
"alignment",
"bioinformatics",
"medical",
"microbiology",
"streptococcus",
"microbial",
"pathogens",
"biological",
"databases",
"bacterial",
"biofilms",
"rodents",
"eukaryota",
"sequence",
"databases",
"database",
"and",
"informatics",
"methods",
"genetics",
"biology",
"and",
"life",
"sciences",
"computational",
"biology",
"amniotes",
"genomic",
"databases",
"organisms"
] |
2018
|
Function of BriC peptide in the pneumococcal competence and virulence portfolio
|
DNA double-strand breaks ( DSBs ) , which are formed by the Spo11 protein , initiate meiotic recombination . Previous DSB-mapping studies have used rad50S or sae2Δ mutants , which are defective in break processing , to accumulate Spo11-linked DSBs , and report large ( ≥ 50 kb ) “DSB-hot” regions that are separated by “DSB-cold” domains of similar size . Substantial recombination occurs in some DSB-cold regions , suggesting that DSB patterns are not normal in rad50S or sae2Δ mutants . We therefore developed a novel method to map genome-wide , single-strand DNA ( ssDNA ) –associated DSBs that accumulate in processing-capable , repair-defective dmc1Δ and dmc1Δ rad51Δ mutants . DSBs were observed at known hot spots , but also in most previously identified “DSB-cold” regions , including near centromeres and telomeres . Although approximately 40% of the genome is DSB-cold in rad50S mutants , analysis of meiotic ssDNA from dmc1Δ shows that most of these regions have substantial DSB activity . Southern blot assays of DSBs in selected regions in dmc1Δ , rad50S , and wild-type cells confirm these findings . Thus , DSBs are distributed much more uniformly than was previously believed . Comparisons of DSB signals in dmc1 , dmc1 rad51 , and dmc1 spo11 mutant strains identify Dmc1 as a critical strand-exchange activity genome-wide , and confirm previous conclusions that Spo11-induced lesions initiate all meiotic recombination .
Meiosis results in the faithful and efficient division of a diploid genome into four haploid gametes . After one round of DNA replication , cells undergo two rounds of chromosome segregation . Recombination between homologous chromosomes ( homologs ) occurs during prophase of the first division . Meiotic recombination promotes genetic diversity , but its main role is to ensure interhomolog association during the first meiotic division [1] . This association is absolutely required for efficient homolog separation , and defects in meiotic recombination result in chromosome nondisjunction [2] . Meiotic recombination is initiated by DNA double-strand breaks ( DSBs ) [3] . DSBs are formed by Spo11 , a homolog of the catalytic subunit of a type II DNA topoisomerase [4 , 5] . Spo11 is conserved among eukaryotes , and loss-of-function Spo11 mutants have been shown to be meiotic recombination–defective in many organisms [6–11] . DSBs form by a mechanism that involves the covalent attachment of Spo11 to break ends [5 , 12] . Subsequent to DSB formation , Spo11 is removed by endonucleolytic cleavage [13] , and break ends undergo 5′ to 3′ resection to create 3′ end single-strand tails [14] . This produces a substrate for Dmc1 and Rad51 , which are eukaryotic RecA homologues that catalyze the strand-invasion step of meiotic DSB repair by interhomolog recombination [15 , 16] . Dmc1 is expressed only during meiosis and is responsible for the bulk of meiotic DSB repair , whereas Rad51 is required for homologous recombination during vegetative growth and also contributes to meiotic recombination [17 , 18] . Meiotic DSBs form in early meiosis I prophase , after premeiotic S phase [19] . DSB formation appears to be co-regulated with DNA replication in two ways . Replication and DSB formation both require active cyclin-dependent kinase ( Cdc28 ) and the B-type cyclin Clb5 [20–23] . DNA replication and DSB formation also are temporally linked at the chromosome level , in that delaying replication of the left arm of chromosome III ( chr III-L ) causes a similar delay in DSB formation specifically on that chromosome arm [19] . Although it remains to be determined how DSB formation can be temporally linked to replication , it also appears that DSBs can form in the absence of DNA replication , because mutants lacking the replication licensing factor Cdc6 form meiotic DSBs in the absence of bulk DNA replication [24] . Hochwagen and Amon have proposed that replication initiation activates a checkpoint system that prevents DSB formation in unreplicated DNA [25] . Certain point mutations in RAD50 ( rad50S ) and in MRE11 ( mre11–58 and mre11S ) , as well as deletions of the SAE2/COM1 gene , which encodes a protein that appears to regulate activity of the Mre11/Rad50/Xrs2 complex , have been widely used in characterizing early steps in DSB formation and in determining DSB distributions [3 , 26–31] . In these mutants , hereafter referred to as rad50S-like , DSBs accumulate unprocessed and unrepaired , with Spo11 covalently attached [5] , which has facilitated the chromosome- and genome-wide DSB mapping ( [32–34] and reviewed in [31 , 35] ) . In these rad50S-like mutants , DSB hot spots are distributed unevenly . Chromosomes are partitioned into large ( ≥50 kb ) domains with many DSB hot spots , alternating with domains of similar size where DSBs are reduced or absent , even though potential DSB sites ( i . e . , chromatin nuclease hypersensitive sites [36 , 37] ) are present [38] . These “DSB-cold” regions are generally found at chromosome ends and adjacent to centromeres , but occur at many other locations as well [33 , 34] . Schizosaccharomyces pombe rad50S mutants also show a nonuniform DSB map , with most breaks occurring at sites in tight ( <3 kb ) clusters separated by ∼50 kb DSB-cold regions [39 , 40] . Two independent observations suggest that studies using rad50S-like mutants have underestimated DSB levels in S . cerevisiae . First , although DSBs rarely occur in a 30-kb centromere-proximal region of chromosome III , both the standard genetic map [41] and genetic studies [38] indicate that crossovers frequently occur in this region . Second , delaying DSB formation by 1 h on chr III-L causes a 4- to 5-fold reduction in DSB levels on that chromosome arm in rad50S-like mutants , but not in wild-type cells [19] . These findings suggest that DSB maps from rad50S-like mutants underestimate true DSB frequencies in some parts of the genome . To test this suggestion , we analyzed DSB distributions in mutants lacking Dmc1 or both Rad51 and Dmc1 strand-exchange proteins . In these mutants , Spo11 is removed from break ends , and unrepaired DSBs accumulate with 3′-ended ssDNA tails [17] . We reasoned that purification of this ssDNA could be used to enrich break-adjacent sequences , and thus to map DSBs at a whole-genome level . Using this new mapping strategy , we obtained a whole-genome distribution of meiotic DSBs that is considerably more uniform than was previously described , with substantial DSB levels in regions previously thought to be DSB-free . Our whole-genome DSB analysis also confirms that Spo11 forms all the lesions that initiate meiotic recombination , and that Dmc1 is a critical meiotic recombinase in all regions of the genome .
Borde et al . showed that deleting all active replication origins from chr III-L caused a delay in DSB formation in wild-type cells but did not alter DSB levels . By contrast , the same origin-deleted chr III-L showed a 4- to 5-fold reduction in DSBs in the rad50S-like mutant sae2Δ , which is unable to remove Spo11 from DSB ends [19] . We extended this analysis to dmc1Δ mutants , where DSBs accumulate at a stage after Spo11 is removed from break ends . Southern blots of pulsed-field gels were used to detect DSBs along the entire chromosome ( Figure 1A ) . In agreement with previous data , late DSB formation on chr III-L was associated with a 4-fold reduction in DSBs in sae2Δ cells . In contrast , dmc1Δ mutants showed similar DSB frequencies on normal and DSB-delayed chromosome arms . This observation suggests that DSB levels measured in dmc1Δ mutants might better represent recombination activity in wild type . Consistent with this suggestion , wild-type cells showed similar frequencies of crossing-over in wild type and DSB-delayed chr III-L ( Figure 1B ) . Meiotic intragenic recombination on chr III-L has also been shown to be independent of DSB timing [19] . Comparison of DSB patterns on normal chromosomes III in sae2Δ and dmc1Δ strains ( Figure 1A ) revealed two other differences . First , the fraction of chromosomes that suffer DSBs was greater in dmc1Δ ( 75%–80% ) than in sae2Δ ( 50%–60% ) , and the total DSB frequency ( 1 . 2 DSBs/chromosome ) in dmc1Δ was substantially greater than in sae2Δ ( 0 . 8 DSBs/chromosome ) . Second , DSBs formed in the center of the chromosome ( region III ) in about 20% of chromosomes in dmc1Δ cells , but in almost none in sae2Δ ( Figure 1A , previously reported by Blat et al . [42] and by Dresser et al . [43] ) . The substantial DSB signal seen in region III in dmc1Δ mutants is sufficient to account for the meiotic recombination observed in this region in wild-type cells ( Figure 1B ) , in contrast to the very low frequency of DSBs seen in this region in rad50S-like mutants ( [32 , 34 , 42] , Figure 1A ) . The above data indicate that more DSBs are formed in dmc1Δ mutants than in rad50S-like mutants , and that DSBs are artificially low in some regions in rad50S-like mutants . We therefore developed a strategy to measure DSB levels in recombinase-deficient strains , taking advantage of the ssDNA that accumulates on each side of DSBs in dmc1Δ mutants [17] and in rad51Δ dmc1Δ mutants [44 , 45] . Benzoyl naphthoyl DEAE ( BND ) cellulose , which selectively binds ssDNA [46–48] , was used to enrich for these DSB-associated sequences ( see Materials and Methods ) , which were compared to DSB-associated sequences prepared from rad50S mutants by immunoprecipitation of Spo11-linked DNA [33 , 34] . Quantitative PCR analysis of DNA prepared by both methods ( Figure 2 ) showed similar enrichment at two DSB hot spots ( YCR047c and YGR176w [32 , 34 , 49] ) relative to ribosomal RNA genes , where meiotic DSBs are absent [50] and meiotic recombination is infrequent [51 , 52] . No ssDNA enrichment was seen in the DSB-negative spo11-Y135F dmc1Δ and spo11Y135F rad51Δ dmc1Δ mutants , indicating that all meiotic ssDNA at these sites originates from Spo11-induced DSBs . We mapped break-associated DNA sequences with oligonucleotide-based microarrays [53] that offer greater resolution ( average interelement distance of 290 nucleotides [nt] ) as well as more uniform element-to-element hybridization properties than do the spotted PCR-product microarrays used previously [33 , 34] . To allow direct comparison between different array datasets , we developed a background-based normalization procedure , rather than the more commonly used median normalization , because the latter method results in an artifactual lowering of array signals when a positively skewed experimental signal ( typical of signals from chromatin immunoprecipitation [ChIP]-chip experiments ) is compared to a more symmetrically distributed background signal [54] . Enrichment values ( Table S1 ) were background-normalized using the median signal from a set of probes located >2 kb from either end of the coding sequences of 19 large genes ( see Materials and Methods , Table S2 ) . Because the vast majority of DSBs occur in promoter regions [32 , 37 , 55] , these probes are unlikely to be present in either meiotic ssDNA or in Spo11-associated DNA . Background normalization resulted in datasets with very similar dynamic ranges , irrespective of the DSB enrichment method used , either ssDNA from dmc1Δ or Spo11 ChIP material from rad50S ( Figure 3 , Tables S1 and S7 , and Figures S1–S3 ) . Similar DSB signals were seen in both dmc1Δ and in rad50S at three of the strongest previously identified DSB hot spots ( Figure 3A ) [32 , 34] , and all of the top 50 DSB hot spots in rad50S were also present among the dmc1Δ hot spots ( Table S3 ) . The current oligonucleotide ( oligo ) -array analysis of Spo11-linked DNA from rad50S is in good agreement with previous analyses using PCR-product arrays [33 , 34] . Whereas Borde et al . identified 585 DSB hot spots in sae2Δ mutants [34] , the increased dynamic range and resolution of oligo-arrays allowed identification of 1 , 306 DSB hot spots ( peak values of 2–30 times background ) in rad50S samples , and these included most of the previously identified hot spots ( Tables S4 and S5 ) . To ask if all ssDNA detected in a dmc1Δ mutant was DSB-associated , we first examined the Spo11-dependence of the meiotic ssDNA signal . No enrichment of ssDNA above background was observed in any region in a spo11Y135F dmc1Δ mutant ( Figure 3 , Table S1 , and Figure S3 ) , consistent with Spo11-catalyzed DSBs being the primary lesion-initiating meiotic recombination genome-wide . We also asked if all DSBs repairable by homologous recombination are detected in dmc1Δ mutants . Dmc1 and Rad51 both catalyze strand exchange during meiosis , and Dmc1 has been shown to be responsible for the bulk of meiotic DSB repair at a few defined DSB sites [17 , 18] . We asked if any regions of the genome contained substantially more ssDNA in rad51Δ dmc1Δ than in dmc1Δ strains , as might be expected if the majority of DSBs in some regions were repaired in a Rad51-dependent , Dmc1-independent manner . Similar DSB distributions were seen in the two mutant backgrounds ( Tables S1 and S7 , Figure S2 ) , and no DSB peaks were 2-fold greater in rad51Δ dmc1Δ than in dmc1Δ ( Figure S2 , Table S1 , and unpublished data ) . In addition , substantially lower ssDNA and DSB levels were detected in meiotic DNA sample taken at the same time in meiosis from a rad51Δ single mutant ( Tables S1 and S7 , Figure S4 , and unpublished data ) . These results indicate that Dmc1 is critical for the majority of meiotic DSB repair reactions in all regions of the genome . For this reason , further ssDNA distribution analysis will use data from dmc1Δ single mutants . Although some DSB hot spots display similar enrichment in dmc1Δ- and rad50S-derived material , disparity between the two mutants is seen in many parts of the genome ( Figure 3 and Figure S1 ) . The majority of nonbackground array elements display greater enrichment in dmc1Δ than in rad50S ( Figure S5 and Table S7 ) ; in 40% of array elements , the dmc1Δ/rad50S signal ratio was greater than 2 ( Figure S5 ) . In addition , DSB-cold regions were notably absent from dmc1Δ mutants . We identified 260 DSB-cold regions longer than 10 kb ( all elements <2× background ) in rad50S , representing about 4 . 8 Mb , or 40% of the single-copy genome . Only 28 of these regions ( about 370 kb , or 3% of the single-copy genome ) were also DSB-cold in dmc1Δ mutants ( Table S1 and Figure S5 ) . The majority of the >10-kb regions were at loci expected to be DSB-cold: 14 were near chromosome ends; eight contained very large open reading frames , and one contained a centromere . Thus , a substantially greater fraction of the genome is DSB-associated in dmc1Δ than in rad50S mutants , and only a very small fraction of regions can be described as DSB-cold . The discordance between dmc1Δ and rad50S is further illustrated by an examination of the number of DSB hot spots/genome and inter–hot spot distances at different peak intensity thresholds ( Figure 4A and Table S5 ) . At all peak intensities , the number of DSB hot spots in dmc1Δ exceeded the number in rad50S . When the strongest hot spots are considered ( peak/background > 5 ) , approximately five times more hot spots are present in dmc1Δ than in rad50S ( Figure 4A ) . At a lower threshold ( peak/background > 2 ) , about twice as many hot spots are present in dmc1Δ as are in rad50S . This convergence is consistent with the suggestion that DSBs are formed at the same sites in both mutant backgrounds , but the DSB intensity in rad50S is substantially less than in dmc1Δ at many sites ( [19] , see also Figures 1 and 5 ) . As expected from the greater DSB hot spot density , inter–hot spot distances in dmc1Δ are substantially less than in rad50S ( Table S5 ) . At twice the background threshold , the mean inter–hot spot distance is about 5 . 5 kb in dmc1Δ but about 8 . 5 kb in rad50S . This discordance is even greater with stronger DSB hot spots ( ≥5 × background ) , with a mean interpeak distance of 9 . 5 kb for dmc1Δ and about 35 kb for rad50S . This discordance is reflected in differences in calculated fractions of the genome within a given distance of the nearest DSB ( Figure 4B ) . In dmc1Δ , more than 70% of the single-copy yeast genome is less than 2 . 5 kb from a DSB peak that is twice background , as compared to less than half of the genome in rad50S . Taken together , these observations clearly indicate that DSBs occur more frequently , and are more uniformly distributed , in dmc1Δ than in rad50S . Studies of rad50S-like mutants indicate that DSBs are absent from sequences within 20 kb of centromeres and within 40–50 kb of chromosome ends [33 , 34] . The increased DSB density in dmc1Δ prompted a re-examination of DSB signals near centromeres and chromosome ends ( Figure 4C and 4D ) . In contrast to the 20-kb centromere-adjacent DSB-cold regions seen in previous studies , we observed significantly reduced DSB signals only in an 8-kb or 10-kb window for dmc1Δ and rad50S , respectively ( Figure 4C ) . However , individual chromosomes display DSB peaks within this centromere-proximal zone ( Figure 3 , Figure S1 , and unpublished data ) [56] , and average dmc1Δ element signals in the 2 kb immediately centromere-proximal are significantly greater than background ( p < 0 . 001 , Mann-Whitney test ) . Taken together , these data suggest that DSBs are absent from centromeres themselves , with the likelihood of DSB formation rising with distance over the next 8–10 kb . DSB signals are also significantly reduced , relative to the genome-wide average , in a ∼60-kb region at chromosome ends ( as defined by the S . cerevisiae reference sequence ) . Average DSB signals in the ∼20 kb closest to chromosome ends are close to background . In dmc1Δ , the DSB signal is ∼2/3 of the genome-wide average in the next 20 kb . In the region 40–60 kb from chromosome ends , DSBs in both dmc1Δ and rad50S approach , but are still significantly below , the genome-wide average ( 80%–90% , p <0 . 0001 , Mann-Whitney test ) . These data indicate that DSBs are absent from the 20 kb closest to most chromosome ends , and are present at modestly reduced frequencies in the next 20 kb . These conclusions must be conditioned by the fact that uncertainty exists regarding the precise distance of many array elements from chromosome ends in our study . In particular , sequences near some chromosome ends are known to differ between the strain in which this study was performed ( SK1 ) and S288c , the reference sequence strain used in microarray design ( [32] , E . J . Louis , personal communication ) . Southern blot analysis of DSBs in selected regions confirmed the conclusion , from whole-genome data , that regions exist where DSB levels in dmc1Δ are substantially greater than in rad50S . Similar DSB frequencies were measured at the YCR047c hot spot in dmc1Δ and rad50S , both on arrays ( Figures 3A and 5A ) and on Southern blots ( Figure 5E ) , although DSB fragments differed in size , due to the 5′-to-3′ single-strand resection that occurs in dmc1Δ but not in rad50S . Agreement between array and Southern-based DSB frequencies in dmc1Δ and rad50S was also observed at a second DSB hot spot ( YDR187c; Figure S6 ) . Southern blot analysis at several loci also confirmed the regional discordance between dmc1Δ and rad50S microarray data . In the central region of chromosome III , significant DSB levels were detected near the promoter regions of YCR011c , YCR020c , and YCR022c in dmc1Δ but not in rad50S , in both microarrays ( Figure 5A ) and Southern blots ( Figure 5C and 5D ) . A 3-fold difference between dmc1Δ and rad50S in DSBs in the region surrounding YCL011c seen in microarrays ( Figure 5A ) was also confirmed on Southern blots ( Figure 5B ) . Similar validation was obtained from analysis of other discordant loci , including a 30-kb region at the end of chromosome XIII and at YIR020c ( Figure S6 ) . The above Southern blot analysis confirms the microarray-based identification of regions of discordance between rad50S and dmc1Δ DSB maps . In one such region—the centromere-proximal region of chromosome III—genetic measures of recombination in wild type are more consistent with the DSB levels seen in dmc1Δ than in rad50S-like mutants ( Figure 1; see also [38] ) . It was therefore of interest to determine whether breaks were frequent or infrequent in wild-type cells in regions of discordance between rad50S- and dmc1-based DSB map . We analyzed DSB levels and timing in wild-type cells in the YCR047c concordant regions and in three discordant DSB sites ( YOR347c , YLR436c , and YDL220c ) , where breaks are present in dmc1Δ and almost absent in rad50S ( Figure 6 and Figure S7 ) . Substantial DSB levels were detected in wild type at all four loci . DSB peaks occurred at 3–3 . 5 h after initiation of sporulation at all loci; DSBs tended to occur at discordant loci later than at the concordant locus , although asynchrony in DSB formation and culture-to-culture variation precludes accurate temporal assignment ( Figure 6 , Figure S7 , and unpublished data ) . These data confirm the conclusion that dmc1Δ mutants are better than rad50S-like in predicting whether or not DSBs occur at a given locus in wild type .
All known homologous recombination mechanisms produce ssDNA , which is bound by RecA-like strand-exchange proteins and used to initiate recombination by invading a second duplex DNA molecule [15] . In mutants lacking RecA-like proteins , lesion-associated ssDNA is expected to accumulate . Approaches that detect this ssDNA should therefore detect the ensemble of recombination initiation events , irrespective of mechanism , as long as the ssDNA formed is stable . We have presented here the mapping and quantification of meiotic DSBs in S . cerevisiae , based on microarray analysis of break-associated ssDNA isolated by BND cellulose enrichment . A similar approach has been used by Blitzblau and coworkers , with similar conclusions [56] . This method has also been used to identify ssDNA regions that accumulate when mitotic replication is blocked [48] . Given the sensitivity and selectivity of this method ( >200-fold , Figure 2 ) , it should provide a powerful way to detect lesions that initiate homologous recombination , and should be generally applicable to many organisms , including mammals . This method provides a powerful way to detect early recombination intermediates , but care must be taken when interpreting results . First , different ssDNA-containing intermediates may not be equally stable . For example , only recombination can repair ssDNA tracts with 3′ ends , the primary precursor in meiosis [14 , 55 , 57] . On the other hand , ssDNA with 5′ ends or ssDNA gaps can be filled by DNA polymerases . Such lesions might not persist as ssDNA , and thus would have been under-detected in our assay . Second , since 5′ to 3′ resection continues over time in dmc1Δ mutants [17 , 18] , early-forming DSBs might be associated with more ssDNA than late-forming DSBs , thus giving a relatively stronger signal on arrays . However , since all DSBs we examined in dmc1Δ display a similar resection size ( Figure 6 , Figure S7 , and unpublished data ) , we believe that a biased representation of early and late breaks is unlikely . The whole-genome analysis of meiotic DSB distributions shows that some DSB hot spots show similar signals in dmc1Δ and in rad50S , but a greater number display a dmc1Δ/rad50S signal ratio of 2-fold or greater , and many DSB sites are detected in dmc1Δ but not in rad50S ( Figure 4 , Figures S1 and S5 , and Table S1 ) . This finding is confirmed by Southern blot analysis of DSBs in selected regions in dmc1Δ and rad50S ( Figure 5 and Figure S6 ) with a linear relationship between DSB frequencies on Southern blot and on microarray ( Figure S8 ) . Break processing–capable but repair–defective mutants such as dmc1Δ are also better predictors of DSB locations in wild type than rad50S-like mutant ( Figure 6 and Figure S7 ) and DSB frequencies in dmc1Δ agree with genetic distances measured at the chromosomal level ( Figure 1 ) . In addition , integrated whole-genome DSB signals from microarray analysis of dmc1Δ and rad51Δ dmc1Δ predict 140–170 DSBs/meiotic nucleus ( Table S7 ) , which approaches the genetic map-based estimate of 180–270 DSBs/nucleus [58] . This stands in contrast to the much lower estimate of about 44 DSBs/nucleus from rad50S data ( Table S7 ) . We conclude that the distribution of ssDNA-enrichment signals in DNA from dmc1Δ mutants is currently the most accurate representation of the relative distribution of DSBs in wild-type cells , although it is likely that it underestimates true DSB frequencies . The use of rad50S-like mutants to map DSBs in budding and fission yeast has resulted in a DSB landscape that is dominated by DSB hot spot clusters separated by 50–200-kb cold regions [32 , 34 , 39 , 40] . In these maps , DSBs are reduced or absent from large ( ∼40 kb ) regions at chromosome ends and near centromeres [32 , 34] . Although a similar DSB pattern is seen in rad50S mutants in this study ( Figures 1 and 2 and Figure S4 ) , a very different DSB map emerges when meiotic ssDNA is analyzed in dmc1Δ mutants ( Figures 3 and 4 and Figures S1 and S2 ) . About twice as much of the genome displays a significant DSB signal ( 2× background ) in dmc1Δ than in rad50S ( 70%–80% versus 35% ) , and the overall DSB signal in dmc1Δ is about twice that seen in rad50S ( Figure S5 and Table S7 ) . This conclusion is confirmed by Southern blot studies and by pulsed-field gel analysis of DSBs on normal chromosomes III , where DSBs are substantially more as frequent in dmc1Δ than in rad50S ( Figure 1 ) . As a consequence , the S . cerevisiae genome can no longer be described as being composed of large DSB “hot” and “cold” regions . Instead , recombination initiation events are more broadly distributed , with the majority ( >70% ) of loci being within 2 . 5 kb of DSB hot spots with detectable break frequencies ( Figure 4B ) . Because of the increased number of DSB hot spots detected in dmc1Δ mutants , it will be important to revisit the sequence , chromatin and chromosome context determinants of DSB hot spots described from a subset of dmc1Δ-rad50S conserved DSBs hot spots [32 , 33 , 35 , 42 , 59–61] . Our current analysis is of insufficient resolution to test suggestions involving chromatin structure or specific sequences . The increased number and intensity of DSB in dmc1Δ relative to rad50S also makes it important to revisit previous studies that used rad50S-like mutants to examine the influence of factors such as global transcription regulators and chromatin modifiers on meiotic DSB locations and levels [35 , 38 , 62 , 63] . These studies examined only a subset of the DSBs that are formed during normal meiosis , and effects detected in these studies might have involved factors that impact the nonphysiological under-representation of some DSBs in rad50S-like mutants . Replication timing is one factor that has been shown to affect DSB levels in rad50S-like mutants , but not in wild type or dmc1Δ ( [19] , see Figure 1 ) . It is unlikely that differences in replication timing can account for all of the differences that we observe between dmc1Δ and rad50S DSB distributions . Mitotic replication timing patterns [64 , 65] do not correspond well with genome-wide patterns of concordance and discordance between dmc1Δ and rad50S ( unpublished data ) . Furthermore , concordant and discordant DSB sites can be found within a single 10-kb region , a distance that is considerably less than the distances ( ∼40 kb ) over which substantial differences in replication timing are observed ( see Figure S5 ) . It will be of considerable interest to identify the factors that determine why , at many sites , DSBs are recovered frequently at in dmc1Δ and wild type , but infrequently in rad50S-like mutants . While we have assumed that this discordance reflects a failure to form DSBs at some sites in rad50S-like mutants , it is possible that breaks are formed transiently at some sites but rapidly repaired without resection , perhaps by reversal of the initial Spo11 cleavage reaction ( e . g . , [66] ) . Previous studies using rad50S-like mutants suggested that DSBs are largely absent from regions within 50 kb of chromosome ends and in ∼40 kb regions around centromeres [33 , 34 , 62] , consistent with the need to exclude recombination from centromeres and chromosome ends to prevent chromosome segregation dysfunction [2 , 67 , 68] . The current analyses indicate that these DSB-cold regions are considerably smaller than previously suggested . While repression of meiotic recombination has been clearly documented for the chromosome III centromere ( [69–71] , T-C Wu , M . Lichten , unpublished data ) , our data indicate that this may not be true for all other chromosomes . Similarly , whereas the first ∼20 kb from chromosome ends are DSB cold , the following ∼30-kb display substantial DSB activity in single-copy sequences , a finding consistent with recent studies of crossing-over near chromosome ends ( A . Barton , D . Kaback , personal communication; S . Chen , J . Fung , personal communication ) . It should also be noted that ectopic meiotic exchange is reported to occur frequently between repeated genetic elements immediately adjacent to chromosome ends [72] , implying that DSBs can form in these elements as well . These repeated elements were not included in the microarrays used in our analysis . Previous studies have shown that most meiotic recombination is initiated by Spo11-catalyzed DSBs [3] , but this has not been confirmed on a genome-wide basis . In addition , studies at individual test loci [17 , 45 , 73] showed that meiotic DSB repair can occur in a Dmc1-dependent , Rad51-independent manner , but do not exclude the possibility that repair in other regions is Dmc1-independent and Rad51-dependent . We find no regions where ssDNA enrichment values in rad51Δ dmc1Δ are more than 2-fold greater than those in dmc1Δ ( Table S1 , Figure S3 ) . This confirms Dmc1 as a critical strand-exchange activity for meiotic DSB repair in all single-copy regions . This conclusion does not exclude an important role for Rad51 in meiotic recombination , but the observation of significantly lower levels of meiotic ssDNA in a rad51Δ single mutant ( Tables S1 and S7 and Figure S4 ) suggests that substantial meiotic DSB repair can occur in its absence during budding yeast meiosis , consistent with reports of meiosis-induced activities that inhibit Rad51 [74] . Similarly , Spo11-independent lesions such as nicks and/or DSBs , if processed to form substrates for strand invasion , should be detected as ssDNA in the absence of Dmc1 and Rad51 . We detected no meiosis-specific ssDNA enrichment in a spo11-Y135F dmc1Δ strain , either at three DSB hot spots ( Figures 2 and 3 ) or in the genome as a whole ( Figure S2 and Table S1 ) . This observation strongly supports the previous conclusions that Spo11-catalyzed DSBs initiate the vast majority of , if not all , meiotic recombination events . The general absence of large DSB-hot and DSB-cold regions that we observe in S . cerevisiae is consistent with the relatively uniform distribution of estimated crossover activity per unit distance over large intervals in the budding yeast genome as a whole [41] . It stands in contrast to the highly punctuated crossover maps , with pronounced recombination hot spots separated by 0 . 1–1 Mb of relatively inactive regions , in several multicellular organisms , in particular in those inferred from human linkage disequilibrium data [75–80] . Comparisons between these highly punctuated recombination maps and the DSB map from yeast rad50S-like mutants have suggested that the molecular-level yeast picture might be an appropriate paradigm for what occurs at the molecular level in other organisms; however , the absence of marked “hot” and “cold” regions in the new yeast DSB map indicates otherwise . If distance comparisons are made in terms of overall chromosome length , then the distribution of recombination hot spots in higher eukaryotes more closely resembles the relatively uniform distribution of DSB hot spots in S . cerevisiae . In addition , this latter distribution is consistent with the relatively uniform distributions of early cytological structures that are presumed to reflect early interhomolog interactions at the onset of crossover and noncrossover recombination events [81–84] . In particular , the use of DSB-mediated interhomolog interactions at sites of random collision to drive homolog pairing during early meiosis 1 prophase [57 , 85] , would be expected to select for a relatively uniform DSB distribution where the distance between hot spots scaled with chromosome size .
All S . cerevisiae strains used ( Table S6 ) are isogenic to the SK1 background [86] . All markers were introduced by transformation and by genetic crosses between transformants . Genetic distances were determined by tetrad analysis , using the Stahl lab calculator , available at http://rd . plos . org/pbio . 0050324 . To minimize ssDNA formation during purification , we used direct lysis in phenol/chloroform and digested DNA with restriction enzymes to produce fragments similar in size to those produced in chromatin immunoprecipitation procedures . Cells were harvested from 25 ml of a sporulation culture 5 h after initiation of sporulation , which was identified as a time when most DSB formation is complete by Southern blot–based studies of dmc1 mutants , as well as by cumulative curve analysis [87] of Southern blots of wild-type strains ( unpublished data , see also Figure 6 and Figure S7 ) . For a detailed protocol , see Protocol S1 . Pulsed-field gels used DNA prepared in agarose plugs as described [38] . Agarose gel electrophoresis , pulsed-field gel electrophoresis , Southern blotting , and hybridization with radioactive probe were as described [19] . Radioactive signal on filters was detected and quantified using a Fuji FLA 5100 scanner and ImageGauge 4 . 1 software . For pulsed-field gels in Figure 1 , densitometry traces were divided into ∼720 bins , and DSB frequencies were corrected to account for signal reduction due to coincident cutting in the bin in question and in probe-proximal bins , using the following formula: where DSBi = DSB signal in bin i . This correction assumes that DSBs are distributed randomly with respect to each other along chromosomes , so that for a chromosome with a DSB at locus i , the likelihood of additional DSBs between i and the chromosome end is the same as the likelihood of DSBs between i and the chromosome end in an unselected population of chromosomes . ChIP of hemagglutinin ( HA ) -tagged Spo11 from a rad50S strain was performed as described with only minor modifications , using cells from a culture 5 h after initiation of sporulation [33 , 34] . ssDNA enrichment on BND cellulose ( Sigma ) used batch purification [88]; for detailed protocols , see Protocol S1 . Enrichment of DSB-associated DNA was estimated by quantitative PCR ( details in Protocol S1 ) . Input and ssDNA-enriched material from BND cellulose fractionation and Spo11-ChIP fractions and whole-cell extracts from Spo11 immunoprecipitates were amplified using a previously described random priming amplification procedure with minor modifications [33] ( details in Protocol S1 ) . Background normalization of fluorescence signals in each channel was performed using a subset of probes for which the presence of meiotic ssDNA is unlikely ( Table S2 ) . Southern blot analyses and fine-structure mapping of DSB hot spots have shown that meiotic DSBs are generally absent from protein coding regions [32 , 55] . Given an average shear size of 1 kb in Spo11 ChIP and 1 kb resection tract in DNA from dmc1Δ mutants , we reasoned that the hybridization signal of array elements located at least 2 kb from the 3′ and 5′ ends of protein coding sequences are likely to represent background . The median fluorescence intensity of elements meeting this criteria , selected from the largest open reading frames in the budding yeast genome ( 294 array elements representing 0 . 7% of the total number of elements ) was used to normalize the fluorescence intensity for each spots in each channel . For each hybridization , we then calculated the ratio of the background-normalized Cy-5 ( experimental ) channel fluorescence versus the average of five independent background-normalized hybridizations with Cy-3 labeled genomic DNA . All experiments were performed in duplicate from independent cultures; the data presented in Table S1 are the average of the two independent ratios for each mutant background . MATLAB ( v . 7 . 4 . 0 ) , Microsoft Excel ( v . 11 . 3 . 3 ) , and Graphpad Prism ( v . 4 . 0b ) were used for computational and statistical analysis; program code will be supplied upon request . DSB hot spots were identified in microarray hybridization data using a MATLAB implementation of the PeakFinder program initially developed for PCR product arrays [89] . A seven-element running average was first applied to background-normalized ratios , and DSB peaks at different threshold levels were identified using the first derivative of this de-noised data . A seven-element running average results in a loss of some individual peaks , especially in gene-dense regions , but this window size is the smallest that avoids peak doubling , as the random labeling protocol used here results in two closely-spaced peaks flanking each DSB site . Peak coordinates determined for various thresholds were then used to calculate interpeak distances and the fraction of genome within a given distance from a DSB site .
The Entrez Protein ( http://www . ncbi . nlm . nih . gov/sites/entrez ? db=protein ) accession numbers for the proteins corresponding to the genes discussed in this paper are: CDC28 ( NP_009718 ) , CDC6 ( NP_012341 ) , CLB5 ( NP_015445 ) , DMC1 ( NP_011106 ) , MRE11 ( NP_013951 ) , RAD50 ( NP_014149 ) , RAD51 ( NP_011021 ) , recA ( AAQ91336 ) , SAE2 ( NP_011340 ) , SPO11 ( NP_011841 ) , YCL011c ( NP_009916 ) , YCR007c ( NP_009933 ) , YCR011c ( NP_009937 ) , YCR019w ( NP_009946 ) , YCR020c ( NP_009947 ) , YCR022c ( P25620 ) , YCR045c ( NP_009974 ) , YCR046c ( NP_009975 ) , YCR047c ( NP_009976 ) , YCR048w ( NP_009978 ) , YCR051w ( NP_009980 ) , YCR052w ( NP_009981 ) , YDL220c ( NP_010061 ) , YGR176w ( P32475 ) , YIR020c ( P40575 ) , YLR436c ( NP_013540 ) , YLR437c ( NP_013541 ) , YLR438w ( NP_013542 ) , YLR439w ( NP_013544 ) , YLR440c ( NP_013545 ) , YOR347c ( NP_014992 ) . The microarray data used in this paper are deposited at http://www . ncbi . nlm . nih . gov/geo/ with accession number GSE8981 .
|
During meiosis , the two copies of each chromosome present in the full ( diploid ) genome come together and then separate , forming haploid gametes ( sperm and eggs , in animals ) . Recombination , which swaps DNA between chromosomes , is critical for chromosome pairing and separation , and also promotes genetic diversity in the next generation , providing the feedstock for evolution . DNA double-strand breaks ( DSBs ) , which are formed by the conserved Spo11 nuclease , initiate meiotic recombination . DSB mapping is thus an alternative to standard genetic analysis for determining where meiotic recombination occurs . DSBs have been most extensively mapped in budding yeast mutants that fail to remove Spo11 from break ends , blocking further recombination steps . Paradoxically , those studies indicated that DSBs are absent from large regions where recombination was known to occur . We developed a new DSB mapping method that purifies and analyzes the single-strand DNA formed at breaks after Spo11 removal . This new map shows that DSBs ( and by inference , recombination ) actually occur frequently throughout almost all of the budding yeast genome , in a distribution that is consistent with recombination's roles in chromosome pairing and in generating genetic diversity . This new mapping method will be useful for studying meiotic recombination and DNA damage repair in other organisms .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods",
"Supporting",
"Information"
] |
[
"cell",
"biology",
"molecular",
"biology",
"genetics",
"and",
"genomics"
] |
2007
|
Mapping Meiotic Single-Strand DNA Reveals a New Landscape of DNA Double-Strand Breaks in Saccharomyces cerevisiae
|
Neural populations respond to the repeated presentations of a sensory stimulus with correlated variability . These correlations have been studied in detail , with respect to their mechanistic origin , as well as their influence on stimulus discrimination and on the performance of population codes . A number of theoretical studies have endeavored to link network architecture to the nature of the correlations in neural activity . Here , we contribute to this effort: in models of circuits of stochastic neurons , we elucidate the implications of various network architectures—recurrent connections , shared feed-forward projections , and shared gain fluctuations—on the stimulus dependence in correlations . Specifically , we derive mathematical relations that specify the dependence of population-averaged covariances on firing rates , for different network architectures . In turn , these relations can be used to analyze data on population activity . We examine recordings from neural populations in mouse auditory cortex . We find that a recurrent network model with random effective connections captures the observed statistics . Furthermore , using our circuit model , we investigate the relation between network parameters , correlations , and how well different stimuli can be discriminated from one another based on the population activity . As such , our approach allows us to relate properties of the neural circuit to information processing .
In the search for clues about the function of neural circuits , it has become customary to rely upon recordings of the activity of large populations of neurons . These measurements exhibit the concerted responses of neurons in different conditions , such as presentations of different stimuli . With the stimulus-dependent , high-dimensional statistics of neural responses in hand , one can ask two questions: How are these statistics generated in the neural population ? What purpose , if any , do they serve ? While the first question is mechanistic and the second is functional , the two are intimately linked . We address these questions by relating observed population activity to the output of different circuit models . In particular , we examine the correlated response variability that arises from diverse circuit architecture , and , in turn , its effect on stimulus representation . Along the mechanistic line of research , a number of network models have been proposed to explain the relation between anatomical and physiological parameters , on the one hand , and the noise correlations , on the other hand . The way in which correlations emerge , and may be suppressed , in recurrent networks was analyzed in detail with simplified biophysical models [1–4] , as well as more abstracted models using binary neurons [5] or Poisson neurons [6] . As an alternative to recurrent connections , shared external input ( e . g . , from top-down afferents ) or shared gain fluctuations can also be at the origin of correlations in neural populations . This and similar mechanisms were exploited in recent studies [7 , 8] , in particular to account for observations in both visual cortex [9–12] and retina [13 , 14] . In modeling stochastic neural activity , models of Poisson neurons or generalizations thereof have proved eminently useful . These have been employed to investigate experimentally measured statistics in the context of both feed-forward [15] and recurrent networks [16] , as well as in the context of models with shared gain fluctuations [9 , 13] . Poisson models can capture a wide range of statistics; their parameters can be viewed as an abstraction from the biophysical properties of neurons . An important motivation to understand the origin of the statistics of neural activity is the latter’s connection with the representation of sensory information in neural populations ( for a recent review , see Ref . [17] ) . A much studied example is that of the influence of recurrent connections on the properties of the mean response of neurons , or tuning curves [18–20] . Beyond the study of first-order statistics , a great deal of work has been devoted to the investigation of the relation between correlated variability in population response and the accuracy of stimulus representation . Several classic studies have pointed to the fact that noise correlations can be harmful to information coding [21–23] , while a number of more recent studies indicate that noise correlation can also have a beneficial impact on coding , depending on its fine structure [23–26] . For example , the effect of noise correlation can depend appreciable on its stimulus dependence [13 , 14] or on the physiological heterogeneity present in the population [27–29] . The way in which noise correlations arise from circuit properties has been examined in simulated networks [18] and analyzed theoretically in the framework of Poisson neurons [30 , 31] . Recent work has also investigated the limits on the representation of information imposed by properties of neural circuits , via the structure of correlations these induce [32 , 33] . Here , we aim at relating possible mechanistic origins of correlation to the statistics of neural population activity , on the one hand , and at relating circuit properties to the representation of information , on the other hand . Rather than starting from a specific circuit model or relying on simulations of a certain network architecture , we want to be able to infer circuit properties of fairly general models for large and interconnected populations from observed activity . We achieve these aims by establishing mathematical expressions that relate circuit properties to population activity statistics; in particular , to variances and covariances in the population . These expressions connect the various parts of our investigation . We obtain these expressions in the framework of the Hawkes model [34] , made up of coupled Poisson neurons . We examine three types of circuit architectures , namely , recurrent networks , feed-forward networks , and networks with shared gain fluctuations . Instead of focusing on specific realizations of these networks , however , we consider ensembles of such networks , defined through distributions of random connections . This allows us to identify generic signatures of circuit architectures in the correlation structures they induce , which can be used to distinguish them based on recorded activity . We then analyze experimental data , namely , populations recordings in mouse auditory cortex [35] , in the light of our model results , and we show that the structure of noise correlations in auditory cortex is in agreement with predictions from a recurrent network . Finally , we turn to the relation between circuit architecture and information coding . Using the derived mathematical expressions , we analyze the effect of network parameters on the representation of information , in the case of different circuit architectures .
The data set was first reported on in Ref . [35] which discussed the structure of the average population responses as a function of stimulus . The data consists of the activity of neural populations ( 46-99 neurons ) recorded with calcium imaging in the auditory cortex of mice . Animals were isoflurane anesthetized ( 1% ) . Signals were obtained from neurons labeled with the synthetic calcium indicator OGB1 . Fluorescence was measured at 30 Hz sampling rate , and firing rates were inferred from temporal deconvolution of the fluorescence signal . Up to 6 neural populations in each of 14 animals were recorded . The data points we use are the average firing rates over a window of 250 ms after presentation of each of 65 different sound stimuli , each for 15 trials . Sound stimuli consisted in 50-70 ms presentations of pure frequency tones or complex broadband sounds generated from animal calls or musical pieces . Onset and offset were smoothed , and the stimuli were played at different volumes . Thus , the data points represent population responses evoked by a short sound presentation , time-averaged over a 250 ms window . Responses were measured relative to spontaneous activity; hence , negative responses occurred if the stimulus-evoked firing rates were smaller than the spontaneous ones . While the precise timing of spikes can convey a large amount of information [36] , in many studies this temporal dimension is neglected , and the representation of stimuli is considered in terms of spike counts in given time windows . This approach is legitimate in cases in which sensory coding occurs on a relatively slow time scale; here , we also assume this form of spike count coding . We denote the vector of population activity in trial T of stimulus s is by r ( s , T ) , and the average response across trials for this stimulus by r ( s ) ≡ 〈r ( s , T ) 〉T . If not explicitly stated otherwise , here , and throughout , we consider time-averaged neural responses , both when referring to experimental data and in the theoretical analysis; i . e . , the vector r ( s , T ) represents the neural activity averaged across a chosen time window . To measure covariability between neurons i and j across trials , in response to a given stimulus , we use the noise covariances , defined as C i j ( s ) = cov ( r i ( s , T ) , r j ( s , T ) ) T = ⟨ r i ( s , T ) r j ( s , T ) ⟩ T - ⟨ r i ( s , T ) ⟩ T ⟨ r j ( s , t ) ⟩ T . ( 1 ) A measure of the strength of pairwise noise correlations across stimuli is the average correlation coefficient , defined as c i j N = 〈 C i j ( s ) C i i ( s ) C j j ( s ) 〉 s , ( 2 ) where 〈 . 〉s denotes the average over all stimuli presented . This quantity is to be contrasted with the signal correlation , c i j S = cov ( r i ( s ) , r j ( s ) ) s var ( r i ( s ) ) s var ( r j ( s ) ) s , ( 3 ) which measures the similarity of average responses across stimuli . To quantify the way in which the orientation of the high-dimensional distribution changes across stimuli , we use the variance of the population activity projected along the direction of the ( normalized ) mean response , r ¯ ( s ) = r ( s ) / | r ( s ) | , i . e . , σ μ 2 ( s ) = ∑ i j C i j ( s ) r ¯ j ( s ) r ¯ i ( s ) . ( 4 ) This quantity can be compared to the variance projected along the diagonal direction , d ¯ = ( 1 , … , 1 ) T / N , σ d 2 ( s ) = ∑ i j C i j ( s ) d ¯ i d ¯ j , ( 5 ) which corresponds to a uniform averaging of the covariances . To compare these quantities across stimuli , σ μ 2 and σ d 2 are normalized by the sum of the variances , σ a l l 2 ( s ) = ∑ i C i i ( s ) . If , for a given stimulus , all neurons are equally active on average , then r ¯ = d ¯ and σ d 2 = σ μ 2 . As we show below , different circuit models predict distinct stimulus dependencies of σ μ 2 and σ d 2 . The discrepancies are most apparent for the stimuli for which the average population response differs strongly from a uniform population response . A deviation from a uniform response can be measured by the angle between mean response and diagonal , or cos ( d , r ) = r ¯ T · d ¯ . For a graphical illustration of these measures , see Fig 1A and 1B . Correlations in the activity of neurons can have other origins besides recurrent connections . In parallel with the recurrent network model , we consider two alternative prototypical models in which correlations originate from shared inputs or from shared gain fluctuations , respectively . The three different scenarios are illustrated in Fig 1C . In S1 Appendix we show how , formally , these models can also be cast as special cases of the recurrent model . In order to evaluate the influence of correlation on neural coding , we examine the discriminability of a set of stimuli from the population activity . In the models , a stimulus , s , evokes activity characterized by an average response vector , r ( s ) , and a covariance matrix , C ( s ) . We seek a simple measure for evaluating the possibility of attributing a given population response to one of two discrete stimuli , s1 and s2 , unambiguously . For this , we assume that the high-dimensional distributions of responses are Gaussian , and project the two distributions on a single dimension in which we calculate the signal-to-noise ratio as our measure . The best projection follows from Fisher’s linear discriminant analysis: the most informative linear combination of neuron responses , denoted by w ⋅ r , is achieved if the vector w points in the most discriminant direction , w = ( C ( s 1 ) + C ( s 2 ) ) - 1 ( r ( s 1 ) - r ( s 2 ) ) . ( 21 ) The mean and variance of the projected distributions onto the normalized direction , w ¯ , are w ¯ T r ( s i ) and σ s i 2 = w ¯ T C ( s i ) w ¯ , for i ∈ {1 , 2} . We can then define the signal-to-noise ratio , as S = | w ¯ T · ( r ( s 1 ) - r ( s 2 ) ) | σ s 1 + σ s 2 . ( 22 ) Larger values of S correspond to better discriminability . To quantify the effect of correlation on discriminability , we compare the quantity S , defined in Eq ( 22 ) , with the quantity Sshuffled obtained from a ‘shuffled data set’ , in which responses are shuffled across trials ( in experimental data ) or off-diagonal covariance elements , Ci ≠ j ( s ) , are set to 0 ( in models ) . According to this procedure , we compare the accuracy of the code in the full ( experimental or model ) data with that in artificially generated data in which noise correlations are removed but which keeps the average populations responses and single-neuron variances unchanged . A ratio Sshuffled/S smaller than unity indicates that correlation is beneficial to discriminability . We note that S is an approximation for a measure based on the optimal linear classifier , for which the threshold separating the two one-dimensional projected distributions has to be calculated numerically . A measure akin to S used in similar contexts is the linear Fisher information [23] , valid for continuous stimuli . An advantageous property of S is its invariance under linear transformations: if all responses , r , are fed into another network whose output , Br , results from a product with an invertible matrix B , S does not change . This obtains because the mean responses are transformed into Br ( s ) , while the covariances are transformed into BC ( s ) BT . Intuitively , a simple matrix multiplication is accommodated for by a corresponding change in the most discriminant direction w on which we project the population activity .
One population feature which distinguishes the possible mechanisms of generation of noise correlation is the relation between the population-averaged response , 〈r〉 ( s ) = 〈ri ( s ) 〉i = ∑i ri ( s ) /N , on the one hand , and the population-averaged variance , 〈Cii〉i ( s ) = ∑i Cii ( s ) /N , or the noise covariance averaged across pairs , 〈Cij ( s ) 〉i ≠ j = ∑i ≠ j Cij/N ( N − 1 ) , on the other hand . In this section , we analyze these relations for Poisson neurons in a recurrent network with random effective connections , as described in Methods . Using Eq ( 12 ) in which a baseline spiking rate has been included , which reads Cij = ∑k Bik Bjk ( rk ( s ) + a + Vext , k ) for the pairwise covariances , we derive the expression ⟨ C i i ⟩ i ( s ) ≈ N ⟨ B 2 ⟩ ( ⟨ r ⟩ ( s ) + a + ⟨ V ext ⟩ ) ( 23 ) for the average variance , and the expression ⟨ C i j ⟩ i ≠ j ( s ) ≈ N ⟨ B ⟩ 2 ( ⟨ r ⟩ ( s ) + a + ⟨ V ext ⟩ ) ( 24 ) for the average covariance . ( See S1 Appendix for mathematical details and Fig 2 for numerical results . ) The quantity 〈B〉 = ∑ij Bij/N2 is the average strength of the effective connections in neuron pairs and , correspondingly , 〈 B 2 〉 = ∑ i j B i j 2 / N 2 is the average of its square . The variance of the input , averaged across neurons , is 〈Vext〉 , and a denotes a possible constant offset in the observed firing rates ( firing rate baseline ) . Eq ( 24 ) yields a linear relation between the average spiking rate and the covariances . The factor of proportionality—the ‘slope’—has been derived , here , with the assumption that the elements of B are independent . If this is not the case , the slope will also depend on the correlations among the matrix the elements , Bij . However , the linear dependence between rates and ( co- ) variances itself , which results from the nature of the Poisson spike generation , is independent of this assumption . In the linear relations between average variance and spiking rate , and between average covariance and spiking rate , the ratio of the intercept and the slope is identical and equal to the strength of the external noise . In summary , in the recurrent network model , both internally generated noise and external noise are amplified ( or attenuated ) by the recurrent connections , and the parameters in the relations between ( co- ) variances and population response can be related to network and input parameters . In feed-forward networks , both shared input units and global gain fluctuations can yield noise correlations [7 , 9 , 13] . Here , we compare the population signatures of the noise statistics in the case of these two alternatives with those that emerge in a recurrent network model . In a feed-forward network , pairwise covariances are given by Eq ( 17 ) , which can be rewritten as C i j = δ i j ( r i + a ) + ∑ k F i k F j k V ext , k . ( 25 ) Note that , in contrast to the case of a recurrent network , the neural firing rates , ri , only affect diagonal entries , i . e . , variances . The average covariance can be expressed as ⟨ C i j ⟩ i ≠ j ( s ) ≈ N ⟨ F ⟩ 2 ⟨ V ext ⟩ ( s ) , ( 26 ) and does not directly depend on the average population response , 〈r〉 ( s ) . The relation between the average input variance , 〈Vext〉 , and the average population response depends upon on the balance between positive and negative elements in rext , which we measure through the input variability , ρext = var ( rext ) /〈rext〉2 ( see S1 Appendix ) . In a model network with global gain fluctuations , the average covariances scale quadratically with the population response , as ⟨ C i j ⟩ i ≠ j = V ext ⟨ r ⟩ 2 . ( 27 ) This is a direct consequence of the fact that pairwise covariances are proportional to the product of the firing rates of the two neurons in the pair . By contrast , the single-cell variances and the population-averaged variance each have both quadratic and linear contribution in r and 〈r〉 , respectively ( see also e . g . [9] ) . ( Further details are provided in S1 Appendix . ) Additional discrepancies between the different scenarios can be related to the detailed structure of the matrix C , which determines the shape of the response distribution . By ‘shape’ we refer to the geometric orientation and extent of the multi-dimensional ellipsoid cloud that corresponds to the population responses , in the space spanned by the responses of the individual neurons . A two-dimensional sketch of such an ellipse , and the geometric interpretation of the quantities we use , are depicted in Fig 1A ( see also Methods ) . Loosely speaking , the orientation of the response distribution depends on the network architecture: the long axis of the ellipsoid , i . e . , the direction of the largest variance , lies along the diagonal in the cases of the recurrent network model and the feed-forward model with shared inputs , but instead along the direction of the mean response in the case of the gain-fluctuation model ( Fig 1B ) . ( The “diagonal direction” corresponds to a population response pattern in which all neurons are equally active . ) To capture this dependence quantitatively , we consider the projected variances , σ μ 2 and σ d 2 ( see Fig 2 , and Methods for details ) . Apart from differences in the relations between population-averaged response and variance , Fig 2A–2C , we find that , in the gain fluctuation model , but not in the other two models , the variance projected along the direction of the average response is approximately constant across stimuli ( Fig 2D–2F ) . In recurrent and feed-forward network models in which the connections induce strong correlations , a large proportion of the variance lies along the diagonal direction , independently of the stimulus , in contrast to the outcome of the gain-fluctuation model ( Fig 2G–2I ) . The variance projected along the direction of the average response is large only if the latter happens to be similar to the diagonal direction . For a given data set , a plot of the relations between population response and ( co ) variance can thus be used as a simple test for the consistency of the data with the different circuit models . In the simple scenarios we analyze here , average responses and covariances are related because they both depend on the network architecture . By expressing signal correlations , c i j S , and noise correlations , c i j N , in terms of network and input parameters , we can derive a relation between these two sets of quantities ( Fig 3 ) . For a random transfer matrix , B , all neuron pairs are statistically equivalent , and the strength of correlations can be characterized by the average of the pairwise signal and noise correlations . In a recurrent network model obtained from a random matrix B in which elements are independent , the network architecture affects noise and signal correlations through an effective quantity , namely , the signal-to-noise ratio of the elements of the transfer matrix , ρ = var ( B ) /〈B〉2: c N ≡ ⟨ c i j N ⟩ i ≠ j ≈ 1 1 + ρ , ( 28 ) c S ≡ ⟨ c i j S ⟩ i ≠ j ≈ 1 + ( N - 1 ) c in 1 + ρ + ( N - 1 ) c in ( 29 ) ( see S1 Appendix , Eqs ( 19 ) and ( 26 ) ) . Here , cin denotes the strength of signal correlations in the input across stimuli . Both signal and noise correlations are larger in networks with more homogeneous entries ( smaller ρ ) . If the average input for pairs of neurons across stimuli is uncorrelated ( no input signal correlations , cin = 0 ) , noise and signal correlations are identical on average . However , due to a prefactor which grows with system size ( N − 1 in Eq ( 29 ) ) , even weak input signal correlations are strongly amplified and can yield a large effect , Fig 3A . Because noise in the population response to a given stimulus is produced internally in the process of spike generation , noise covariances are unaffected by this mechanism , so that strong signal correlations can coexist with weak noise correlations , Fig 3B . These relations depend on network parameters and hold for the average correlations ( across pairs in a network ) . The pairwise noise and signal correlation coefficients , c i j N and c i j 2 , can vary widely , but numerical calculations indicate that they are correlated ( across pairs ) in a given network , Fig 3C . For the corresponding relations in the other two models , see Table 2 . Details of the derivations are reported in S1 Appendix , as well as in previous work on the gain-fluctuation model ( see Discussion ) . Since the network architecture affects both signal and noise correlations , it is natural to ask how the discriminability of stimuli is affected in turn , as it depends on both quantities . Indeed , noise correlations can affect the coding properties of a population appreciably [24] . Noise correlations are referred to as favorable for the discrimination of a pair of stimuli when the shapes of the two correlated response distributions ( corresponding to the two stimuli ) are such that variability occurs predominantly in a direction orthogonal to the informative direction ( Fig 4 ) . The relevance of correlations can be quantified by comparing the discriminability of stimulus pairs in the case of the full population output and in the case in which noise correlations are removed by shuffling trials . A number of earlier studies have considered the influence of correlated variability on coding [23 , 26 , 28 , 29 , 32] . Most of these , however , did not consider the dynamical or mechanistic origins of correlations . Some recent investigations have related coding with correlated variability to questions of dynamics and circuits [7 , 9 , 11 , 13 , 14] , but these have focused on feed-forward architecture and , primarily , small populations of neurons . By contrast , we consider here larger , recurrent networks . In Fig 4 , we analyze , in the recurrent network model , the interplay of signal and noise correlations that influences stimulus discrimination . A dominant portion of the variance occurs along the diagonal ( Fig 4A ) , so that stimulus discriminability depends on the angle between the diagonal and the line connecting the two average population responses ( corresponding to the two stimuli ) . In the case of a family of stimuli ( Fig 4B ) , discriminability depends on the distribution of the many angles corresponding to different choices of pairs of stimuli . Here , we illustrate this quantity ( Fig 4C ) , and we compare discriminability in the recurrent model to that for independent neurons ( Fig 4D ) , in the case of a population with 60 neurons . We find that the more heterogeneous our recurrent connections , i . e . , the larger the parameter ρ , the more frequently small cosines , i . e . , favorable angles , occur ( Fig 4C ) . We then compare discriminability ( through the signal-to-noise ratio , S , defined in Eq ( 22 ) in the recurrent network model with that in a population of independent neurons obtained by shuffling across trials ( Fig 4D ) . We find that the effect of shuffling depends on network heterogeneity as well ( via the distribution of the noise correlations ) ; however , when averaged over all pairs of stimuli , the effect of correlations is negligible and does not depend on the network’s heterogeneity . In the case of more homogeneous recurrent networks , both the beneficial and harmful effect of noise correlations are boosted , for different stimulus pairs . We note that similar effects can be observed in the model of feed-forward network with shared input and in the model of feed-forward network with gain fluctuation; see S1 Fig . In S2 Appendix we analyze idealized two- and four-population models of recurrent networks which provide further intuition on the relation between network structure and the representation of stimulus information . We illustrate the way in which the accuracy of stimulus representation depends on the synaptic strengths; among other things we show that , by adequately varying the within-population and cross-population synaptic strengths , stimulus representation can improve while firing rates and variance remain unchanged . Intriguingly , this improvement is accompanied by an increase of the noise entropy in the network , rather than the more commonly expected decrease of the noise entropy . We use the theoretical results presented in the previous sections to analyze the responses of populations of neurons to different stimuli . The data set contains the firing rates , collected during a certain time interval , in response to the presentation of different sound stimuli in a number of trials ( see Methods and Ref . [35] ) . We will compare the properties of covariances ( across trials ) and average responses with the predictions of our different models . Based on the models , we can then evaluate the effect of network-generated variability on stimulus discrimination . The trial-to-trial variability in single-neuron output is large ( supra-Poisson , Fig 5A ) ; neurons with larger average response exhibit also a larger variability in their responses . This tendency is observed across different neurons responding to a given stimulus , for individual neurons across stimuli , and for the average population response across stimuli . The pairwise correlations in trial-to-trial variability are quantitatively large and pairs with high noise correlations tend to have strong signal correlations , Fig 5B . In other words , because signal correlations measure the similarity of average responses across different stimuli , neurons with similar tuning properties also have similar trial-to-trial variability . Additionally , in populations with strong average signal correlations , noise correlations are strong as well . These relations between noise and signal correlations can be reproduced in the recurrent network model ( Fig 3 ) . However , similar results may be possible also in alternative scenarios for the generation of correlated variability . We compare the different models in the next section . Motivated by our investigation of network models , we first examine the relation between the shape of the response distributions for each stimulus and the pattern of average responses for the entire set of stimuli . In Fig 6A and 6B , we display the normalized standard deviation projected on the direction of the mean response , σμ/σall , and the normalized standard deviation projected on the diagonal direction , σd/σall , as functions of the angle between average response and diagonal , represented as cos ( d , r ) ( see Methods for the definitions of these quantities ) . We observe a strong dependence on cos ( d , r ) of the variance projected on the direction of the mean response , but not of the variance projected on the diagonal direction . The dependence of σμ/σall on the stimulus is much stronger in nearly all the measured populations , Fig 6C . This behavior is consistent with a network model , either feed-forward or recurrent , but not with a model with shared gain fluctuations , where a large part of the variance is consistently in the direction of the mean response ( see Sec . “Population signatures of noise statistics in feed-forward network models” ) . We note , however , that in the data firing rates are measured relative to the spontaneous activity; this is reflected in the presence of negative values in the stimulus-evoked activity . Such an offset could affect our comparison because we do not know the true value of the mean response . To account for this possibility , we searched for the best possible offset , assuming the model of gain fluctuations for the stimulus-dependent covariances . We then corrected the firing rates by this offset , and evaluated the stimulus dependence of σμ and σd , as before ( see S1 Appendix for details ) , but found no qualitative change in the results . However , we did not test for mixed models , and it is possible that part of the correlated variability can be explained by common gain fluctuations . Based on the results presented above , we conclude that , of our three scenarios , feed-forward and recurrent network models are more consistent with the data . In the following , we fit the parameters of these models to the data , to find which of these two provides a better fit . For this comparison , we analyze the dependence of the average variances , 〈Cij〉i ≠ j , and covariances , 〈Cii〉i , on the population-averaged response , 〈r〉 ( s ) . In the recurrent network ( Eqs ( 23 ) and ( 24 ) ) , both variances and covariances increase linearly with 〈r〉 , while in the feed-forward network ( Eq ( 27 ) ) , mean covariances are not expected to depend strongly on 〈r〉 . To examine whether the experimental data is consistent with the feed-forward model , we fitted parameters to the activity of each experimentally observed population ( see S1 Appendix ) , and generated a surrogate set of responses and covariance matrices , with a matching number of neurons and stimuli . We followed this procedure because , here and in the following , we do not fit a full connectivity matrix , but generate a matrix , F , with random entries from a distribution with parameters to fit the data at the population level . In particular , we obtained values for the variability in the input ensemble , ρext = var ( rext ) /〈rext〉2 , and in the network connection strengths , ρ = var ( F ) /〈F〉2 . We found that ρext , ρ > 1: for both input and network elements , the variance was ( appreciably ) larger than the mean ( see Fig 7A and 7B ) . A large value of ρext is needed to explain the high variability of average responses across stimuli , while the variability of the network connections captured by the parameter ρ is related to the magnitude of the observed correlations . With these parameters , the statistics of the distribution of average responses are well reproduced ( Fig 7C ) . As mentioned in Sec . “Population signatures of noise statistics in feed-forward network models” , for large values of ρext the feed-forward model does not predict a strong dependence of 〈Cij〉i ≠ j on 〈r〉 ( s ) . We quantify this relation by the ratio of slope to intercept in a linear fit of the data; this ratio measures the strength of dependence of average covariances on population responses . In the feed-forward scenario , we find that the slopes of the linear fits in the model relative to the intercepts are too low in comparison to what is observed in the experiments ( Fig 7D and 7E ) . The feed-forward model cannot reproduce both the large variability of neural responses across stimuli and the increase of the average covariance with the average response . In a recurrent model , this increase is expected , because noise generated by the neurons is propagated through the network: the covariances are proportional to the average response ( see Eq ( 24 ) ) . Moreover , it predicts that the ratios of intercept and slope in the linear behaviors of variances and covariances are the same ( this ratio scales with the variance of the external input ) . Indeed , linear fits to these relationships reveal approximately consistent ratios of intercept and slope ( panel F ) ; the estimated parameters are summarized in Fig 7G and 7H . The external noise , estimated from the ratio intercept/slope , turns out to be of the same magnitude as the average rate . Interpreted in terms of the model , the noise resulting from Poisson spike generation thus contributes as much to neuron variance as the external input . This combined variability is propagated through the network , and , on average , multiplied by a factor N〈B〉2 > 1 , corresponding to the slope in Eq ( 24 ) . If this factor is larger than one , both average response and noise are amplified by the recurrent connections . In summary , we find that a model of recurrently connected neurons with random effective connections captures the observed activity statistics , in particular the relations between average response and average covariances as well as the consistent shape of population fluctuations across stimuli . Feed-forward networks models with either shared input or common gain fluctuation are not consistent with all of these observations . We note , however , that our conclusions are based on a relatively small data set; a larger number of trials and measurements of absolute values of firing rates would be desirable for a more stringent test . We quantify the influence of noise correlation on stimulus discrimination by the ratio Sshuffled/Soriginal , where the signal-to-noise ratio , S , is calculated for the data before and after shuffling trials . The quantity S is defined in Eq ( 22 ) and denotes , for a pair of stimuli , the difference in average response divided by the standard deviation of the responses , both projected on the most discriminant direction . Larger values indicate that stimuli are easier to discriminate; if Sshuffled/Soriginal is larger than one , then removing correlations by shuffling trials improves stimulus discrimination . Across pairs of stimuli , this signal-to-noise ratio varies strongly in the data ( Fig 8A ) . On average , stimuli are slightly easier to discriminate in the shuffled data , i . e . , noise correlations are weakly harmful to the coding of the stimulus set used in the experiments Fig 8B . A closer analysis of the response distributions reveals that , to a large degree , the effect of shuffling can be explained by the relative locations of the two mean responses with respect to the diagonal , Fig 8C and 8D , as measured by cos ( d , r ( s1 ) − r ( s2 ) ) ( Fig 4 ) . This is because , due to the noise correlations , the main direction of variability is along the diagonal . The overall effect of shuffling on stimulus discrimination depends on the relation between noise and signal correlations , Fig 8E and 8F: stronger signal correlations lead to a stronger dominance of angles that are unfavorable if noise is aligned along the diagonal direction , and in this case shuffling correlations away will benefit stimulus discrimination . In the above , we investigated the influence of noise correlation on stimulus coding . A more biologically relevant question pertains , rather , to the influence of circuit architecture and parameters on coding . These govern the response statistics , which in turn influence the coding performance . Upon varying circuit properties , in general both average responses and their higher-order statistics can change [18 , 30] . The overall effect of the network stimulus information during this transformation of the input into the spiking output depends on the noise generated intrinsically—by the Poisson spike-generation mechanism . Noise in the input-output transformation necessarily destroys some of the information contained in the input , so that this transformation is lossy . Stronger excitatory connections generically amplify both average responses and the trial-to-trial variability . Specifically , the variance of Poisson neurons scales with firing rate . In a feed-forward network , this is the only source of noise , and the input to the neurons increases with the strength of the feed-forward connections . As a result , the signal-to-noise ratio improves with stronger excitatory connections: the enhancement in the firing rates overcompensate the increase in noise . By contrast , in a recurrent network , noise from Poisson spike-generation is fed back into the network and this causes fluctuations of the firing rates themselves . These fluctuations are also amplified by the recurrent connections , and consequently the signal-to-noise ratio decays with stronger excitatory connections . Formally , the distinction between feed-forward and recurrent networks is embodied in dependence of the ( co ) variances on network parameters and average responses ( Eqs 17 and 12 ) . To examine the combined effects of recurrent connections on average responses as well as variability around these and its correlation , it is instructive to consider a simplified , recurrent network model; we discuss such a model in detail in S2 Appendix . While the theoretical framework is not a new one , the advantage of this simple model is that the effect of recurrent connections can be calculated explicitly . We show that , provided the recurrent connections are sufficiently strong , shuffling correlations will reduce stimulus discriminability . This effect arises because the shuffling removes the effect connections have on correlations , but not their effect on the average responses and single cell variances .
We studied correlated variability theoretically using the idealized framework of Poisson neurons and random effective networks . These simplifications allow us to focus on qualitative differences arising from different network architectures; in turn , the identified qualitative trends can be tested against experimental data without having to infer the detailed connectivity structure . On the down side , we cannot fit individual pairwise correlations . Also missing is a thorough investigation of the effects of more biologically realistic , non-random connections . In particular , specific connectivity structures which may give rise to different response regimes [40] , such as weakly correlated activity , remain to be investigated . A broad range of quantitative work has been devoted to examining the origins of variability in the activity of neural populations . Recurrence has been invoked to explain population variability in the absence of microscopic stochasticity [41 , 42] , and more recent work along has extended this line of thought to cover a greater range of response regimes of asynchronous [43–45] . Population-wide fluctuations have been identified as contributing strongly to the observed correlations between the responses of cortical neurons to sensory stimuli [10] . Global , multiplicative gain fluctuation has been proposed also as a mechanism for the generation of such correlations [11] , and its implication on the form of the correlation has been analyzed and fitted successfully to cortical data [9] . Similar models capture noise correlations , and their stimulus dependence , in the retina [13 , 14] , as well as the large correlations observed under anesthesia [12] . Finally , global gain fluctuation models have also been used to reproduce the relations between noise and signal correlations [7] . In general , it is difficult to differentiate the mechanism of shared input from that of common gain fluctuation , and even from that of recurrence , as the origin of correlated variability , based on measured activity . Here , instead of focusing on temporal dynamics [16 , 46 , 47] , we approach the problem by considering the relation between the network architecture and the structure of population activity statistics , when a neural population is presented with an ensemble of stimuli . In principle , the observation of population responses to a battery of stimuli provides sufficiently many constraints to tease apart different network models , and infer the parameters in the corresponding connectivity matrices . In practice , and especially for large neural populations , the number of trials accessible in experiments constrains the precision of measured covariances and the resolution of the fits . Here , we took a macroscopic view , by considering measurable effects at the population level , rather than aiming to estimate the weights of individual ( effective ) connections . As such , we compare population quantities derived from our models to population quantities extracted from the data , rather than to individual , pairwise correlations or other such microscopic quantities . Our theoretical analyses approximate the spike generation mechanism in neurons by a Poisson process; as such , single-cell variability scales with the spiking rate . If the noise generated within the network were purely additive , and , hence , independent of spiking rates , it would be indistinguishable from external noise that is simply filtered through the network [38] . In the Poisson model , by contrast , the interplay of internally generated noise and variability due to external signals imprints its signature on the dependence of covariances on firing rates . Even if Poisson variability happens to be a faithful model of single-cell variability , still our models make the limiting assumption of linearly interacting Poisson processes . Clearly , nonlinear transfer functions can affect the relation between spiking rates and noise correlations in individual pairs of neurons [8 , 48] . The tractability of the model , however , allows us to obtain explicit relations between observable statistics , which can be used as a starting point to interpret the origin of correlated variability . We used a recurrent network with random effective connections to estimate the influence of the variability generated in the network dynamics on stimulus representation . Based on the paradigm that response variability amounts to noise that downstream neurons have to cope with , a number of authors have argued that noise correlations can benefit information coding , provided they suppress variability along the direction relevant for stimulus discrimination [24–26 , 29] . The effects of additive and multiplicative origins of noise correlation have been examined in Refs . [11 , 13 , 14] . As in earlier work , we quantified the effect of noise correlation on coding by comparing the true or model data with their modified versions obtained from shuffling trials: this allows for a comparison between the correlated population and a mean- and variance-matched independent population . In our work , however , we compared the outcome of this procedure in the case of three different network architectures , and in the context of large populations of ( model ) neurons presented with a battery of stimuli . Recent theoretical studies have examined the impact of noise and correlations generated in recurrent networks . Noise internally generated in a network via spike generation can in principle be averaged out in feed-forward or recurrent networks , depending on the connectivity of the network , by increasing the population size and the redundancy of the population code [30] . However , external noise and ‘sub-optimal processing’ [33] can limit the amount of information in the output activity , a fact that is reflected by the specific structure of the correlations ( rather than their magnitudes ) [32] . However , such structures are difficult to pinpoint in detail in measured activity [32] . In experiments , it thus remains difficult to evaluate the influence of network-generated noise on stimulus coding . In the present study , our goal was to take experimentally observed statistics of the population activity as a starting point , and to interpret the structure of the average responses and noise correlations based on outcomes of model networks . Varying network properties such as the architecture of connections changed both the pattern of average activity and the noise correlations , both of which influence the accuracy of the neural code . As our mathematical results indicate , the effect of recurrent dynamics on information coding depends primarily on the magnitude of the noise generated internally . Our toy model , as well as the investigation presented in Refs . [18 , 30] , suggests that , generically , recurrent amplification does not improve information coding . The reason is essentially that , in a recurrent network , the noise inherent to the spike generation mechanism is fed back in the network as ‘input noise’ , which in turn amplifies the ‘output noise’ . This amplification harms the signal-to-noise ratio . By contrast , in a feed-forward network , only the ‘output noise’ in spike affects the code; in Poisson neurons , an increase in this variability is overcompensated by an increase in the firing rate , so that the signal-to-noise ratio improves . Our recurrent network model allows us to estimate the ‘amplification factor’ in an experimental population , as well as the consequent strength of the noise , provided simplifying assumptions such as that of a random effective network architecture . We tested the consistency of different scenarios involving interacting Poisson neurons with measured neural population responses , and we interpreted the observed variability in terms of a recurrent network model . The measured population activity was given as spike counts in 250 ms bins , so that correlations were defined over a relatively slow timescale and , therefore , did not take rapid temporal modulation ( as noted in , e . g . , Ref . [49] ) into account . Both experimental variability and correlations were high [50] , so that the effects of correlations were potentially strong , and amenable to an analysis such as ours . In general , variability and correlations can be high in anesthetized animals [11] , but noise and signal correlations were noted to be somewhat weaker in experiments that are comparable to the one we consider [51] . While it is difficult to rule out that part of the correlations result from experimental artifacts relating to the calcium imaging technique used in the experiments ( e . g . , scattering of fluorescent light by the neuropil ) , intracellular recordings of rates were consistent with calcium recordings [35] . A related point concerns the identification of the deconvolved fluorescence signal with the neural response: if the latter is inferred up to a multiplicative factor , the numerical values of the inferred covariances will be affected by the square of this factor . Such a factor , however , would not change the functional form of the relations between firing rates and covariances , but only the numerical values of inferred parameters . Furthermore , calcium imaging operates on a slow time scale as compared to the dynamics of individual spikes . In this work , we were interested in the coarse population activity and variability , and therefore considered spike counts in a fixed window , thereby disregarding the precise temporal dynamics of neural activity . The latter may indeed be relevant , especially in auditory processing [52 , 53] . In comparing the data to our model results , we did not to attempt at inferring the strength of individual ( effective ) connections between neurons; this would require a much larger volume of data . Instead , we focused on the parameters of the distribution of effective , random connections . This approach , in the context of a recurrent network model , allowed us to capture not only the scaling of variances and covariances with firing rates , but also the relation between signal correlation and noise correlation , and the effect of shuffling away covariances on stimulus discrimination . Furthermore , differences among experimental populations of neurons could be attributed to changes in the effective connectivity within populations . Recent studies [7 , 9 , 11–14] have proposed the mechanism of common gain fluctuation to explain correlated variability across a population . In this work , we identify similarities and differences of correlated variability that arises in recurrent models vs . feed-forward models such as the common-gain-fluctuation model . Estimated model parameters indicate recurrent amplification and suggests that the portion of the variability that can be traced back to spike generation in the population—noise harmful to coding—is comparable in magnitude to the external input noise . Interpreting correlations in the framework of a random network is obviously not a new idea; our results illustrate the fact that subtle differences in the output statistics can lead to very different interpretations on network architecture . Our data analysis suffers from an unknown offset due to putative baseline spiking rates , the limited volume of the data , and the low temporal resolution . In this context , we compared different prototypical scenarios , but did not consider mixed models . In experimental populations of neurons , the statistics of activity can obviously result from a combination of feed-forward and recurrent processes . An likely scenario is one in which a feed-forward network provides correlated input to a recurrent network from which recordings are carried out . For simple choices of model stimulus sets , pure network models will not explain all aspects of the data . Examining the relative contributions of recurrent and feed-forward processes may be relevant to understanding the origin and role of correlated variability . As a first step , here , we provide some characterizations of the outcome of each process taken individually .
|
The response of neurons to a stimulus is variable across trials . A natural solution for reliable coding in the face of noise is the averaging across a neural population . The nature of this averaging depends on the structure of noise correlations in the neural population . In turn , the correlation structure depends on the way noise and correlations are generated in neural circuits . It is in general difficult to identify the origin of correlations from the observed population activity alone . In this article , we explore different theoretical scenarios of the way in which correlations can be generated , and we relate these to the architecture of feed-forward and recurrent neural circuits . Analyzing population recordings of the activity in mouse auditory cortex in response to sound stimuli , we find that population statistics are consistent with those generated in a recurrent network model . Using this model , we can then quantify the effects of network properties on average population responses , noise correlations , and the representation of sensory information .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"methods",
"Results",
"Discussion"
] |
[
"medicine",
"and",
"health",
"sciences",
"neural",
"networks",
"engineering",
"and",
"technology",
"nervous",
"system",
"signal",
"processing",
"signaling",
"networks",
"random",
"variables",
"neuroscience",
"covariance",
"mathematics",
"network",
"analysis",
"computational",
"neuroscience",
"coding",
"mechanisms",
"computer",
"and",
"information",
"sciences",
"animal",
"cells",
"neural",
"pathways",
"probability",
"theory",
"cellular",
"neuroscience",
"neuroanatomy",
"cell",
"biology",
"anatomy",
"signal",
"to",
"noise",
"ratio",
"neurons",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"physical",
"sciences",
"computational",
"biology"
] |
2018
|
Interpretation of correlated neural variability from models of feed-forward and recurrent circuits
|
The equine-associated obligate pathogen Burkholderia mallei was developed by reductive evolution involving a substantial portion of the genome from Burkholderia pseudomallei , a free-living opportunistic pathogen . With its short history of divergence ( ∼3 . 5 myr ) , B . mallei provides an excellent resource to study the early steps in bacterial genome reductive evolution in the host . By examining 20 genomes of B . mallei and B . pseudomallei , we found that stepwise massive expansion of IS ( insertion sequence ) elements ISBma1 , ISBma2 , and IS407A occurred during the evolution of B . mallei . Each element proliferated through the sites where its target selection preference was met . Then , ISBma1 and ISBma2 contributed to the further spread of IS407A by providing secondary insertion sites . This spread increased genomic deletions and rearrangements , which were predominantly mediated by IS407A . There were also nucleotide-level disruptions in a large number of genes . However , no significant signs of erosion were yet noted in these genes . Intriguingly , all these genomic modifications did not seriously alter the gene expression patterns inherited from B . pseudomallei . This efficient and elaborate genomic transition was enabled largely through the formation of the highly flexible IS-blended genome and the guidance by selective forces in the host . The detailed IS intervention , unveiled for the first time in this study , may represent the key component of a general mechanism for early bacterial evolution in the host .
The genomes of host-adapted bacteria , including endosymbionts and obligatory intracellular pathogens , go through reductive evolution [1] , [2] , [3] . Such changes are partly due to a reduced pressure to maintain genes that are not essential for survival in the host . Similarly , decreased efficiency of purifying selection , resulting from the reduced population size from a restricted life , results in inactivated genes , including beneficial genes , through genetic drift [3] . During the early stage of the genome reduction process , the majority of genes are lost as large chromosomal fragments spanning multiple genes . Such genome reduction has been documented in diverse bacterial groups , including Firmicutes , Chlamydiae , Spirochetes , and γ-Proteobacteria [1] , [3] , [4] , [5] , [6] , [7] . Most of these bacteria have large expansion of IS elements ( insertion sequences ) , and thus it has been suggested that the IS elements may play an essential role during the genome reduction process [1] , [3] , [8] , [9] , [10] . Burkholderia pseudomallei and Burkholderia mallei belong to the ß-Proteobacteria family , and are the causative agents of melioidosis and glanders , respectively [11] , [12] , [13] , [14] , [15] , [16] , [17] . B . mallei has very recently ( ∼3 . 5 myr ) evolved from a clone of B . pseudomallei through extensive genome reduction [18] , [19] , accounting for as much as 1 . 41 Mb or 20% of the genome , as estimated by the size difference between the genomes of B . mallei ATCC 23344 and B . pseudomallei K96243 [18] , [20] , [21] . Concomitant with this process , B . mallei became constantly associated with mammalian hosts , specifically equines [22] , [23] , while B . pseudomallei maintains an opportunistic pathogenic lifestyle [17] . Preliminary analyses of the two type strains , B . mallei ATCC 23344 and B . pseudomallei K96243 , have suggested that genome reduction and rearrangement in B . mallei were mediated by IS elements that are widely spread throughout the genome [20] , [21] . Genes that have been deleted from the B . mallei genome but are maintained in B . pseudomallei include genes that are required for environmental survival . Many of these genes encode metabolic functions for the synthesis of metabolites or the utilization of various sugars and amino acids , without which bacterial propagation in the environment could be significantly hindered [20] . While the genomic reduction during bacterial restriction to their hosts has been well documented [1] , [8] , [10] , most of the stepwise processes have not yet been elucidated . The B . mallei genome has unique significance , as it is much younger than the other genomes in which the genome-reductive evolutionary processes have been most studied to date , including Buchnera ( >150 million years ) and other much older groups [1] , [3] , [4] , [5] , [6] , [7] . The studies with these older genomes have been challenging due to the subsequent genomic- and nucleotide-level mutations that accumulated over a long evolutionary history . In this study , we dissected 10 genomes each of B . pseudomallei and B . mallei to understand the early-stage processes that drive genome-reductive evolution in host-associated bacteria .
It is well known that bacteria specialized to a ( host ) niche , often have a large number of IS elements compared to their free-living relatives [1] , [3] , [8] , [9] , [10] . Likewise , by comparing genome sequences , we found that three types of IS elements , ISBma1 , ISBma2 , and IS407A , were significantly increased in B . mallei compared to B . pseudomallei ( Fig . 1A ) . By contrast , other types of IS , including IS1356 , ISBma3 , ISBma4 , and ISBma5 were found in low copy number in both species of bacteria . These elements appeared to be mostly degenerate evolutionary remnants ( i . e . , part of the IS disrupted or deleted ) of the Burkholderia lineage . ISBma1 , ISBma2 , and IS407A also had degenerate elements in each species; the ISBma1 elements had the highest levels of degeneration ( 44% ) , followed by ISBma2 ( 20% ) , and by IS407A ( 5% ) ( Fig . 1A ) . Intriguingly , up to almost 90% of ISBma1 , ISBma2 , and IS407A ( 88 . 5% , 86 . 1% , and 89 . 6% , respectively ) were found to be present at the corresponding loci in all 10 B . mallei strains , when examined after the rearranged genomic fragments in each strain were aligned against a reference genome of B . pseudomallei K96243 ( Fig . 1B; for a scaled map with the IS insertion sites in all B . mallei and B . pseudomallei strains , see Fig . S1; for the patterns of genomic rearrangements in the strains of each species , see Fig . S2; for the actual comparative blast data , see Tables S1 and S2 ) . In contrast to these “core” elements , those elements that were not present in all ( singletons and those found in a few strains ) , collectively referred to as “accessory” elements , were much less common . That the core elements , expected to be associated with the speciation of B . mallei from B . pseudomallei , accounted for most of the elements clearly reflects the common origin of B . mallei strains from a clone of B . pseudomallei [18] , [20] . More importantly , it also suggests that further transpositions were significantly slowed after subsequent geographical segregation of the bacteria . There are 13 core elements in B . mallei that have matching IS elements located at the same sites in B . pseudomallei strains ( Table 1 ) . These elements were found to be composed of elements of ISBma1 and ISBma2 but not of IS407A . This finding suggests that ISBma1 and ISBma2 have a longer history of association with B . pseudomallei than IS407A does . Among the three largely expanded elements , we found that IS407A and ISBma2 were associated with almost all of the large genomic deletions and rearrangements in the B . mallei strains ( Fig . 2; Figs . S1 and S2; Tables S1 and S2 ) . The only exception to this was a large deletion found in the strain ATCC 23344 and its direct derivatives , FMH , JHU , and GB8 horse 4 [24] , between the 43rd and the 44th elements in chromosome 2 ( Fig . 2; Table S1 ) . No genomic rearrangement was mediated by features other than the two IS elements . ISBma1 , which was significantly increased in B . mallei , was not directly involved in any of the genomic deletions or rearrangements , however as many as 35% of it served as secondary entry points for IS407A . The majority of the core elements of IS407A , 71 . 8% and 63 . 3% in chromosomes 1 and 2 , respectively , mediated rearrangements , deletions , or both ( Fig . 3A ) . By contrast , accessory elements of IS407A contributed less , but were more active in chromosome 2 than in chromosome 1 . By contrast , 50 . 4% and 53 . 2% of the core elements of ISBma2 in chromosomes 1 and 2 , respectively , contributed to rearrangements and/or deletions , and the accessory elements in both chromosomes were very rarely involved ( Figs . 2 and 3 ) . We identified 59 and 28 genomic fragments in chromosomes 1 and 2 , respectively , which were encompassed by core elements of IS407A or ISBma2; these core elements mediated genomic rearrangements in at least one strain ( Figs . 2; Table S1 ) . We referred to these genomic fragments as BRUs ( basic rearrangement units ) , a set of basic units for genomic reduction and rearrangement in B . mallei . The BRUs formed various rearrangement patterns in the B . mallei strains ( Fig . S2A ) . By contrast , B . pseudomallei strains had little variation in genome arrangement among one another due to low levels of IS elements- a few rearrangements were found but were around non-IS repeat sequences ( Fig . S2B ) . When the pattern of the IS insertions and their involvement in genome-reductive and rearrangement processes in strains were used to construct a phylogenetic tree , strains sharing a recent common ancestry ( e . g . , ATCC 23344 and its immediate derivative isolates , FMH , JHU , and GB8 horse 4 ) or common recent geographical origins ( e . g . , strains NCTC 10257 , NCTC 10229 , and 2002721280 from European countries ) were grouped together ( Fig . 3B ) . This phylogenic relationship supports the hypothesis that the accessory IS elements , which provided the major determinants for the tree rather than the common core elements , occurred following the speciation and geographical segregation of the B . mallei strains . By contrast , such patterns were not obvious among the B . pseudomallei strains which did not go through IS element expansions; Australian strains 1655 and 668 did not branch separately from the South Asian strains . The deletions and rearrangements that were mediated by accessory elements were most frequently noted in strains SAVP1 and 2002721280 , which lost virulence after successive passages in laboratory cultures [25] ( Figs . 2 and S1 ) . Most of the extra deletions in these strains were more prominent in chromosome 2 than in chromosome 1 . In SAVP1 , an IS407A-mediated deletion removed a major group of virulence genes encoding the animal-type type III secretion system in the BRU B22 ( Figs . 2 and S1 ) ; this deletion may be a major cause of the avirulence of that strain . By contrast , there is no obvious deletion that may be responsible for the loss of virulence in strain 2002721280 . That the strains SAVP1 and 2002721280 obtained deleterious mutations from in vitro culturing suggests that maintenance of the genomic contents in B . mallei requires selective pressure for survival in the host environment . By contrast , the fully virulent strain PRL-20 showed more frequent deletions and rearrangements mediated by accessory elements than other virulent strains . This strain may represent one of the more evolved ( more genome-reduced ) strains of B . mallei . Although extra deletions and rearrangements were noted , the actual number of the accessory IS elements was not significantly increased in PRL-20 , SAVP1 or 2002721280 . Furthermore , none of the direct derivatives of the strain ATCC 23344 ( i . e . FMH , JHU , and GB8 horse 4 ) had new IS insertions ( Fig . 2 and Table S1 ) . These ATCC 23344 derivatives also did not have genomic rearrangements; the only change found was a single IS407A-mediated deletion located within the BRU B17 in the strain JHU ( Fig . 2 and Table S1 ) . These lines of evidence suggest that B . mallei genomes are structurally flexible with regard to deletions , however perhaps not as much anymore for additional IS transpositions or genomic rearrangements . IS407A elements are known to generate 4-bp target region duplications as direct repeats around them when they transpose [26] . We found that ISBma1 generates 8-bp target region duplications , and that ISBma2 generates longer repeats of various lengths ( 18–26 bp ) ( Table 2; for the entire data , see Table S3 ) . In addition to the various lengths of duplications , these target regions of the three types of IS had different nucleotide compositions and patterns . Most notably , the sequences of ISBma1 contained homopolymers of A and/or T in up to 8-bp stretches of nucleotides ( Fig . 4A ) . The target sequences of ISBma2 had a loose pattern in which the GC-rich central region was encompassed by strands of As and Ts on either side . Target sequences of IS407A had the least characteristic composition . It is intriguing to note that each IS element showed different levels of copy number expansion , ISBma1 with the lowest ( 3 . 3× ) , ISBma2 with an increased level ( 9 . 5× ) , and IS407A with the highest ( 16 . 7× ) ( Fig . 1A ) . Perhaps this difference , at least in part , resulted from the availability of genomic sites suited for insertion targets . There were concordant patterns of disruption of the core elements of one type by another , in that ISBma1 and ISBma2 were intersected by transposed IS407A ( Fig . 4B ) , while the reverse ( IS407A disrupted by ISBma1 or ISBma2 ) was not found . A possible explanation for these insertion patterns may be that ISBma1 and ISBma2 could not transpose into IS407A due to the lack of sites suited for their rather uncommon target preferences , while IS407A did not have this problem . Consistent with this hypothesis , ISBma1 and ISBma2 also did not have self-disrupted elements , while there were several self-disrupted IS407A elements . The involvement of the three IS elements with different target sites increased the total number of IS insertions in the genome . Furthermore , this increase led to further spread of IS407A , because ISBma1 and ISBma2 provided neutral insertion points for the element . This in turn directly improved the efficiency of IS407A-mediated recombination in the genome , resulting in more sophisticated deletions and rearrangements . We estimated that 83 . 7% of IS407A and 65 . 6% of ISBma2 elements in the B . mallei genomes lost their matching target duplicates , while all of the elements from intact ISBma1 elements were maintained ( Table S3 ) . Almost all of the IS407A ( see Table S3 for details ) and all of the ISBma2 elements that contained matching repeats were not involved in genomic rearrangements in B . mallei . This indicates that recombination among the elements were the major cause of the loss of the matching target duplicates . B . mallei still has a high nucleotide-level identity ( 99% ) to B . pseudomallei . Consistent with this , there was no AT-biased genome deviation in B . mallei , unlike that seen in many old symbionts or obligatory host-associated pathogens [1] , [3] . Although the overall identity is still very high , significant nucleotide-level divergence exists , especially at the SSRs ( simple sequence repeats ) , where there are intrinsically high mutation rates [27] . These SSRs were abundant in both B . mallei and B . pseudomallei at corresponding sites in the genomes . However , there were more genes that were disrupted by frameshift mutations in B . mallei compared to B . pseudomallei ( Table S4 ) . Most of these disrupted genes were commonly present in all B . mallei strains , reflecting the clonal origin of the strains . Some of these gene disruptions may have contributed to better adaptation of the bacteria ( increased persistence ) in the host environment or simply became obsolete [28] . One of the most characteristic loss of function or of surface structure in B . mallei is the loss of flagella [20] . A gene essential for flagellum biogenesis , fliP , [29] in the strain ATCC23344 was disrupted by a 65-kb fragment flanked by IS407A elements , and this mutation completely turned B . mallei flagella-less . This disruption in fliP is present in all B . mallei strains ( Table S1; Fig . 2 , between BRUs A2 and A3 ) , implying the significance of losing flagella in the evolution of host-restricted B . mallei . The loss of flagella has been noted in other bacteria , including Bordetella pertussis and Bordetella parapertussis during their host specialization , derived from the strains of Bordetella bronchiseptica , [9] and Yersinia pestis during its conversion from a gut to a systemic pathogen [30] . Additional disrupted genes not present in all strains were found at approximately the same levels as in B . pseudomallei , suggesting that there were no significant increases in mutation rates in B . mallei after geographical segregation . There also was no significant level of erosion of these , so called , pseudogenes by purifying selection at levels high enough to contribute to the actual genome size reduction ( data not shown ) . The extensiveness of the genome-wide reduction and rearrangements as well as additional nucleotide-level mutations may suggest that there is a potential for altered gene expression patterns in B . mallei . A total of 341 potential regulatory genes survived the general IS-mediated genomic reduction in B . mallei ( not taking into account the diverse strain-specific deletions that occurred after speciation ) . Among these genes , only a small fraction ( about 10 ) in each strain had deleterious ( e . g . frameshift , null , or IS-insertion ) mutations ( for the list of the genes , see Table S5 ) . In addition , none of the predicted operons in B . mallei , which correspond to the putative operons previously found in B . pseudomallei K96243 [31] , were disrupted by IS elements ( data not shown ) . We also estimated the potential for changes in promoters . There were 2 , 473 upstream sequences of genes , many of which may overlap or contain promoters , in the reference genome of B . pseudomallei K96243 that have homologous sequences ( with at least 95% identity over at least 95% of their lengths ) in all other strains of B . pseudomallei . We found that up to 99% of these sequences also matched the corresponding regions in B . mallei ATTC23344 at the same homology levels ( see Table S6 for the list of the 2 , 473 upstream sequences , associated gene information , and the blast data ) . Together , all these data from the analyses of the conserved genomic regions suggest that there is only a low potential for the genes in B . mallei to have significantly divergent gene expression patterns from B . pseudomallei . By contrast , there were 56 genes with putative regulatory functions that were lost along with the commonly deleted genomic fragments of the B . mallei genome . These genes include potential global regulatory genes , such as those encoding a quorum-sensing system ( genes BPSS1176 and BPSS1180 in the reference genome of B . pseudomallei K96243 ) , a two-component regulatory system ( the pair BPSS1994 and BpSS1995 in B . pseudomallei K96243 ) , and a number of regulators of various families ( Table S7 ) . Whether the loss of any of these 56 regulatory genes affects the expression of the remaining genes in the B . mallei genome was yet to be examined . To experimentally estimate the possible transcriptomic divergence between B . pseudomallei and B . mallei , we infected female BALB/c mice with B . mallei ATCC 23344 or B . pseudomallei K96243 , employing the previously established aerosol models of acute glanders and melioidosis [32] . Gene expression was compared in the bacteria that colonized the lungs and the spleens of the mice . Both B . mallei- and B . pseudomallei-challenged animals showed increases in the bacterial loads within these organs over time , with B . pseudomallei having slightly faster growth rates ( Fig . 5 ) . In our experience , B . pseudomallei also grew faster than B . mallei in vitro ( data not shown ) . Unlike the mice infected by B . mallei , sampling the B . pseudomallei-challenged animals after 72 hr was not possible due to animal mortality from the more rapid disease progression . When gene expression profiles in the spleens and lungs were compared between B . mallei and B . pseudomallei at middle- ( i . e . , 24 hr for both bacteria ) and late stages ( i . e . , 48 hr for B . pseudomallei and 72 hr for B . mallei ) of infection ( a total of four comparison pairs ) , conserved B . mallei and B . pseudomallei orthologs showed nearly identical patterns with high Pearson correlation coefficient ( R ) values ranging from 0 . 94 to 0 . 97 , regardless of the host tissue type ( Fig . 5 ) . Therefore , there was no indication of significant modifications of the expression schemes in the genes required by B . mallei to thrive in BALB/c mice compared to those in B . pseudomallei . This is consistent with the findings of our previous gene expression studies in culture and in vivo , which also showed similar gene expression patterns in B . mallei and B . pseudomallei [20] , [33] , [34] , [35] . These data suggest that , during the early stage , genomic reduction proceeds conservatively , not seriously affecting the indigenous gene expression patterns . In contrast to B . mallei , most of the transcription units in the insect symbiont Buchnera were altered , most likely due to complex genomic alterations accumulated over a long period of time [2] . In this study , we unveiled the mechanics of genomic deletions and rearrangements that occur in the early stage of bacterial specialization in the host , by conducting comparative analyses of B . mallei and its parental species , B . pseudomallei . It became clear that stepwise IS intervention was the main driving force mediating a large genomic reduction in B . mallei . Expansion of ISBma1 and ISBma2 in a clone of B . pseudomallei set the stage for the wide spread of IS407A , allowing its proliferation to sites , to which the element itself may rarely target . Actual genomic deletions and rearrangements occurred through recombination reactions mainly among IS407A and also among ISBma2 ( Fig . 2 ) . These processes achieved highly efficient deletions of dispensable genomic regions , causing only small disruptions to the portions of the genome that were maintained . This was possible due to the guidance by selective forces in the host and via the intrinsic flexibility of the compactly IS-blended genome . The B . mallei genome currently appears to still be structurally flexible with regard to deletions but is now less flexible with regard to genomic rearrangements and additional transpositions . This may indicate that the genomic evolution in B . mallei has been moving into a second stage , in which large-scale genomic alterations are reduced and nucleotide-level erosion has become more important . On the other hand , a large number of genes disrupted by frameshift mutations in SSRs were found in the B . mallei genome . The loss of function encoded by these genes and of flagella via disruption in fliP by IS407A ( Table S1 ) , could be part of the adaptive evolution for survival in the host environment , which will eventually lead to genome size reduction by erosion over time . Widespread relics of IS elements found in diverse symbionts and obligate pathogens [1] , [3] , [8] clearly suggest that a similar sequential IS intervention , modeled in Figure 6 , may illustrate a general mechanism , by which elaborate genome transition occurs during early bacterial evolution after establishing constant association with the host .
All research involving live animals was conducted in compliance with the Animal Welfare Act and other federal statutes and regulations relating to animals and experiments involving animals and adhered to the principles stated in the Guide for the Care and Use of Laboratory Animals , National Research Council , 1996 . All mouse experiments conducted in the USAMRIID ( US Army Medical Research Institute of Infectious Diseases ) were approved by the Association for Assessment and Accreditation of Laboratory Animal Care International . The type strains for B . mallei ( ATCC23344 ) [20] and B . pseudomallei ( K96243 ) [21] were previously sequenced . Strains FMH , JHU , and GB8 horse 4 were direct derivatives of strain ATCC 23344 after passages in the human or horse , and these strains were also sequenced previously [24] . B . mallei strains NCTC10229 , NCTC10247 , and SAVP1 were sequenced with full closure and manually annotated as previously described [20] . The remaining three strains ( 2002721280 , ATCC10399 , and PRL-20 ) were sequenced to 8× Sanger sequence coverage by the whole genome shotgun method [36] without closure , and assembled using the Celera Assembler [37] , and contigs were oriented by alignment to the reference strain ATCC23344 using PROMER [38] . ORFs were predicted and annotated automatically using GLIMMER [39] , [40] . Pseudo-chromosomes were constructed from the ordered scaffolds , using manual examination where necessary . Similarly , B . pseudomallei strains 1106a , 1710b , and 668 were sequenced with full closure and manual annotation , while 1655 , 406e , S13 , and Pasteur 6068 were sequenced without closure and annotated automatically . For the analyses of genomic deletions and rearrangements in B . mallei and B . pseudomallei , 5 , 799 predicted protein sequences from the B . pseudomallei type strain K96243 were compared with the nucleotide sequences of the genomes of B . mallei ( ATCC 23344 , 2002721280 , ATCC 10399 , FMH , JHU , GB8 horse 4 , PRL-20 , NCTC 10229 , NCTC 10247 , SAVP1 ) and the other strains of B . pseudomallei ( 1106a , 1106b , 1655 , 1710a , 1710b , 406e , 668 , Pasteur , S13 ) using tblastn ( http://blast . wustl . edu ) . For the mapping of the insertions of ISBma1 , ISBma2 , and IS407A in the genomes of B . mallei and B . pseudomallei , the entire sequences of the IS elements were searched against the 20 genomes using blastn ( http://blast . wustl . edu ) . For the analysis of association of the IS elements with genomic deletions and rearrangements in B . mallei and of the target sequences in the genomes , strain ATCC 23344 represented all of its immediate derivatives , FMH , JHU , and GB8 horse 4 , to avoid redundancy in the data , because the three strains showed identical patterns . To compare the patterns of genome rearrangements in the B . mallei strains , the positions of the BRUs in each strain of B . mallei relative to B . pseudomallei K96243 were visualized using a genome-comparative software tool ACT ( [41]; http://www . sanger . ac . uk/Software/ACT ) , and the displays were compared in parallel among the strains . We also examined B . mallei and B . pseudomallei for intergenic regions that potentially containing promoters , putative regulatory genes , and disruptions of putative operons to estimate the possibility of causing gene expression divergence . For intergenic region comparisons , up to 100 bp upstream of the start codon , or up to as much as available if the neighboring gene was closer , of the genes that contain at least 50 bp of an untranslated upstream region were retrieved from the genome of B . pseudomallei K96243 . Then , these sequences ( 2 , 268 and 1 , 566 from chromosomes 1 and 2 , respectively ) were searched against the genomes of B . mallei and B . pseudomallei using blastn ( http://blast . wustl . edu ) , and the length-match as well as the identity values of the orthologous regions were calculated . Putative operons reported by Rodrigues et al . from the genome of B . pseudomallei K6243 [31] were used to match the orthologous gene clusters in the genome of B . mallei ATCC 23344 , and these gene clusters were examined for any disruptions caused by IS elements . All the genome sequences of B . mallei and B . pseudomallei used in this study are available through the Pathema web site ( http://pathema . jcvi . org/cgi-bin/Burkholderia/PathemaHomePage . cgi ) at the J . Craig Venter Institute ( http://www . jcvi . org/ ) . A phylogenetic tree was constructed with the strains of B . mallei and B . pseudomallei based on the insertion patterns of and the role played in the genomic deletions and rearrangements by the three major IS elements , ISBma1 , ISBma2 , and IS407A . All the data used are shown in Tables S1 and S2 and Figure 2 . Bootstrapped maximum parsimony trees were calculated using the PAUP package with default parameters , and a consensus tree was produced from the bootstrap replicates . Branches with bootstrap scores of less than 50 were collapsed in the tree . Among the duplicated target regions encompassing the IS elements ISBma1 , ISBma2 , and IS407A , those regions that had perfectly matching sequences were first collected . Then , among the sequences from unmatched pairs , those that occurred in more than two strains were assumed to be un-mutated valid sequences and , therefore , were added to the data pool for the analysis . Strain ATCC 23344 represented all its direct derivatives ( FMH , JHU , and GB8 horse 4 ) in this analysis to avoid redundancy in the data . The collected sequences were aligned with Clustal X , and the alignments were graphically visualized using Sequence logos [42] . Exposure of mice to bacterial aerosol was performed as described by Roy et al . [43] . Fresh overnight cultures of B . pseudomallei DD503 [44] and B . mallei ATCC 23344 were prepared in LB or in LBG ( LB supplemented with 4% glycerol ) , respectively , at 37°C with aeration ( 250 rpm ) . Thirty female BALB/c mice six to eight weeks old ( National Cancer Institute , Frederick , MD , USA ) were infected with these bacteria: nine mice each with B . pseudomallei and B . mallei for the gene expression studies , and six mice each for the bacterial load assays . The mice infected with B . mallei received an inhaled dose of 7 . 2×103 cfu ( 7 . 2×LD50 ) , and those mice infected with B . pseudomallei received 1 . 8×104 cfu ( 18×LD50 ) , as estimated by colony counting on agar plates . The infected mice were provided with rodent feed and water ad libitum and maintained on a 12-hr light cycle . After 24 and 48 hr ( for both B . mallei and B . pseudomallei ) or 72 hr ( for B . mallei ) of infection , five mice from each point in time were euthanized in a CO2 chamber , and their spleens and lungs were removed . Due to animal mortality , a 72 hr point in time was not possible for B . pseudomallei . The organs from two randomly picked mice were saved for bacterial load estimations , and the rest were homogenized in 1 ml of Trizol ( Invitrogen Corp . , Carlsbad , CA , USA ) using a Tissue-Tearor ( BioSpec Products , Bartlesville , OK , USA ) . Total RNA was purified according to the manufacturer's recommendations ( Invitrogen Corp . , Carlsbad , CA , USA ) . The bacterial load in the mouse organs was estimated as described by Ulrich and DeShazer [32] . Total RNA , both bacterial and mouse , from the same organ types from three mice was pooled to compensate for potential individual variation . These pooled RNA samples were used for the experiments without further purification of the bacterial RNA because RNA from mice does not cross-hybridize to the B . mallei microarray at a level affecting the legitimate interactions between the B . mallei array and the Burkholderia transcriptome [35] . The B . mallei whole genome array used in this study for both B . mallei and the closely related B . pseudomallei ( average gene identity at the nucleotide level of 99% ) was described in detail previously [33] . The B . mallei- and B . pseudomallei-infected organ samples were paired for the hybridization reactions based on early and late pathological states . A total of eight hybridization reactions or four different comparisons were performed , each of which was replicated in flip-dye pairs and the final ratios were calculated as log2 ( B . pseudomallei gene expression intensity/B . mallei gene expression intensity ) . Labeling of the probes , slide hybridization , and slide scanning were carried out as previously described [35] . The independent TIFF slide images from each channel were analyzed using TIGR Spotfinder to assess the relative expression levels , and the data were normalized using a local regression technique LOWESS ( LOcally WEighted Scatterplot Smoothing ) with the MIDAS software ( <http://www . jcvi . org/cms/research/software> , The J . Craig Venter Institute , Rockville , MD , USA ) . The resulting data were averaged from triplicate genes on each microarray and from duplicate flip-dye arrays for each experiment .
|
It has been known for some time that bacteria undergo genome-reduction when they transition from a free-living state to a constantly host-restricted state . High levels of IS element expansion were also found in these bacteria , and the IS elements were suggested to play a role in genome reductive evolution . Here we provide evidence for stepwise IS actions as the exclusive mechanism that mediates bacterial genomic changes during the early stage of constant host-bacterial association , by unveiling the processes that resulted in the development of B . mallei genome . We show the details of the multi-level interplay of IS elements , which facilitate the wide spread of the IS copies , and the overall mechanics in genome reduction and rearrangement . These processes appeared to operate as chain reactions mediating elaborate genomic transition , without seriously affecting the original gene expression patterns . The absence of differential gene expression in the resulting genome suggests that changes in transcriptional regulation that are often observed in other old bacterial genomes may take place subsequent to the IS-mediated steps , along with gradual nucleotide-level changes .
|
[
"Abstract",
"Introduction",
"Results/Discussion",
"Materials",
"and",
"Methods"
] |
[
"evolutionary",
"biology/microbial",
"evolution",
"and",
"genomics",
"evolutionary",
"biology/genomics",
"evolutionary",
"biology/bioinformatics",
"evolutionary",
"biology/evolutionary",
"and",
"comparative",
"genetics"
] |
2010
|
The Early Stage of Bacterial Genome-Reductive Evolution in the Host
|
Reporting bias in the literature occurs when there is selective revealing or suppression of results , influenced by the direction of findings . We assessed the risk of reporting bias in the epidemiological literature on health-related behavior ( tobacco , alcohol , diet , physical activity , and sedentary behavior ) and cardiovascular disease mortality and all-cause mortality and provided a comparative assessment of reporting bias between health-related behavior and statin ( in primary prevention ) meta-analyses . We searched Medline , Embase , Cochrane Methodology Register Database , and Web of Science for systematic reviews synthesizing the associations of health-related behavior and statins with cardiovascular disease mortality and all-cause mortality published between 2010 and 2016 . Risk of bias in systematic reviews was assessed using the ROBIS tool . Reporting bias in the literature was evaluated via small-study effect and excess significance tests . We included 49 systematic reviews in our study . The majority of these reviews exhibited a high overall risk of bias , with a higher extent in health-related behavior reviews , relative to statins . We reperformed 111 meta-analyses conducted across these reviews , of which 65% had statistically significant results ( P < 0 . 05 ) . Around 22% of health-related behavior meta-analyses showed small-study effect , as compared to none of statin meta-analyses . Physical activity and the smoking research areas had more than 40% of meta-analyses with small-study effect . We found evidence of excess significance in 26% of health-related behavior meta-analyses , as compared to none of statin meta-analyses . Half of the meta-analyses from physical activity , 26% from diet , 18% from sedentary behavior , 14% for smoking , and 12% from alcohol showed evidence of excess significance bias . These biases may be distorting the body of evidence available by providing inaccurate estimates of preventive effects on cardiovascular and all-cause mortality .
The literature on the association between behavioral risk factors ( e . g . , smoking , alcohol , physical inactivity , and unhealthy diet ) and cardiovascular diseases—the single largest cause of death globally [1]—has grown exponentially in the last decades [2–39] . Observational epidemiological studies are the dominant design assessing these associations , since clinical trials cannot always be ethically or logistically conducted [40] . Systematic review methods are used to synthesize and evaluate this growing body of evidence . It is important to evaluate the methodological risks of bias in systematic reviews [41] , as well as the impact that reporting bias can have on the findings of reviews [42 , 43] . Reporting bias is one of the most common biases identified in the literature . It includes selective publication of studies or outcomes of studies [44 , 45] based on factors other than the study quality , such as nominally statistically significant results ( P < 0 . 05 ) [46 , 47] or authors’ “pedigree” [44 , 45 , 48] . These practices threaten the completeness and validity of scientific evidence [46] by distorting the estimates of causal effects of interventions or exposures on diseases [49] . The extent of reporting bias could differ between bodies of evidence consisting of randomized trials , such as drug studies , compared to observational studies , such as studies of health behavior . Different levels of reporting bias in the literature on health behavior may lead to inaccurate estimates of preventive effects on cardiovascular and all-cause mortality and therefore offer incorrect guidance for policymaking . To gain a better understanding of the potential reporting bias in the literature on health-related behavior and cardiovascular disease mortality and all-cause mortality , we examined reporting and other risks of bias in a sample of systematic reviews published between 2010 to 2016 . Our analysis also provided a comparative assessment of the reporting bias between health-related behavior and statins used in primary prevention .
The majority of the systematic reviews exhibited a high overall risk of bias ( n = 44 , 90% ) ( Fig 2 ) . Among the four ROBIS domains , domain 1 ( study eligibility criteria ) presented the best scores , with 32 ( 65% ) out of 49 reviews showing a low risk of bias . In domain 2 ( identification and selection of studies ) , 2 ( 4% ) reviews were scored as unclear , 40 ( 82% ) showed a high risk , and 7 ( 14% ) a low risk of bias . Whereas , in domain 3 ( data collection and study appraisal ) , 7 ( 14% ) reviews were scored as unclear , 28 ( 57% ) scored with high risk , and 14 ( 29% ) with low risk of bias . Finally , in domain 4 ( synthesis and findings ) , 2 ( 4% ) review was scored as unclear , 30 ( 61% ) with high risk , and 17 ( 35% ) with low risk of bias ( Fig 2 and S3 Table ) . Comparing risk of bias in the reviews across research areas , sedentary behavior performed worst in domain 1 ( study eligibility criteria; 70% of reviews were regarded as having high risk of bias ) . All research areas performed poorly in domain 2 ( identification and selection of studies ) , with high risk of bias ranging from 70% in smoking reviews to 90% in both sedentary behavior and statin reviews . Alcohol ( 70% ) and diet ( 60% ) reviews presented high risk of bias in domain 3 ( data collection and study appraisal ) . Sedentary behavior ( 90% ) , smoking ( 70% ) , and diet ( 70% ) reviews presented high risk of bias in domain 4 ( synthesis and findings ) . Overall , statin reviews presented the best scores in the ROBIS assessment compared to other research areas . Among statin reviews , a low risk of bias was identified in 60% in domain 1 , 10% in domain 2 , 50% in domain 3 , and 60% in domain 4 ( Table 1 and Fig 3 ) . We identified 111 meta-analyses ( exposure–outcomes associations ) that were performed across the 49 included reviews . On average , each meta-analysis synthesized results from 9 primary studies ( ranging from 2 to 81 ) , including 331 , 688 participants ( ranging from 595 to 3 , 674 , 042 ) and 19 , 012 deaths ( ranging from 33 to 320 , 252 ) ( Table 2 and S4 Table ) . Of the 111 meta-analyses , 72 ( 65% ) showed a nominally statistically significant result at P < 0 . 05 . Nominally statistically significant results ( P < 0 . 05 ) were found in 92% of the meta-analyses from sedentary behavior and 100% of the meta-analyses from physical activity and smoking . Alcohol and statin reviews had 38% and 45% of meta-analyses with P < 0 . 05 results , respectively ( Table 2 and S4 Table ) . We conducted a sensitivity analysis by restricting the sample in each research area to meta-analyses with ≥10 primary studies . In this subsample ( n = 29 ) , 86% of the meta-analyses showed statistically significant results at P < 0 . 05 , as compared to 65% in the entire sample of meta-analyses . These results varied by research area , ranging from 60% in statin meta-analyses to 100% in physical activity , sedentary behavior , and smoking meta-analyses ( Table 3 ) . Small-study effect was present in 31% of the meta-analyses . The proportions of meta-analyses in the sensitivity analysis with small-study effect were 80% for physical activity , 50% for sedentary behavior , 29% for alcohol , and 33% for smoking . Diet and statin meta-analyses had no evidence of small-study effect . Around 38% of the health-related behavior meta-analyses with ≥10 primary studies presented small-study effect , as compared to zero in statin meta-analyses ( Table 3 ) . Excess significance was identified in 27% of the meta-analyses with ≥10 primary studies: 100% of the meta-analyses for sedentary behavior , 50% for diet , 40% for physical activity , 20% for alcohol , and 17% for smoking . Around 33% of the health-related meta-analyses with ≥10 primary studies showed evidence of excess significance , as compared to zero in statin meta-analyses ( Table 3 ) . Overall , after excluding small individual studies ( with <200 deaths ) from meta-analyses , results from small-study effect and excess significance tests did not change ( S5 Table ) .
This study aimed to assess the extent of reporting bias among recent meta-analyses that examined the associations of health behavior and statins with cardiovascular and all-cause mortality . We found evidence of reporting bias across all health-related behavior areas . The degree of reporting bias varied by the method used to assess it . Reporting bias was present in 20% ( according to excess significance test ) or 18% ( according to small-study effect test ) of all meta-analyses included ( health behavior and statins ) . Evidence of reporting bias was found in between a quarter and one-fifth of health-related behavior meta-analyses ( 22% small-study effect and 24% excess significance ) but in none of the statin meta-analyses ( 0% ) . In lifestyle epidemiology , the interpretation of evidence for researchers and policymakers is challenging for several reasons [61] . As observational studies are the dominant designs in this area , spurious associations can arise due to confounding or several sources of bias . The impact of such biases on statistical findings and interpretation of findings has been poorly reported and discussed [62] . Therefore , meta-analytical synthesis of the evidence in lifestyle health behavior epidemiology may provide precise but spurious results [63] . Reporting bias is a major threat to the validity of the relevant body of evidence . Our results suggest that around 20% of the meta-analyses on health-related behavior and cardiovascular disease mortality and all-cause mortality may be susceptible to reporting biases . The existence of reporting bias in the literature has several explanations . Failure to submit manuscripts of analyses that did not produce statistically significant results ( “the file-drawer problem” [46] ) and the low likelihood of publication of small studies ( regardless of statistical significance ) [44] are two possible reasons . The selective reporting of certain analyses with statistically significant results is another likely source of reporting bias [44 , 46 , 47] . Each of the research areas we examined is likely to be linked to variable levels of reporting bias due to the different economics , dynamics , and conflicts of interest in each discipline [64 , 65] . Interpreting the literature as a whole is challenging , considering the numerous biases that may affect the reliability and integrity of the scientific enterprise [66 , 67] . To obtain a complete picture of the evidence ( i . e . , without reporting bias ) , it is important to know the results from all conducted studies on a given research question [68] . In our study , results from meta-analyses of health-related behavior and cardiovascular disease mortality and all-cause mortality were more likely to be affected by reporting bias compared to statin meta-analyses ( 22% and 24% versus 0% , respectively ) . The literature of health-related behavior is almost exclusively composed by observational studies , whereas statins are most often studied using randomized controlled trials . Reporting bias may be less frequent among trials than observational studies because several efforts to increase transparency and reproducibility of results have been adopted over the history of randomized controlled trials [69] . These include the mandatory registration of all clinical trials in humans and disclosure of all results [70] . As of more recently , data sharing statements of clinical trials are also required [71] . Observational epidemiologic studies should embrace these reproducible research practices to reduce reporting bias in the literature [68–70 , 72] . These practices could involve key elements of the scientific process , including ( a ) methods ( e . g . , rigorous training in statistics ) , ( b ) reporting and dissemination ( e . g . , disclosure of conflicts of interest ) , ( c ) reproducibility ( e . g . , open data ) , ( d ) evaluation ( e . g . , pre- and postpublication peer review ) , and ( e ) incentives ( e . g . , funding replication studies ) [72] . Improving methodological training involves aspects of both research design and statistical analyses—for example , correct interpretation of P values [73] , acknowledging the importance of statistical power , and improving the accuracy of effect sizes [72] . Protecting against cognitive biases is another major issue that has been overlooked [72] . Protecting against conflict of interests , especially financially related , is an imperative to achieve reproducible science . In addition to disclosure of potential conflicts of interest , promoting preregistration of study procedures and analytical plan may prevent reporting bias favoring positive results [72] . Funding replication of studies and encouraging openness in science and reproducibility practices by making datasets , scripts , and software publicly available may increase transparency and credibility of scientific claims [72] . For instance , food industry–sponsored studies are more likely to report conclusions favorable to the sponsors [74] but frequently lack transparency on acknowledgment of the funding source [75] . Further examples of reproducibility practices have been described and discussed by Munafò and colleagues [72] . To our knowledge , our analysis is the first comparative assessment of reporting bias across different fields of health-related behavior and statins . Our findings were based on well-established statistical tests developed to detect different aspects of reporting bias , as well as a complementary assessment of the risk of bias of systematic reviews using the ROBIS tool . We selected the ROBIS tool as it has greater specification to assess risk of bias compared to other tools . For instance , the “Assessing the Methodological Quality of Systematic Reviews” ( AMSTAR ) that has been used to evaluate the methodological quality of systematic reviews has constructs that are more related to quality of reporting than risk of bias [76 , 77] . Risk of bias is linked to methodological quality of systematic reviews but provides further evaluation on how methodological limitations were considered to form conclusions . In this sense , the ROBIS tool is increasingly being used to assess risk of bias not only in systematic reviews [41 , 76 , 78] but also in guideline committees that evaluate evidence level ( e . g . , Australian government , National Health and Medical Research Council ) . Our ROBIS tool results showed that most of the systematic reviews had high risk of bias . Similar findings have been observed in previous studies appraising risk of bias in other research areas using the ROBIS tool [76 , 78] . For instance , 18 ( 58% ) out of 31 systematic reviews evaluating the effectiveness of intra-articular hyaluronic acid injection in treating knee osteoarthritis had high ( n = 16 ) or unclear ( n = 2 ) risk of bias [78] . Another survey assessing systematic reviews about psoriasis found that most reviews ( 86% ) were classified as high risk of bias [76] . It is noteworthy that high risk of bias was found even for systematic reviews exhibiting high methodological quality as assessed through AMSTAR [76] . Our ROBIS assessment indicated that identification and selection of studies ( i . e . , appropriate range of databases , terms and filters used , and efforts to minimize errors in selection of studies ) are major concerns . These biases in the review process could explain , at least in part , reporting bias results obtained from small-study effect and excess significance tests . The synthesis and findings domain also revealed potential risk of bias due to insufficient inclusion of studies and appropriate synthesis of estimates . This domain also reflects between-study variation , robustness of findings ( e . g . , sensitivity analyses ) , and biases in synthesis findings ( i . e . , if evaluated by systematic reviews ) . We used small-study effect and excess significance tests to appraise reporting bias in the literature , which are the most commonly recommended and used methods [79] . However , results from these tests might also reflect methodological and clinical heterogeneity , or even chance [42] . In fact , most meta-analyses contained moderate to high heterogeneity ( based on I2 statistic; S4 Table ) . Results from an Egger test ( small-study effect ) can give spurious false positive results due to correlation between log of effect size and its variance , especially in the presence of heterogeneity between studies in a meta-analysis . An alternative better-performing test has been proposed by Peters to identify reporting bias in meta-analyses , but it requires data from a 2 × 2 table [80] . Such data were rarely reported in individual studies in the meta-analyses of observational studies . As also noted by Tsilidis and colleagues [81] , meta-analyses commonly use maximally adjusted relative risks rather than unadjusted relative risks calculated from 2 × 2 tables . For such data , the use of the Egger test is appropriate . The egger test and excess significance test have low power to detect reporting bias and do not give indication about what the sources of bias are . Therefore , we performed sensitivity analyses , retaining only meta-analyses with ≥10 primary studies . In this subsample of meta-analyses , evidence of reporting bias was higher than the entire sample ( small-study effect: 31% versus 18%; excess significance: 27% versus 20% ) . Differences between primary results and sensitivity analyses are likely related to low power of reporting bias tests , which could lead to false negative results in the former group of meta-analyses . Therefore , our estimates of reporting bias in the meta-analyses are possibly conservative . The ranking of research areas according to levels of reporting bias was also different between the main analysis and the sensitivity analysis ( i . e . , meta-analyses with ≥10 primary studies ) . For instance , meta-analyses of sedentary behavior appeared most sensitive to this restriction , as the estimated proportion of reporting bias increased when calculated with either the small-study effects ( from 9% to 50% ) or excess significance tests ( from 18% to 100% ) . A possible explanation could be the small fraction of meta-analyses with ≥10 primary studies ( 2 out of 12 ) in this relatively new research field [82] . It is important to acknowledge that certain methodological decisions we made may have introduced bias in the sample of reviews selected or may compromise the generalizability of our findings . We excluded systematic reviews on alcohol published in Chinese language ( n = 2 ) , which potentially have high risk of bias [83] . In addition , we restricted our analyses to systematic reviews published in this decade only ( 2010–2016 ) , which explains the small number of included meta-analyses in some research areas . This may have limited comparisons of the extent of reporting bias between research areas investigated . Our results may not provide a complete historical assessment of reporting bias in these areas . Nevertheless , our results reflect reporting bias in the literature of recent and relevant public health topics and from a time period when reporting standards have been improving due to , e . g . , the widespread use of various manuscript reporting checklists [84] . Recent systematic reviews contain a higher number of primary studies than older systematic reviews and synthesize evidence of emerging fields that have flourished only recently ( i . e . , sedentary behavior ) . In conclusion , we found evidence of reporting bias in approximately one-fifth of recent meta-analyses of observational studies of health-related behavior ( physical activity , sedentary behavior , smoking , alcohol consumption , diet ) and cardiovascular and all-cause mortality . Such a level of reporting bias may , to some extent at least , distort conclusions arising from this body of evidence . Contrarily , we found no evidence of reporting bias in meta-analyses of randomized controlled trials of statins .
We searched Medline ( through PubMed ) , Embase ( i . e . , excluding Medline ) , Cochrane Methodology Register Database , and Web of Science for systematic reviews published between 2010 and 2016 . We restricted our search to recent systematic reviews for several reasons . These systematic reviews belong to a “birth cohort” of systematic reviews published after the launch of the Meta-analysis of Observational Studies in Epidemiology ( MOOSE ) [85] and Preferred Reporting Items for Systematic Reviews and Meta-Analyses ( PRISMA ) [86] guidelines and are expected to have lower risk of bias . As we were interested in comparing levels of bias across different research areas , this restriction may have reduced confounding due to date of publication . We restricted the search , as well as the successive phases of our study , to systematic reviews aiming to investigate the associations of health-related behavior ( tobacco , alcohol , diet [fat , fruits and vegetables , salt , and sugar] , physical activity , and sedentary behavior ) and statins with cardiovascular disease mortality ( overall cardiovascular mortality and cause-specific deaths from cardiovascular disease ) and all-cause mortality . We accepted any definition for the exposures and the outcomes as defined in the original systematic reviews . The keywords used in the search are described in S1 Text , and files exported from databases during search strategy with all studies screened and selected are available at https://osf . io/wpb69/ . Systematic reviews were screened and selected ( by two reviewers , and disagreements solved by a third reviewer ) based on the following eligibility criteria: ( i ) sought to investigate an exposure–outcome association in a nonclinical population; ( ii ) systematically searched for primary studies and performed a meta-analysis ( i . e . , weighted summary effect size ) using results from primary studies; ( iii ) selected only observational studies ( cohort and case-control studies ) if a health-related behavior review and only randomized controlled trials if a statin review; ( v ) reported data from each primary study included in the meta-analysis ( S1 Text ) . We decided a priori that a random sample of up to 20 systematic reviews per research area ( tobacco , alcohol , diet , physical activity , sedentary behavior , and statins ) would be included to compare levels of reporting bias in the relevant literature . If our search retrieved fewer than 20 meta-analyses in a given research area , we included them all . A similar study-selection strategy was recently used in a study evaluating publication bias in meta-analyses of individual studies [87] . These methods were decided a priori as described in the analysis plan available at https://osf . io/wpb69/ ( not published prior to the identification and selection of systematic reviews ) . Reporting bias could be related to overall risk of bias in a review . Therefore , four reviewers ( JPRL , NC , AF , LP ) , working in pairs , independently assessed the risk of bias in the included systematic reviews using the ROBIS tool [41] . ROBIS comprises three phases: ( 1 ) assess relevance; ( 2 ) identify concerns with review process; ( 3 ) judge risk of bias in the review . To assess relevance , we extracted the target question from each review using the PICOS acronym ( participants , interventions , comparisons , outcomes ) or equivalents for etiological questions ( participants , exposure , comparisons , outcomes ) . In phase 2 , we assessed the risk of bias in four domains related to the review process: ( 1 ) study eligibility criteria; ( 2 ) identification and selection of studies; ( 3 ) data collection and study appraisal; and ( 4 ) synthesis and findings . Questions included in each of the four domains are available in S3 Table . Questions were answered as “Yes , ” “Probably Yes , ” “Probably No , ” “No , ” and “No Information , ” with “Yes” indicating low risk of bias . In phase 3 , we summarized the concerns identified in each domain during phase 2 and risk of bias in the review as low , high , or unclear . Further details on the ROBIS tool are described elsewhere [41] . For each meta-analysis performed in the selected systematic reviews , we assessed the extent of reporting bias in the included literature via small-study effect [42] and excess significance tests [88] . To perform these tests , we extracted necessary data ( e . g . , effect size , confidence intervals , sample size , and number of events [deaths] ) for each primary study included in the main meta-analysis performed in the systematic reviews . We also used these data to reperform the meta-analyses ( i . e . , using random effect models , which was used in the majority of the original meta-analyses ) . We did this to describe the number of meta-analyses with nominally statistically significant results at P < 0 . 05 ( S1 Text ) . Small-study effect test ( also known as regression asymmetry test , proposed by Egger and colleagues ) evaluates whether smaller studies tend to overestimate the effect size estimates compared to larger studies . For this matter , the test evaluates whether the association between effect size ( e . g . , relative risk , odds ratio ) and precision ( standard error ) is greater than might be expected by chance . We considered a P value < 0 . 10 as a statistical significance threshold for small-study effect bias ( i . e . , suggesting evidence of reporting bias ) , as initially proposed by Egger and colleagues [42 , 89] and consistently used in the literature [42 , 66 , 81 , 87 , 90 , 91] . Excess significance test evaluates whether the O differs from the E . The E in each meta-analysis was obtained from the sum of power estimates of each primary study . The power estimate of each primary study depends on the plausible causal effect of each research area ( e . g . , smoking and cardiovascular mortality ) , which was assumed to be the effect of the most precise primary study ( smaller standard error ) in each meta-analysis [88] . We considered P < 0 . 10 ( one-side P < 0 . 05 for O > E ) as a statistical significance threshold for excess significance bias [43 , 88] . The excess significance is reported as a proportion of studies , with the higher proportion indicating more excess significance ( O > E ) and thus more evidence of reporting bias . Due to the low power of these bias tests , we performed a sensitivity analysis excluding meta-analyses with fewer than 10 studies to analyze the impact in the results . We also performed a sensitivity analysis excluding small individual studies ( fewer than 200 deaths ) within meta-analyses to evaluate whether results reflect reporting bias among small studies only . We performed all statistical analyses using Stata version 15 . 0 ( College Station , TX ) .
|
In the scientific literature , reporting bias occurs when communication and publication of results are influenced by the direction of findings . Reporting bias can distort scientific evidence and may misguide subsequent clinical and public health efforts . Our study provided an assessment of the degree of reporting bias in the literature on health-related behavior ( smoking , alcohol , diet , physical activity , and sedentary behavior ) and statins and their association with cardiovascular disease and mortality . We analyzed recently published systematic reviews . Most of the systematic reviews ( 90% ) had a high risk of bias related to study eligibility criteria , identification and selection of studies , data collection and study appraisal , and synthesis and findings . We found evidence of reporting bias in about one-fifth of health-related behavior meta-analyses but none of the statin-related meta-analyses . Readers should be aware of the extent of reporting bias in these research areas when interpreting meta-analytical results .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"medicine",
"and",
"health",
"sciences",
"chemical",
"compounds",
"behavioral",
"and",
"social",
"aspects",
"of",
"health",
"drugs",
"diet",
"physical",
"activity",
"organic",
"compounds",
"cardiovascular",
"medicine",
"mathematics",
"nutrition",
"statistics",
"(mathematics)",
"pharmacology",
"alcohols",
"research",
"and",
"analysis",
"methods",
"public",
"and",
"occupational",
"health",
"meta-research",
"article",
"mathematical",
"and",
"statistical",
"techniques",
"chemistry",
"research",
"assessment",
"statins",
"organic",
"chemistry",
"meta-analysis",
"systematic",
"reviews",
"cardiovascular",
"diseases",
"biology",
"and",
"life",
"sciences",
"physical",
"sciences",
"statistical",
"methods"
] |
2018
|
Reporting bias in the literature on the associations of health-related behaviors and statins with cardiovascular disease and all-cause mortality
|
Filoviruses are capable of causing deadly hemorrhagic fevers . All nonsegmented negative-sense RNA-virus nucleocapsids are composed of a nucleoprotein ( NP ) , a phosphoprotein ( VP35 ) and a polymerase ( L ) . However , the VP30 RNA-synthesis co-factor is unique to the filoviruses . The assembly , structure , and function of the filovirus RNA replication complex remain unclear . Here , we have characterized the interactions of Ebola , Sudan and Marburg virus VP30 with NP using in vitro biochemistry , structural biology and cell-based mini-replicon assays . We have found that the VP30 C-terminal domain interacts with a short peptide in the C-terminal region of NP . Further , we have solved crystal structures of the VP30-NP complex for both Ebola and Marburg viruses . These structures reveal that a conserved , proline-rich NP peptide binds a shallow hydrophobic cleft on the VP30 C-terminal domain . Structure-guided Ebola virus VP30 mutants have altered affinities for the NP peptide . Correlation of these VP30-NP affinities with the activity for each of these mutants in a cell-based mini-replicon assay suggests that the VP30-NP interaction plays both essential and inhibitory roles in Ebola virus RNA synthesis .
Filoviruses such as Ebola ( EBOV ) and Marburg viruses ( MARV ) are nonsegmented negative-sense RNA viruses that can cause deadly hemorrhagic fevers with up to 90% fatality [1] . The impact of EBOV is highlighted by the recent outbreak in West Africa involving over 28 , 000 cases and claiming more than 11 , 000 lives [2] . Key to the viral life cycle are the components of the viral nucleocapsid . The nucleocapsids of all nonsegmented negative-sense RNA viruses carry a viral RNA-dependent , RNA polymerase ( L ) , a phosphoprotein polymerase co-factor ( P or VP35 ) and a nucleoprotein ( N or NP ) , which encapsidates the viral genome . In the Mononegavirales order of viruses , L and NP interact through the phosphoprotein to carry out viral RNA synthesis . Filoviruses are unusual among mononegaviruses in that they encode an additional nucleocapsid component , VP30 . VP30 is a multifunctional protein and acts as a transcriptional activator [3] . EBOV VP30 promotes read-through of an RNA hairpin in the NP open reading frame to enhance viral transcription [4] . EBOV VP30 also assists stop-start transcription at gene junctions to promote transcription of downstream genes [5] . The N-terminal portion of VP30 contains phosphorylation sites , a zinc-binding site , and a RNA-binding site . Phosphorylation in the N-terminal region regulates association of EBOV VP30 with the nucleocapsid and alters the balance of viral transcription and RNA replication [5–8] . Binding of zinc is important for its transcriptional enhancement activity , and capacity to bind RNA may facilitate the interaction of VP30 with the viral genome [9 , 10] . The C-terminal domain of VP30 ( CTD , amino acids 139–288 ) forms a conserved dimer of two globular , α-helical domains assembled by the extension of one α-helix from each protomer across the dimer interface to contact the adjacent protomer [11 , 12] . EBOV VP30 binds directly to nucleocapsid components L and NP [13–15] . Hartlieb , et al . determined that it is the VP30 CTD that contains the binding site for the viral NP [11] . The uniqueness of filovirus VP30 among nonsegmented negative-sense RNA viruses has left many unanswered questions concerning its role in viral RNA transcription and replication . Despite available functional data , very little is known about the mechanisms by which VP30 carries out these functions or interacts with other viral components . Here , we have mapped the filovirus VP30-binding site on NP and determined crystal structures of the complexes for EBOV and MARV . We have used these structures to explore the function of the VP30-NP interaction using rational mutagenesis and mini-replicon assays . This work shows the importance of the EBOV VP30-NP interaction for viral RNA synthesis and provides clues to understand how this interaction contributes to the regulation of the filovirus nucleocapsid complex .
We first demonstrated that the EBOV VP30 CTD binds to the NP C-terminal portion ( 360–739 ) . The N-terminal half of NP ( a . a . 1–390 ) forms an ordered structure and plays important roles in NP-NP oligomerization and RNA binding [16 , 17] . In contrast , the C-terminal half of NP ( residues 391–739 ) is mostly disordered , although a short folded domain exists at its C terminus ( residues 645–739 ) [18] . Hypothesizing that VP30 may recognize a portion of NP within the C-terminal half of NP , we carried out isothermal titration calorimetry ( ITC ) using the VP30 CTD , which Hartlieb et al . has previously shown interacts with NP [11] and the C-terminal portion of NP . This experiment demonstrated that the EBOV VP30 CTD binds to the NP C-terminal portion with a KD of 21 . 3 ± 4 . 5 μM . We next used biolayer interferometry ( BLI ) to fine map the VP30-binding site on Ebola virus NP . We generated several biotinylated EBOV NP peptides: 360–420 , 419–496 , 560–617 and 644–739 based on sequence conservation among the five ebolavirus species . We immobilized the EBOV NP peptides on streptavidin-coated biosensors and bound them to EBOV VP30 CTD . Due to difficulties in curve fitting and the dimeric VP30 CTD likely acting as a multivalent binder , these BLI experiments were performed qualitatively . VP30 CTD bound to NP 560–617 , but not to any of the other NP peptides . We further used a series of shorter overlapping NP peptides: 560–580 , 580–600 , 600–617 , 560–600 and 580–617 , to define the VP30-binding site more narrowly . Only EBOV NP 580–617 and 600–617 bound to VP30 CTD ( S1 Fig ) . This result strongly suggested that the primary binding site for EBOV VP30 CTD lies within NP 600–617 . This region of NP contains a stretch of amino acids that is conserved across the ebolaviruses and shares homology with cuevaviruses and marburgviruses ( Fig 1A ) . To quantitatively assess the molecular interaction of EBOV NP 600–617 with VP30 CTD , we performed ITC using NP 600–617 attached to T4 lysozyme as a carrier protein . The fusion of NP to lysozyme permits the preparation of this protein recombinantly , the accurate measurements of protein concentration using UV absorbance and the efficient dialysis prior to ITC , which would have otherwise been complicated for a short peptide . We observed no binding of VP30 CTD to T4 lysozyme alone ( S2 Fig ) . EBOV VP30 CTD binds to the NP 600–617 peptide with a KD of 5 . 69 ± 0 . 04 μM , a modestly higher affinity than for NP 360–617 , confirming the findings from the BLI experimentation that this NP peptide is sufficient for VP30 binding . Because the EBOV NP 600–617 peptide is highly conserved across ebolaviruses and has homology across filoviruses , we performed cross-species binding experiments using ebolavirus NPs 600–617 and MARV NP 552–569 with VP30 CTDs ( MARV VP30 146–281 ) using ITC ( Fig 2 ) . Significant cross-reactivity is noted , especially among the ebolaviruses . Among the ebolavirus NPs , that of Ebola virus ( EBOV ) appears to be a stronger binder of the ebolavirus VP30s . Among the VP30s , that of Sudan ebolavirus ( SUDV ) is a stronger binder of the ebolavirus NPs . MARV VP30 recognizes its own NP peptide with reasonable affinity , 14 ± 1 μM . Cross-reactivity across genera also occurs . However , this binding is weaker than the binding observed within members of the ebolavirus genus , as expected from the weaker conservation of the MARV NP peptide and VP30 CTD to those of the ebolaviruses . These results are supported by previous experiments demonstrating that MARV VP30 was able to support transcription , albeit with much lower efficiency , when exchanged for EBOV VP30 in an EBOV mini-replicon reporter assay [3] . These data show that the interaction of each VP30 CTD with its NP 600–617 is conserved across the filoviruses . To further confirm our mapping of this interaction on the viral NP and show its importance in RNA synthesis , we performed viral mini-replicon assays . In this assay , cells are transfected with plasmids expressing the minimal Ebola virus protein components for RNA synthesis: NP , VP35 , VP30 and L , along with a plasmid expressing a negative-sense RNA encoding firefly luciferase flanked by Ebola virus leader and trailer sequences [20] . In this system , proper assembly and function of the RNA synthesis machinery leads to expression of the reporter . Leung , et al . has previously shown that when a truncated NP 1–600 is used in place of the full-length NP ( 1–739 ) , that no mini-genome activity is observed in this assay [21] . Conversely , Watanabe , et al . has demonstrated that NP 1–600 is sufficient for reporter production in similar assays [17] . Our own results with NP 1–600 indicate a complete loss of reporter activity consistent with the results of Leung , et al . [21] ( Fig 3 ) . The NP 1–600 truncation has been shown to form the helical oligomers typical of non-segmented negative-sense RNA viruses , to form nucleocapsid-like structures in the presence of VP35 and VP24 [17] , and to contain the NP RNA-binding site [16 , 22] . The NP 1–600 truncation , however , excludes our newly identified VP30-binding site . Increasing the length of the NP protein construct to residues 1–620 restores luciferase activity validating our VP30-binding site on NP ( residues 600–617 ) . As a further check on the importance of this site within NP , we exchanged the full-length , wild-type NP for an NP in which the conserved amino acids within 600–617 were scrambled ( 605-PVARPYAP-612 ) . The scrambled mutant NP showed large reductions in binding to the VP30 CTD in vitro by ITC ( S2 Fig ) and in luciferase activity in the mini-replicon assay ( Fig 3 ) . The in vitro binding data show that the NP 600–617 is necessary for VP30 binding and results from the coupled transcription/replication mini-replicon assays show that reporter activity is dependent on the NP-VP30 interaction . To understand the molecular basis for the VP30-NP interaction , we sought to obtain high resolution X-ray crystallographic data for the complex . The rather low affinity between VP30 CTD and NP 600–617 precludes the co-purification of the complex . Presumably , when both VP30 and NP are full-length and oligomeric , multivalent binding greatly increases the avidity of this interaction . However , for crystallographic studies we used truncated viral proteins to reduce conformational heterogeneity that may impede crystal formation . To overcome the low-affinity of the interaction , we expressed the EBOV NP peptide ( 600–627 ) as a fusion to the N terminus of EBOV VP30 CTD , and the MARV NP peptide ( 552–579 ) as a fusion to the N terminus of MARV VP30 CTD ( 146–281 ) . These portions of NP include the identified VP30-interaction sites , as well as an additional ten amino acids that act as linkers to the VP30 CTDs . Both protein fusions crystallized and X-ray diffraction data were collected . The EBOV NP-VP30 fusion protein diffracted to 2 . 20 Å with ten protomers in the asymmetric unit , while the MARV NP-VP30 construct diffracted to 3 . 25 Å with eight protomers in the asymmetric unit ( Table 1 ) . Each of the VP30 CTD protomers in both complex structures is bound to a NP peptide . Amongst the copies of EBOV VP30 in the asymmetric unit , the N-terminal-most amino acid visible is 140 or 141 while the visible C-terminus is amino acid 265 or 266 . Amongst copies of NP , visible residues range from 602 or 603 to 612 or 614 . All of the EBOV NP peptides bind to VP30 in an identical fashion . The high resolution of the data allows visualization of many solvent molecules , some near the interfaces of NP and VP30 as well as between VP30 protomers . There also appears to be a clustering of sulfate ions from the crystallization condition near the VP30 dimer interfaces . This clustering reflects the presence of several basic residues near the VP30 dimer interfaces . Amino acids from the EBOV NP peptide contacting VP30 are primarily those that are highly conserved across ebolaviruses . NP contacts VP30 on the EBOV VP30 globular C-terminal domain , distal to the VP30 dimer interface ( Fig 4A ) . The NP-VP30 interface is primarily hydrophobic . Very few NP side chains make hydrogen bonds to VP30 . However , NP R612 participates in a salt bridge to VP30 D202 and NP Y611 forms a hydrogen bond to VP30 E197 ( Fig 4B and 4C ) . The NP peptide makes several hydrogen bonds using its main-chain atoms to the VP30 side chains of Q203 , R213 , Q229 and W230 . Analysis of the EBOV VP30-NP complex structure with a sequence alignment of ebolavirus VP30s ( Fig 1B ) shows that of the 21 VP30 amino acids within 5 Å of the NP peptide , 18 are conserved across ebolaviruses . Of the three that show some polymorphism , proline at VP30 position 206 is conserved among variants of EBOV , but is a serine in other species of the ebolavirus genus ( SUDV , Reston , Bundibugyo and Taï Forest ebolaviruses ) . Glycine at EBOV VP30 position 198 is an asparagine in SUDV , and serine at position 221 is instead an alanine or asparagine in other ebolaviruses . Polymorphisms at positions 198 or 221 are not predicted to have a strong impact on NP-binding , as these residues are peripheral to the binding site . Of the NP residues , the eight contained within 605-APPAPVYR-612 are strictly conserved across all ebolaviruses with the exception of R612 , which is a lysine in SUDV ( Fig 1A ) . Amino acids in ebolavirus NPs flanking this sequence are poorly conserved , but appear not to make intimate contacts . In the MARV VP30-NP complex structure , NP amino acids 555 to 564 and VP30 147 to 273 are visible in all of the protomers . The overall structure of the MARV VP30 CTD is remarkably similar to that of EBOV VP30 CTD despite sharing only 35% sequence identity across the CTD regions visualized in the crystal structures ( Fig 4D ) . The conformation of the bound MARV NP peptide is also remarkably similar to that of EBOV especially across the NP proline-rich region ( Fig 4E and 4F ) . The conserved binding mode of these NP peptides is supported by the ITC binding data above where cross-genus binding is possible , albeit with lower affinity than the species- and genus-matched interaction pairs . However , the conformation of the NP peptide is more divergent between EBOV and MARV N-terminal to the proline-rich region . In this region of the NP-VP30 complex , there is evidence for divergence and the evolution of compensatory mutations . In EBOV NP , A605 packs against the EBOV VP30 207–229 α-helix , but in MARV NP , the presence of amino acid L557 at the equivalent position turns the NP backbone towards the C-terminal half of the MARV VP30 208–223 α-helix . The altered NP backbone conformation gives rise to compensatory mutations including E209Q and R213G ( EBOV to MARV , EBOV numbering ) . The E209Q allows the MARV NP main chain to hydrogen bond at this position and avoid an electrostatic clash with the NP main-chain carbonyls . The R213G change allows the MARV NP peptide to avoid a clash with this side chain . In addition , other amino acid differences make the MARV and EBOV proteins less apt cross-genus binding partners . These MARV amino acid differences include the lack of a basic residue in NP at 564 ( EBOV NP R612 ) to form a salt bridge with VP30 D209 ( EBOV VP30 D202 ) , as well as VP30 G236 ( EBOV VP30 Q229 ) , which can no longer make hydrogen-bonding contacts to the NP backbone . The lack of these interactions in MARV NP and VP30 may result in the modestly lower affinity observed in ITC experiments and make MARV NP and VP30 poor binding partners for their ebolavirus protein counterparts . Comparison of the EBOV and MARV VP30-NP complex structures presented here with the first structure of the EBOV VP30 C-terminal domain ( 2I8B . pdb [11] ) shows a small conformational difference in the VP30 α-helix composed of residues 201–216 , where the 2I8B . pdb conformation of this α-helix would clash with the NP peptide in our complex structures ( Fig 5 ) . To explore a potential conformational change experimentally , we used ITC to measure the heat capacity change upon binding of EBOV VP30 to NP ( ΔCp ) and the extrapolated temperature at which the entropy of the binding event is zero ( TS ) . We used these two values to calculate the buried non-polar surface area upon binding and the number of amino acids rigidified upon binding [23 , 24] ( S3 Fig ) . The buried non-polar surface area calculated from the ITC experiments is 520 Å2 , which is similar to the 620 Å2 of total buried surface area observed in the EBOV NP-VP30 crystal structure . The calculated number of amino acids rigidified upon binding is approximately ten amino acids , similar to the 10–13 amino acids of NP observed in the complex structure . Because this region of NP is predicted to be disordered [25] , the observed rigidification can likely be attributed to the NP peptide . The buried non-polar surface area and amino acid rigidification data suggest that the binding site on VP30 is pre-formed . To reconcile this finding with the differences observed between 2I8B . pdb and the structures presented here , we further compared our structures to the Reston ebolavirus ( RESTV ) VP30 CTD ( 3V7O . pdb [12] ) and a more recently determined structure of the EBOV VP30 CTD ( 5DVW . pdb [26] ) . In both of these un-complexed structures , the VP30 α-helices composed of residues 201–216 align closely to the EBOV and MARV VP30-NP complexes , supporting our finding of a preformed NP-binding site on VP30 . A closer examination of 2I8B . pdb reveals that the modest difference in conformation is likely induced by a neighboring protomer in the crystal lattice . With crystal structures of the VP30-NP complexes in hand , we designed targeted EBOV VP30 mutations to alter the affinity of NP for VP30 . To strongly impact the NP-VP30 interaction , we mutated our structure-selected amino acid positions to arginine , with the exception of a partially buried tryptophan residue , which we conservatively mutated to phenylalanine ( W230F ) . We expressed the mutant VP30 CTD proteins recombinantly and tested their ability to interact with the NP peptide using ITC . Despite the high concentrations of protein used in these assays , no binding was observed for VP30 E197R , W230F , Q203R or S234R , indicating that the equilibrium dissociation constants for these mutant dimeric VP30 CTDs with monomeric NP peptide is likely greater than 200 μM . The negative impact of these VP30 CTD mutants on NP peptide binding validates the NP-VP30 interaction sites observed in the crystal structures . Two of our EBOV VP30 CTD mutants showed a reduced , yet measurable , affinity for NP peptide: D202R and Q229R ( Table 2 ) . Finally , a single mutant , P206R , showed an increased affinity for the NP peptide . P206 lies in EBOV VP30 α-helix 201–216 . Despite being conserved in Ebola viruses , the presence of a proline residue in the middle of a helix suggests that this proline may destabilize the helical conformation of this region leading to a non-ideal NP peptide-binding site . This notion is further supported by the presence of a serine at position 206 in other ebolaviruses and a valine in MARV . Indeed , SUDV VP30 recognizes EBOV NP with higher affinity than does EBOV VP30 , despite being otherwise highly conserved across the NP-binding site ( Figs 1B and 2 ) . We further examined the interactions of these mutants and NP in the context of full-length proteins using co-immunoprecipitation of NP with HA-tagged VP30 ( Fig 6A ) . Under our stringent wash conditions ( 50 mM TrisCl pH 8 . 0 , 300 mM NaCl , 2 mM EDTA , 10% glycerol , 1% Igepal CA-630 , 2 mM beta-mercaptoethanol ) and treatment of lysates with benzonase nuclease to examine protein-protein interactions , only the wild-type VP30 and the VP30 P206R pulled down co-transfected NP . Curiously , VP30 P206R , itself , immunoprecipitated poorly on the anti-HA beads despite its presence in the soluble protein fraction and pulling down a significant amount of NP . This result was highly reproducible and was only observed in the presence of co-transfected NP and not co-transfected VP35 ( Fig 6B ) . These results suggest that the poor pull down of VP30 P206R is a result of its intimate interaction with the highly oligomeric NP and is not due to improper protein folding . As EBOV VP30 has also been reported to interact with VP35 [6] we performed a similar co-immunoprecipitation analysis using FLAG-tagged full-length VP35 . For none of the VP30 variants , including the wild-type , do we observe VP35 pulldown with VP30 ( Fig 6B ) . This is congruent with a recent study which showed that the previously observed VP35-VP30 associations are mediated by RNA and are sensitive to nucleases [15] . Non-segmented negative sense RNA virus nucleoproteins are chaperoned as monomers by the viral phosphoproteins ( filovirus VP35 ) , prior to their incorporation into the large oligomeric , RNA-bound complexes that are the templates used by the phosphoprotein-polymerase protein complex for RNA synthesis . Having shown that VP30 and VP30 mutants interact with NP , we sought to determine which NP oligomeric form binds VP30 . To do this , we separated the NP-binding activities of VP35 into the chaperoning region ( VP35 1–80 ) [16 , 21] and the oligomeric C-terminal region ( VP35 80–340 ) thought to be important for VP35’s activity as a polymerase co-factor , as has been shown for other non-segmented negative-sense RNA viruses [16 , 28–30] . By pulling down on each of these FLAG-tagged VP35 truncations , we are able to select for either the chaperoned , monomeric NP or the oligomeric , RNA-bound NP . In the case of the chaperoned , monomeric NP , the co-transfection of any of our VP30 constructs , including wild-type , greatly diminishes the pulldown of NP with VP35 1–80 . None of these VP30 constructs , nor wild-type VP30 , pulled down with the VP35 1–80 chaperoned NP ( Fig 6C ) . For the oligomeric , RNA-bound NP selected by VP35 80–340 , the only VP30 variant to pull down is the P206R mutant , consistent with the higher affinity of this mutant for NP ( Fig 6D ) . That wild-type VP30 did not pull down with NP in the presence of VP35 80–340 , as it did with NP alone ( Fig 6A ) , is likely reflective of the stringent washes used for the pulldown , the weaker affinity of the wild-type VP30 compared to VP30 P206R , and the less robust isolation of secondary interaction partners by co-immunoprecipitation analysis . To determine the effects of the EBOV VP30 mutants on binding to NP in cells , we used fluorescence analysis of transfected cells ( Fig 7 ) . On its own , the oligomeric NP forms large intracellular inclusions . In virus-infected cells , these inclusions are the sites of viral RNA synthesis [31 , 32] . Though ordinarily presenting diffuse cytoplasmic localization , VP30 and VP35 localize to these inclusions when co-expressed with NP [32 , 33] . Our fluorescence analysis of the VP30 mutants parallels the ITC data where wild-type VP30 , D202R and P206R robustly localize to the NP inclusions , Q229R shows a less robust localization , and E197R , Q203R , W230F and S234R are considerably more diffuse . These data also support the co-immunoprecipitation data above by confirming that VP30 interacts with the oligomeric , RNA-bound NP in the punctate inclusions and not a monomeric form of NP [34] . To complement this binding and interaction data , we also assessed the activity of the VP30 mutants in the EBOV mini-replicon assay , which tests their ability to support RNA synthesis ( Fig 8 ) . Surprisingly , the firefly luciferase activities from the mini-replicon assays display a complex pattern when compared to VP30-NP interaction affinities . The VP30 P206R mutant , which possesses higher affinity for NP , shows less mini-replicon activity . Conversely , the two VP30 mutants with decreased affinities ( D202R and Q229R ) show wild-type or better mini-replicon activity . The four VP30 mutants with affinities too low to be assessed in the ITC experiment showed reduced or background levels of mini-replicon activity . In infected cells , viral replication generates three distinct populations of RNA: genomic viral RNA ( vRNA ) , complementary RNA ( cRNA ) and messenger RNA ( mRNA ) . The negative-sense vRNA acts as a template for the synthesis of positive-sense cRNA and mRNA . The cRNA , in turn , acts as a template for the synthesis of additional vRNA , producing additional copies of the viral RNA genome . One of the proposed functions of VP30 is regulation of the balance of various viral RNA species during viral replication . Mutagenesis of VP30 can alter the ratios of vRNA , cRNA and mRNA in infected and transfected cells [6 , 7] . Thus , we sought to better define the functional consequences of altering VP30-NP interactions with respect to each of these RNA species , in the hope of better understanding the specific processes the VP30-NP interaction mutants would impact . We evaluated the RNAs produced from mini-replicon transfected cells using qRT-PCR ( Fig 9 ) . Although we observe differences in the relative levels of vRNA , cRNA , and mRNA for the different VP30 mutants , the most striking effect of the VP30 mutants is on the overall amounts of all RNAs produced , which correlates with the firefly luciferase activities . These data suggest that the VP30-NP interaction plays a primary role in the regulation of overall RNA synthesis activity by the viral nucleocapsid complexes . Particularly striking are the high levels of viral RNAs seen with the VP30 D202R mutant , especially for the cRNA . However , the amount of mRNA for the other VP30s appears to follow the relative amount of vRNA suggesting that the transcriptional activity is relatively unaltered ( at most ~2-fold ) by changes in the VP30-NP interaction . While we found that VP30 is required for transcription , we were surprised that in the absence of VP30 , neither vRNA nor cRNA is produced , contrary to previous reports that VP30 is not necessary for viral replication [3 , 35] . These previous studies used a mini-replicon system using a different vector backbone for expression of the ebolavirus proteins , different plasmid ratios and different reporter genes . However , these systems are conceptually identical and these minor differences are not expected to account for differing results . Our finding that VP30 is essential for viral RNA transcription and replication is supported by our data from the VP30 W230F and S234R mutants which showed undetectable binding in the ITC analysis , minimal luciferase activity in the mini-replicon system , and RNA profiles similar to those of the–L and–VP30 negative controls .
In this work , we have identified the region of filovirus NPs necessary and sufficient for interaction with VP30 . We broadened our initial screening for the EBOV NP-VP30 interaction to confirm that this interaction also occurs in SUDV and MARV , highlighting the conserved nature of the NP-VP30 interaction and suggesting an important role for this interaction in the filovirus life cycles . Our crystal structures of the EBOV and MARV NP-VP30 complexes have revealed the molecular determinants of these interactions and have allowed us to structurally design mutants specifically targeting the VP30 and NP interfaces . We used our structure-based mutagenesis of the EBOV NP-binding site on VP30 to assess the binding of NP to VP30 using a myriad of assays ( Fig 10 ) . The ITC data indicate a range of mutant VP30 CTD affinities for the NP 600–617 peptide , including some mutants for which NP binding was not apparent . These data are supported by the fluorescence data , which show VP30 localization to NP inclusions for only those VP30 constructs with high affinity binding to NP in ITC . Those VP30 CTD mutants that were not observed to bind NP 600–617 in ITC ( E197R , Q203R , W230F and S234R ) could possibly bind NP if higher protein concentrations were used in the ITC assay . However , further increasing the protein concentrations of the VP30 CTD and NP 600–617 beyond those used in the ITC experiments presented is not feasible . A further limitation to the interpretations of the ITC data is that it uses highly truncated , dimeric VP30 and monomeric NP to accurately measure the affinity of the molecular interaction . Although VP30-NP binding was not observed for some VP30 mutants in ITC experiments , the oligomerization of full-length VP30 and NP is expected to increase the avidity of binding as well as to increase the NP-affinity differences among the VP30 mutants . Analogously , in co-immunoprecipitation experiments , only wild-type VP30 and VP30 P206R could be shown to interact with NP despite measurable interactions between the NP 600–617 peptide and VP30 CTD mutants D202R and Q229R in the ITC assay . This indicates that while assays such as ITC , immunofluorescence and co-immunoprecipitation can reveal protein-protein interactions , they sample a select range of interaction affinities and may exclude biologically relevant interactions . It was previously concluded that the VP30-NP interaction was not necessary for viral transcription based on data showing that a VP30 E197A mutant maintained activity in a mini-replicon assay , but did not pull down NP in co-immunoprecipitation experiments , although the VP30 E197A mutant did partially co-localize with NP inclusions in cells [6 , 11] . The data we present here demonstrate that modifying the affinity of VP30 for NP has differing effects on the outcome of mini-replicon reporter assays , and that these affinities are differentially assessed by assays such as co-immunoprecipitation , ITC , and co-localization in immunofluorescence analysis . A possible explanation for the observed mini-replicon activity for VP30 mutants that are negative for binding to NP , as assessed by assays such as ITC , immunofluorescence or co-immunoprecipitation , is that even a weak or transient interaction of VP30 with NP is capable of promoting RNA synthesis activity . Examples of weak VP30-NP binding permitting RNA-synthesis activity are the previously characterized VP30 E197A mutant [11] and the VP30 Q229R mutant , which though showing a weak affinity in ITC experiments , possesses a wild-type level of RNA synthesis . This is in contrast to the VP30 W230F and S234R mutants that showed neither NP binding in in vitro assays nor activity in mini-replicon assays and may represent complete knockouts for the VP30-NP interaction . Comparison of the mini-replicon assay luciferase activities and qRT-PCR with the mutant VP30-NP interactions as assessed by ITC , co-immunoprecipitation and fluorescence localization of VP30 to NP inclusions reveals an interesting and unexpected trend ( Fig 10 ) . Along the gradient of high-affinity to low-affinity VP30-NP interactions , there appears to be a transition in the effects of VP30-NP affinity on RNA synthesis , belying at least two functions for the VP30-NP interaction in viral RNA synthesis . Beginning on the high-affinity end of the gradient , high-affinity NP-binding VP30 P206R mutant actually decreases viral RNA synthesis activity and modestly weakening the VP30-NP interaction increases RNA synthesis activity . The RNA synthesis activity reaches a maximum for the VP30 D202R mutant and then transitions such that further decreases in the VP30-NP affinity decrease RNA synthesis . The first phase of this affinity gradient suggests that the VP30-NP interaction is inhibitory for RNA synthesis while the second phase is reflective of the essential nature of VP30 for RNA synthesis as we see in our -VP30 control samples . The VP30-NP binding and RNA synthesis activity data suggests multiple functions for the VP30-NP interaction in the filovirus life cycle . A possible mechanism accounting for the two phases of RNA synthesis activity with VP30-NP affinity is that VP30 serves as a positive RNA synthesis co-factor and also stabilizes the nucleocapsid . High-affinity VP30-NP interactions may over-stabilize the nucleocapsid and restrict access of the viral polymerase to the viral genome . Conversely , modestly decreasing the affinity of the VP30-NP interaction would allow greater access of the polymerase to the viral genome . However , further decreases in VP30-NP affinity strongly impact VP30’s function as an RNA synthesis activator , impairing polymerase function . The strength of this hypothesis remains to be tested by future studies of nucleocapsid assembly , stability and polymerase function . Our data show that EBOV , SUDV and MARV VP30s recognize a short peptide in the C-terminal region of their respective NPs . In addition , we have described the molecular interaction of these components with crystal structures and shown this interaction to be essential for viral RNA synthesis . Although the VP30-NP interaction is essential for RNA synthesis , it appears that this interaction serves to reduce the RNA synthesis activity and that to an extent , modestly destabilizing this interaction gives a boost to viral RNA synthesis ( Figs 8 and 9 ) . This suggests that the VP30-NP interaction tunes the overall RNA synthesis activity to a level ideal for viral infection , rather than maximal RNA synthesis . Because the VP30-NP interaction is essential and highly conserved and because the strength of this interaction modulates RNA synthesis activity , therapeutic drugs either blocking or stabilizing this interaction may be efficacious in treating filovirus infections . Escape mutations arising in VP30 or NP in response to such therapeutic drugs are predicted to have disregulated RNA synthesis activity possibly resulting in attenuated viruses .
Full length EBOV NP was detected by Western blotting using human antibody KZ51 ( gift from Dennis Burton ) . EBOV NP truncations and mutants were detected by Western blotting using rabbit anti-2A peptide antibody ABS31 ( Millipore ) . EBOV VP30 and VP30 mutants for the mini-replicon experiments were detected by Western blotting with a polyclonal rabbit anti-VP30 antibody ( gift from Yoshihiro Kawaoka and Pete Halfmann ) . EBOV VP35 in the mini-replicon experiments was detected by Western blotting using a mouse anti-VP35 antibody ( gift from Victor Volchkov ) . In co-immunoprecipitation experiments , HA-tagged VP30 proteins were detected by Western blotting using a rabbit anti-HA antibody ( Rockland Immunochemicals ) and FLAG-tagged VP35 proteins were detected using a rabbit anti-FLAG antibody ( Cell Signaling Technology ) . Secondary antibodies used for Western blotting were goat anti-mouse or -rabbit or -human Fc HRP conjugates ( Thermo Fisher ) . For immunofluorescence , HA-tagged VP30 was detected with 16B12 ( BioLegend ) and a goat anti-mouse Alexa 647 conjugate ( Thermo Fisher ) . NP and VP30 protein constructs for bacterial expression were cloned into pET46 ( Novagen ) , which contains an N-terminal hexahistidine tag and enterokinase cleavage site . For producing biotinylated NP peptides , pET46 was modified to additionally include an N-terminal AviTag . Bacterially expressed proteins were produced in Rosetta2 pLysS E . coli ( Novagen ) . 1 L cultures were grown at 37°C to an OD600 of 0 . 4 and induced with isopropyl-β-D-1-thiogalactoside ( IPTG , 0 . 5 mM final concentration ) . The temperature was reduced to 25°C and expression was carried out overnight . Bacteria were harvested by centrifugation and resuspended in Ni-NTA binding/wash buffer ( 50 mM HEPES pH 7 . 4 , 300 mM NaCl , 30 mM imidazole , 2mM BME ) . Resuspended cells were lysed using a Microfluidizer M110-P ( Microfluidics ) . Lysates were cleared by centrifugation at 25 , 000×g for 30 minutes and then filtered through a 0 . 22 μm filter . 2 mL of Ni-NTA Agarose ( Qiagen ) was added to the lysates and allowed to incubate for 30 minutes . Agarose beads were collected and washed twice with 20 mL of Ni-NTA binding buffer in a gravity column format . Bound proteins were eluted with the same buffer containing 300 mM imidazole . Proteins were concentrated by ultrafiltration ( Amicon , Millipore ) prior to size exclusion chromatography ( Superdex 200 , GE Life Sciences ) in 50 mM HEPES pH 7 . 4 , 300 mM NaCl , 5mM BME . For the expression of biotinylated NP peptides , NP-expressing plasmids were co-transformed with a plasmid carrying E . coli biotin ligase and 1 L cultures were supplemented with 13 mg of D-biotin at the time of expression induction . These peptides were purified identically to other expressed proteins with Superdex 75 used for size exclusion chromatography . For biolayer interferometry ( BLI ) screening of EBOV NP-VP30 protein interactions , all proteins were prepared in 1X BLI buffer ( 50 mM HEPES pH 7 . 4 , 300 mM NaCl , 5 mM BME , 10μg/mL bovine serum albumin and 0 . 002% Tween-20 ) . All BLI experiments were performed using an Octet Red ( Forte Bio ) and streptavidin coated biosensors . Biosensors were equilibrated in 1X BLI buffer for 5 minutes prior to the start of the experiment . Biosensors were successively dipped into wells containing 1X BLI buffer ( 60 s ) , 50 μg/mL biotinylated EBOV NP peptides ( 120 s ) , 1X BLI buffer ( 120 s ) , 10 μM VP30 CTD ( 180 s ) and 1X BLI buffer ( 180 s ) . Because EBOV VP30 CTD is dimeric , complicating quantitative analysis , these experimental data were used qualitatively . Expressed and purified proteins were dialyzed overnight into 25 mM HEPES pH 7 . 4 , 300 mM NaCl , 5 mM BME . EBOV NP 360–739 constructs and NP 600–617 fused to T4-lysozyme were concentrated to 120–140 μM as determined by UV absorbance at 280 nm . VP30 CTD constructs were concentrated to 1 , 200–1 , 600 μM . NP constructs were loaded into the calorimeter cell and VP30 constructs were loaded into the titration syringe . All experiments were carried out using an Auto-iTC200 ( Malvern ) operating at 25°C and 1000 rpm unless otherwise noted . ITC data was processed with Origin software . Reported dissociation constants and errors are the average and standard deviation of three replicates . EBOV mini-genome assays were performed similar those previously described [20] . Briefly , HEK 293T cells ( gift from Dr . Dennis Burton ) were grown to 80–90% confluency in DMEM , 5% FBS , 50 U/mL penicillin and 50 μg/mL streptomycin in 12-well format . Cells were transfected with 2 μg of EBOV L-pCAGGS , 0 . 75 μg VP35-2A-NP-pCAGGS , 0 . 25 μg VP30-pCAGGS , 0 . 5 μg T7 Polymerase-pCAGGS , 0 . 5 μg 3E5E-ffLuc and 0 . 1 μg Renilla luciferase-pCMV ( Promega ) mixed with TransIT-LT1 ( Mirus Bio ) in a 3:1 ratio ( μL/total DNA μg ) . Transfected cells were harvested after 48 hours . For luciferase activities , cells were washed with phosphate-buffered saline and lysed with 1X Passive Lysis Buffer ( Promega ) . Lysates were flash-frozen and thawed , and then cleared by centrifugation prior to measuring luciferase activities using the Dual-Luciferase Assay Reporter System ( Promega ) read on a Spark 10M ( Tecan ) using injectors . Luciferase activity for each transfection is reported relative to wild-type control and normalized for differences in transfection efficiency using the Renilla luciferase activity as an internal control . Transfections were performed at least three times . For VP30 co-immunoprecipitation of NP or VP35 , 0 . 5 μg of N-terminally HA-tagged VP30-pDisplay or mutants were cotransfected in HEK 293T cells with 0 . 5 μg of either full-length NP or full-length N-terminally FLAG-tagged VP35 mixed with 1mg/mL PEI MAX ( Polysciences ) in a 3:1 ( μL/total DNA μg ) in 12-well plates . For VP30 co-immunoprecipitation of NP and truncated VP35 , 0 . 5 μg of HA-tagged VP30-pDisplay was combined with 0 . 25 μg of NP-pDisplay and 0 . 25 μg of either C-terminally tagged VP35 1-80-pDisplay or N-terminally tagged VP35 80-340-pDisplay . Transfected cells were harvested after 48 hours . Cells were washed with 100 μL TBS and lysed in Cell Lysis Buffer ( 50 mM TrisCl pH 8 . 0 , 300 mM NaCl , 2 mM EDTA , 10% glycerol , 1% Igepal CA-630 , 2 mM BME , 250 U/mL benzonase nuclease ( Novagen ) ) for 10 minutes . Lysates were cleared by centrifugation . Lysate supernatants were mixed with 25 μL anti-HA agarose beads ( 3F10 , Roche ) or 25 μL anti-FLAG agarose beads ( M2 , Sigma ) and incubated at 4°C for 30 minutes with agitation . Beads were then washed three times with 800 μL of Cell Lysis Buffer ( without benzonase ) and then mixed with 30 μL of SDS-PAGE sample loading buffer ( 6% SDS , 30% glycerol , 180 mM TrisCl pH 6 . 8 , 0 . 0125% bromophenol blue ) and boiled for 1 minute before loading gels for Western blotting analysis . HEK 293T cells were grown on poly L-lysine coated coverslips . Cells were transfected with 0 . 25 μg of N-terminally HA-tagged VP30-pDisplay and 0 . 25 μg of eGFP-NP-pDisplay with TransIT-LT1 ( MirusBio ) in a 3:1 ratio ( μL/μg DNA ) . 24 hours post-transfection , coverslips were washed in phosphate buffered saline ( PBS ) and cells were fixed with 2% paraformaldehyde for 20 minutes . Coverslips were washed three times with PBS and then quenched and permeabilized ( 0 . 5% Triton X-100 , 20 mM glycine in phosphate buffered saline ) for 20 minutes . Coverslips were washed three times with PBS , blocked ( 2% bovine serum albumin ( BSA ) in PBS ) for 30 minutes and then incubated with a 1/500 dilution of anti-HA antibody in 1% BSA and PBS for 45 minutes . Coverslips were washed three times in PBS and then incubated with a 1/2000 dilution of secondary antibody , goat anti-mouse IgG conjugated to Alexa Fluor 647 in 1% BSA and PBS for 45 minutes . Coverslips were washed with three times with PBS and then stained with 1 μg/mL Hoechst 33342 in PBS for 5 minutes and then washed twice with PBS and once with water . Coverslips were mounted on glass slides using Prolong Gold ( Invitrogen ) . Slides were imaged on a Nikon TE2000 epifluorescence microscope and images were collected and processed with Metamorph software ( Molecular Devices ) . In addition to measuring luciferase activities of mini-genome transfected cells , select reactions were repeated and assessed using qRT-PCR to examine relative amounts of vRNA , cRNA and mRNA . Cells were transfected as above . After 48 hours , RNA was harvested from the cells using the RNeasy Plus Mini Kit ( Qiagen ) . cDNA was generated using 500 ng of RNA and the SuperScript III First-Strand Synthesis System ( Invitrogen ) using primers specific for vRNA ( GACAAATTGCTCGGAATCACAAAATTCC ) , cRNA ( CACGCAGGGAGAGAGGCTAAATATAG ) or mRNA ( poly ( dT20 ) ) . qPCR was performed using primers directed to amplify firefly luciferase ( GCTATTCTGATTACACCCGAGG , TCCTCTGACACATAATTCGCC ) , GAPD ( AAGGTGAAGGTCGGAGTCAA , AATGAAGGGGTCATTGATGG ) or RPLP0 ( GCGACCTGGAAGTCCAACTA , ATCTGCTTGGAGCCCACAT ) . The reactions were run using 2X SYBR Select Master Mix ( Applied Biosystems ) and a CFX384 Touch Real-Time PCR Detection System ( Bio-Rad ) . Data was analyzed using Bio-Rad CFX Manager using the GAPD and RPLP0 as reference controls . A serial dilution of 3E5E-ffLuc plasmid was included to assess the efficiency of amplification of the firefly luciferase cDNAs . qPCR reactions were repeated four times . Due to the low affinity of the EBOV NP 600–617 for VP30 CTD , these proteins were cloned as a fusion protein for structural studies using X-ray crystallography . EBOV NP 600–627 was fused to the N-terminus of VP30 CTD and purified as described above . Crystals were grown using 0 . 3 μL of 20 mg/mL protein mixed with 0 . 3 μL of 2 . 2 M ammonium sulfate , 100 mM sodium acetate pH 4 . 9 in a sitting drop , vapor diffusion setup . Crystals were cryo-protected in 3 . 15 M ammonium sulfate , 50 mM sodium acetate pH 5 . 0 and 5% glycerol prior to cryo-cooling the crystals in liquid nitrogen . MARV NP 552-579-VP30 146–281 fusion protein crystallized in 20% isopropanol , 18% PEG 4000 , 0 . 1 M sodium citrate pH 5 . 6 using 0 . 2 μL 24 . 7 mg/mL protein mixed with 0 . 2 μL mother liquor . MARV crystals were cryoprotected with 20% 2-methyl 2 , 4-pentanediol in mother liquor prior to cryo-cooling in liquid nitrogen . Diffraction data were collected at the National Institute of General Medical Sciences and National Cancer Institute Structural Biology Facility ( GM/CA ) at the Advanced Photon Source , Argonne National Lab at beamline 23ID-D . Diffraction data was processed with XDS [36] and merged with AIMLESS [37] . The EBOV NP-VP30 structure was determined using molecular replacement as implemented in Phaser [38] and 2I8B . pdb [11] as a search model , while for the MARV NP-VP30 molecular replacement , the 5DVW . pdb [26] was used as a search model . The models were rebuilt in Coot [39] and refined using Phenix [40] and Refmac [41] . Analysis of the protein interaction surface was done in PISA [42] . Coordinates and structure factors for the EBOV NP-VP30 complex , 5T3T . pdb , and the MARV NP-VP30 complex , 5T3W . pdb , are deposited in the Protein Data Bank ( www . rcsb . org ) . GenBank accession numbers for proteins used in this study: EBOV NP: AAD14590 . 1 , EBOV VP30: AAD14587 . 1 , SUDV NP: AGB56675 . 1 , SUDV VP30: YP_138525 . 1 , MARV NP: YP_001531153 . 1 , MARV VP30: YP_001531157 . 1 , EBOV VP35: NP_066244 . 1 , EBOV L: NP_066251 . 1 ( http://www . ncbi . nlm . nih . gov/ ) .
|
Filoviruses use a system of proteins and RNA to regulate viral RNA genome transcription and replication . Here , we have determined crystal structures and the biological functions of the protein complex formed by the filovirus transcriptional activator , VP30 , and the core component of the nucleocapsid machinery , NP . The complex of these two essential players represses Ebola virus RNA synthesis and may have played a role in the evolution of filoviruses to tune viral RNA synthesis activity to a level ideal for infection . This interaction is conserved across the filoviruses and may provide an opportunity for therapeutic development .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"luciferase",
"nucleic",
"acid",
"synthesis",
"medicine",
"and",
"health",
"sciences",
"crystal",
"structure",
"pathology",
"and",
"laboratory",
"medicine",
"protein",
"interactions",
"enzymes",
"pathogens",
"condensed",
"matter",
"physics",
"enzymology",
"microbiology",
"negative-sense",
"rna",
"viruses",
"viral",
"structure",
"nucleocapsids",
"viruses",
"filoviruses",
"rna",
"viruses",
"immunoprecipitation",
"co-immunoprecipitation",
"rna",
"synthesis",
"crystallography",
"chemical",
"synthesis",
"research",
"and",
"analysis",
"methods",
"solid",
"state",
"physics",
"proteins",
"medical",
"microbiology",
"oxidoreductases",
"microbial",
"pathogens",
"precipitation",
"techniques",
"biosynthetic",
"techniques",
"physics",
"biochemistry",
"rna",
"ebola",
"virus",
"nucleic",
"acids",
"virology",
"viral",
"pathogens",
"biology",
"and",
"life",
"sciences",
"physical",
"sciences",
"hemorrhagic",
"fever",
"viruses",
"organisms"
] |
2016
|
The Ebola Virus VP30-NP Interaction Is a Regulator of Viral RNA Synthesis
|
Larvae of the tapeworm E . multilocularis cause alveolar echinococcosis ( AE ) , one of the most lethal helminthic infections in humans . A population of stem cell-like cells , the germinative cells , is considered to drive the larval growth and development within the host . The molecular mechanisms controlling the behavior of germinative cells are largely unknown . Using in vitro cultivation systems we show here that the EGFR/ERK signaling in the parasite can promote germinative cell proliferation in response to addition of human EGF , resulting in stimulated growth and development of the metacestode larvae . Inhibition of the signaling by either the EGFR inhibitors CI-1033 and BIBW2992 or the MEK/ERK inhibitor U0126 impairs germinative cell proliferation and larval growth . These data demonstrate the contribution of EGF-mediated EGFR/ERK signaling to the regulation of germinative cells in E . multilocularis , and suggest the EGFR/ERK signaling as a potential therapeutic target for AE and perhaps other human cestodiasis .
A population of pluripotent adult somatic stem cells , well known as “neoblasts” , has been extensively characterized and documented for free-living platyhelminthes ( flatworms ) [1–5] . Neoblasts represent the only proliferative cell population , responsible for cell renewal during homeostasis , development and regeneration [6–11] . Like in its free-living relatives , neoblast-like stem cells and similar cell renewal mechanisms also exist in the other two main groups of flatworms , trematodes ( flukes ) and cestodes ( tapeworms ) , both living a parasitic life [3 , 12–15] . In cestodes , a population of undifferentiated cells , called “germinative cells” , is the only source for cell proliferation . These germinative cells are totipotent and are thought to drive growth and development throughout the life cycle of cestodes [12 , 15–16] . The larval stage of the cestode Echinococcus multilocularis is the causative agent of alveolar echinococcosis ( AE ) , one of the most lethal human helminthiasis [17] . An infection is initiated when the intermediate host ( rodents , humans ) ingests infective eggs produced by adult tapeworms . The oncosphere hatches from the egg and then develops in the liver into cyst-like metacestode vesicles , which grow like tumors and infiltrate host tissue , forming new vesicles and even metastasizes . The metacestode vesicles bud giving rise to brood capsules , which in turn generate protoscoleces by asexual multiplication . Protoscoleces can either mature into adult tapeworms if ingested by the definitive host ( canids ) or develop into metacestode vesicles when distributed in the intermediate host . This unique proliferation potential of E . multilocularis metacestode larvae is believed to be based upon the germinative cells , which are totipotent and have the ability for extensive self-renewal [15] . It has been well documented that the maintenance of pluripotency and self-renewal capacity of stem cells requires a continuous input from cell-extrinsic signals [18–19] . The extrinsic factors initiate various intrinsic signaling cascades which in turn maintain stem cells and regulate their functions . Signaling axes including LIF/gp130/STAT3 , BMPs/BMPRs/Smads , Wnt/Frizzled/β-catenin , PI3K/AKT , and FGF/FGFR have been extensively evidenced to participate in controlling the survival , self-renewal , and differentiation of stem cells [19–20] . Increasing evidence has shown that the epidermal growth factor receptor ( EGFR ) -dependent signaling pathways also play important roles in the maintenance and regulation of stem cells [21–25] . The metacestode larvae of E . multilocularis grow and proliferate in close contact with the intermediate host’s tissues , mainly within the liver . The microenvironment for metacestode development involves a number of host-derived hormones and cytokines , such as insulin , bone morphogenetic protein ( BMP ) , fibroblast growth factor ( FGF ) and epidermal growth factor ( EGF ) [26] . These host-derived factors are thought to bind to parasite receptors and in turn influence the parasite’s growth and development through exhibiting their impacts on the relevant evolutionarily conserved signaling systems within the parasite [26–27] . A recent study has evidenced that host insulin activates E . multilocularis PI3K/AKT signaling pathways and stimulates germinative cell proliferation and larval development [28] . In addition , in vitro cultivation of E . multilocularis larvae and primary cells requires continuous presence of host cells as the feeder cells ( like stem cell cultivation ) or host cell-conditioned medium which contains host-derived growth factors [29] . Together , lines of evidence offer compelling clues that the conserved signaling pathways in E . multilocularis could respond to host factors and may regulate germinative cells which are fundamental for the larval growth and development of the parasite [30] . It has been shown that the Ras/Raf/MEK/ERK signaling in E . multilocularis is activated in response to a host-derived EGF signal , which is most probably mediated by the parasite’s EGFR-like kinase [31–33] . Using in vitro cultivation systems we show here that E . multilocularis EGFR/ERK signaling pathway is activated upon addition of human EGF and promotes germinative cell proliferation during the parasite’s larval growth and development .
All animal experiments were conducted in strict accordance with China regulations on the protection of experimental animals ( Regulations for the Administration of Affairs Concerning Experimental Animals , version from July-18-2013 ) and specifically approved by the Institutional Animal Care and Use Committee of Xiamen University ( Permit Number: 2013–0053 ) . The parasite isolate used in this study was obtained from Hulunbeier Pasture of Inner Mongolia of China [34] and maintained by in vivo propagation of the parasite material in mice ( supplied by Xiamen University Laboratory Animals Center , XMULAC ) . In vitro cultivation of metacestode vesicles was performed using host cell conditioned medium according to a previously established protocol [35] unless otherwise indicated . For the growth assay , vesicles ( diameter < 1mm ) were manually picked up and cultured in 24-well cell culture plates supplemented with different media as indicated in the text . 100 ng/mL recombinant human EGF ( PeproTech , Rocky Hill , NJ ) was used for all experiments unless otherwise indicated . Parasite growth was determined by the measurement of vesicle’s diameter under inverted microscope weekly . Each group contains at least 3 replicates and more than 150 vesicles in total for each group were analyzed . Two-three independent experiments were performed . For the treatment of inhibitors , the EGFR inhibitors CI-1033 and BIBW2992 or the MEK inhibitor U0126 ( Selleck Chemicals , Houston , TX ) was added into the culture medium at a final concentration as indicated . All experiments were performed with exchange of the medium containing the same ingredients every three days . Protoscoleces were collected from parasite material and in vitro cultured in conditioned medium . The vesicle formation process , in which protoscoleces dilate and vacuolate , were examined after 18 days of culture . 40 mM of hydroxyurea was used to treat metacestode vesicles as described before [15] . Vesicles were incubated with BrdU for two days and chromosomal DNA was isolated for BrdU incorporation assay with the colorimetric BrdU ELISA kit ( Roche , Mannheim , Germany ) as described before [28] . Vesicles were incubated with 50 μM of EdU for 4 hours and whole-mount prepared according to Cheng et al . [36] . Click-iT-EdU Alexa Fluor 555 Imaging Kit ( Life Technologies , Shanghai , China ) was used for detection of EdU . For the EdU-BrdU dual labeling , vesicles were incubated with 10 μM EdU for 4 hours , washed and then cultured with no labeling for 44 hours . 10 μM BrdU was next used for continuous labeling for another 24 hours . Vesicles were fixed at the end of the labeling period and whole-mount prepared ( see also S2C Fig ) . After a 45-minute 2N HCl treatment , immunofluorescence was performed for BrdU detection using the anti-BrdU antibody ( clone MoBU , Life Technologies ) followed by EdU detection . DNA was counterstained with 4’ , 6-diamidino-2-phenylindole ( DAPI ) ( Sigma , St . Louis , MO ) for all labeling experiments . For the inhibitor experiments during the recovery from hydroxyurea treatment , vesicles were allowed to recover in the conditioned medium supplemented with EGF . 10 μM CI-1033 or 20 μM U0126 was then added into the medium immediately after the initial EdU pulse . Germinative cell proliferation was analyzed by EdU-BrdU dual labeling after 4 days of recovery . For the quantification of EdU+ or BrdU+ cells , at least 12 random microscopic fields from 4–6 vesicles were captured and the positive cells were manually counted . 3–5 labeling experiments were performed and analyzed for each control and treatment group . Total RNA was extracted from in vitro-cultivated protoscoleces or metacestode vesicles , treated with RNase-free DNase and reverse transcribed into cDNA . cDNAs were processed for RT-PCR analysis using the primers: EmER-qF2 ( 5’-GCG AAT GTA AGC ATT TCA AGT CA-3’ ) and EmER-qR2 ( 5’-TTC ACA AAG TAG CAG AAA GCA CAT-3’ ) for Emer; 617300-qF ( 5’-GCC GCA TCT ATG GAC ACGC-3’ ) and 617300-qR ( 5’-AGT CAT CTT GTG GGA GGA ATCG-3’ ) for Emuj_000617300; 969600-qF ( 5’-CTC TGG GGT GTC TGC TGT CC-3’ ) and 969600-qR ( 5’-TCC CAC AGA GTC ACA CCG TAGG-3’ ) for Emuj_000969600 . Expression of the parasite EGF receptor EmER in Xenopus oocytes was performed according to [37] . Briefly , the full-length coding sequence of EmER was cloned into the Xenopus oocyte expression vector pXT7-flag ( a gift from Dr . Li Guang in Xiamen University ) . Linearized plasmids were then used as the templates for capped mRNA ( cRNA ) synthesis using the T7 mMessage mMachine Kit ( Ambion ) . Oocytes were obtained from Xenopus laevis ( supported by Stem Cell Bank , Chinese Academy of Sciences ) and then microinjected with EmER cRNA or water ( noninjected control ) . Membrane proteins were extracted from 30 oocytes after 48 h of injection and immunoprecipitated by the anti-flag monoclonal antibody ( Sigma ) and analyzed by western blot . For EGF and CI-1033 treatment , oocytes that had been expressing EmER for 48 h were incubated with 10 μM of CI-1033 or DMSO for 4 h followed by 20 minutes of EGF stimulation . Membrane extracts were immunoprecipitated and analyzed by western blot using the anti-flag or anti-phospho-tyrosine ( CST , Beverly , MA ) monoclonal antibodies . Induction of GVBD ( germinal vesicle breakdown ) in EmER-expressing Xenopus oocytes was performed according to [37] . Oocytes that had been expressing EmER for 48 h were pretreated with 10 μM of CI-1033 or DMSO for 4 h and then incubated with EGF . GVBD was monitored after 16 h of EGF incubation . As a positive control for GVBD , oocytes were stimulated with progesterone ( PG ) , the natural inducer . 20–30 oocytes were used for each group and three independent experiments were performed . Lysates of in vitro-cultivated protoscoleces or metacestode vesicles were produced , separated on 12% acrylamide gels and transferred to PVDF membranes . Detection of E . multilocularis ERK and phosphorylated ERK was performed according to Spiliotis et al . [32] using the polyclonal rabbit anti-ERK1/2 ( Stressgen , Victoria , Canada ) and anti-ERK1/2 [pTpY185/187] ( Life Technologies ) antibodies , respectively . The anti-rabbit IgG antibody conjugated with horseradish peroxidase ( Theromo Scientific , Shanghai , China ) was used as a secondary antibody . For all western blot experiments , detection of E . multilocularis actin was performed using the polyclonal anti-β-actin antibody ( CST ) as loading controls . Immunofluorescence was performed using the whole-mount prepared metacestode vesicles as described before [36] . For Histone H3 detection , the anti-phospho-Histone H3 antibody ( Ser10 , 1:200 ) ( CST ) was used . For all immunofluorescence experiments , an Alexa 488-conjugated second antibody ( Life Technologies ) was used and DNA was counterstained with DAPI . Data of three or more experimental repeats are shown as mean ± SD as indicated in the respective figure legend unless otherwise indicated . The mean values of the data from the experimental groups were compared by performing two-tailed Student’s t-test and the P values were indicated as those: *P < 0 . 05 , **P < 0 . 01 , and ***P < 0 . 001 .
To examine the impacts of EGF on the larval growth of E . multilocularis , metacestode vesicles were incubated in host cell-conditioned medium supplemented with recombinant human EGF . We found that 10 ng/mL or higher concentrations of EGF can greatly promote vesicle’s growth ( S1 Fig ) . We then used 100 ng/mL of EGF which showed the most significant effect on parasite’s growth in vitro for further studies . The results show that addition of EGF can stimulate the growth of metacestode vesicles ( Fig 1A ) . Similar results were observed after EGF was added into the Dulbecco’s modified eagle medium ( DMEM ) containing 10% serum ( Fig 1B ) . We also found that EGF greatly promoted the vesicle formation process , in which the protoscoleces dilated and vacuolated ( Fig 1C ) . These results illustrate that the larval growth and development of in vitro-cultivated E . multilocularis larvae could be stimulated by exogenously added EGF , which is probably mediated by an EGFR-dependent signaling in the parasite . Considering that either serum or conditioned medium contains complex ingredients , we then incubated vesicles in the serum-free DMEM only supplemented with EGF . Addition of EGF could not obviously stimulate the growth of vesicles , however , it remarkably promoted their survival ( Fig 1D ) . Although metacestode vesicles could not survive for long in this situation , the method excludes the influence of other host factors in serum and suggests that the parasite is responsive to host EGF stimulation . Further experiments in this study were all performed using host cell-conditioned medium unless otherwise indicated . Given that the germinative cells , a population of stem cell-like cells , drive larval growth and development of E . multilocularis , we then investigated the impacts of EGF on the germinative cells . Vesicles were incubated with 5-bromo-2’-deoxyuridine ( BrdU ) , an analogue of thymidine used for studying cell proliferation by detecting its incorporation into the newly synthesized DNA of replicating cells . The result shows that addition of EGF greatly stimulated the BrdU incorporation in the vesicles ( Fig 2A ) . Since germinative cells are the only cells capable of proliferation in metacestode vesicles , this result suggests that EGF promotes germinative cell proliferation . A dual labeling method through sequential pulses of 5-ethylnyl-2’-deoxyuridine ( EdU ) and BrdU , which has been utilized to verify the self-renewal capacity of adult somatic stem cells in the human blood fluke Schistosoma mansoni [14] , was further applied to determine the effects of EGF on germinative cells . EdU is a newly found analogue of thymidine [38] , which has been shown to be incorporated by proliferating cells of E . multilocularis [15] . Under normal in vitro cultivation conditions , we found that most of the dividing germinative cells in the vesicles could incorporate EdU before a chase period of 44 h , which was further used for EdU-BrdU dual labeling experiments ( S2A Fig ) . The results show that about 48% of cells that initially incorporate EdU are BrdU+ 3 days after an initial EdU pulse ( 749 EdU+BrdU+ / 1554 EdU+ nuclei , 4 independent labeling experiments ) ( Fig 2B ) , indicating that these germinative cells divide and produce proliferation-competent daughter cells that initially incorporate EdU can incorporate BrdU during the second replication . It has been shown that depletion of the germinative cells in E . multilocularis vesicles could be achieved through hydroxyurea treatment for longer periods of time ( e . g . seven days ) and that the germinative cells undergo clonal expansion like stem cells after removal of hydroxyurea [15] . We also performed similar experiments and found that the EdU+BrdU+ cells are highly presented in the clonally growing germinative cells ( S3 Fig ) , suggesting that the EdU+BrdU+ cells are extensively proliferating germinative cells and a part of them might be undergoing self-renewing divisions . To further investigate the effect of EGF on germinative cells , we treated vesicles with hydroxyurea to eliminate most germinative cells [15] . After removal of hydroxyurea , germinative cell proliferation/self-renewal was allowed for recovery in the medium with addition of EGF or not for 4 days , and the EdU-BrdU sequential pulses began on the second day of the recovery ( see also S2B and S2C Fig ) . At the end of the dual labeling period , the results show that addition of EGF induced significantly more numbers of both EdU+ and EdU+BrdU+ cells in the vesicles compared to the controls ( Fig 2C ) . We found that the proportion of EdU+BrdU+ cells with respect to the number of EdU+ cells was also greatly increased ( 25 . 6% and 50 . 5% for the control and EGF-treated groups , respectively , statistical significance P = 0 . 00101 ) , suggesting increased continuous proliferation , and possibly promoted self-renewal of the germinative cells upon EGF stimulation . Together , these results support the findings that addition of EGF can promote the proliferation of germinative cells ( Fig 2A ) , which subsequently drives the larval growth and development of E . multilocularis . These results also suggest that an EGFR-dependent signaling in the parasite may be involved in regulating germinative cell proliferation upon EGF stimulation . E . multilocularis possesses an EGFR-like kinase ( EmER ) which is suggested to interact with host EGF [31 , 33] . Besides EmER , we also found two additional E . multilocularis EGF receptor members ( Emuj_000617300 and Emuj_000969600 ) by analyzing E . multilocularis genome sequence ( http://www . genedb . org/Homepage/Emultilocularis ) and fully cloned and sequenced the respective genes . Comparisons of the putative protein sequences reveal that these EGF receptor members exhibit significant homologies to human EGFR and the EGFR homologue of the closely related schistosome S . mansoni , especially in the tyrosine kinase domains ( S4A–S4D Fig ) . The results of mRNA expression analysis show that these EGF receptor homologues are constitutively present in E . multilocularis metacestode vesicles and protoscoleces ( S4E Fig ) . To investigate whether the E . multilocularis EGF receptor ( s ) respond to EGF stimulation or not , we utilized the Xenopus oocyte expression system , which is a powerful tool for receptor tyrosine kinase research and has been successfully used for studying the EGF receptor ( SER ) in S . mansoni [37] . The results show that the parasite EGFR EmER could be efficiently expressed in Xenopus oocytes with a molecular weight approximately 200 kDa ( Fig 3A ) , and that addition of EGF resulted in the activation of EmER by detection of phosphorylated tyrosine ( Fig 3B ) . It has previously been shown that host EGF could induce germinal-vesicle breakdown ( GVBD ) in S . mansoni SER-expressing oocytes [37] . Similar results were also observed in the EmER-expressing oocytes ( Fig 3C ) , suggesting that addition of exogenous EGF could activate the parasite EGFR in the oocytes and induce GVBD . Using the oocyte system , we also found that CI-1033 ( canertinib ) , an irreversible inhibitor for human EGF receptors [39] , could effectively inhibit the EGF-induced activation of EmER and GVBD ( Fig 3B and 3C ) . We wondered if the impaired EGFR activation would impact E . multilocularis germinative cell behaviors . To this end , we treated vesicles with CI-1033 . The results show that CI-1033 significantly reduced the number of EdU+ cells in the vesicles ( Fig 4A ) . Another EGFR inhibitor , BIBW2992 ( afatinib ) [40] , also exhibited a similar effect to CI-1033 on the germinative cells ( S5A Fig ) . In the EdU-BrdU dual labeling experiments , vesicles were allowed to recover from the hydroxyurea treatment with addition of EGF , and CI-1033 was administrated to the vesicles immediately after the initial EdU pulse . At the end of the labeling period , only 1 . 6‰ of total cells were EdU+ ( 11 EdU+ / 6737 DAPI nuclei ) , and the EdU+BrdU+ cells were hardly detected ( Fig 4B ) . There is a strong accumulation of EdU+ cells in numerous aggregates in some developing vesicles , which indicates that the active proliferation and extensive self-renewal of germinative cells may drive the development of brood capsule and protoscolex in the vesicles [15] . Our studies show that CI-1033 can also abolish the accumulation of EdU+ cells in these cell aggregates ( Fig 4C ) . Further investigations showed that CI-1033 and BIBW2992 can significantly inhibit the larval growth and development upon EGF stimulation ( Fig 4D and S5B Fig ) . These results suggest that an EGFR-dependent signaling in the parasite is required for the promoted germinative cell proliferation and larval growth , and that the signaling could probably be impaired by the EGFR inhibitors initially designed against human EGF receptors . In E . multilocularis , structural and functional homologues to mammalian MAP kinase cascade molecules , such as RAF , MEK and ERK , have been identified and characterized [27] . Previous study has shown that host EGF could induce E . multilocularis ERK activation [32] . Our data show that the basal level of ERK phosphorylation in the vesicles was down-regulated following CI-1033 treatment in a time-dependent manner ( Fig 5A ) . We also found that CI-1033 can inhibit the phosphorylation of ERK induced by EGF ( Fig 5B ) . These results suggest that the EGFR inhibitors could impair the activations of parasite’s EGFR and ERK . We then treated vesicles with U0126 , a MEK inhibitor which effectively suppressed both of the basal and the addition of EGF-induced ERK phosphorylations in the parasite ( Fig 5C and 5D ) . Along with the inhibition of MEK/ERK activity , a remarkable decrease in the number of EdU+ germinative cells was observed in the vesicles ( Fig 5E ) . EdU-BrdU dual labeling experiments also indicate that U0126 has a comparable inhibition effect to CI-1033 on the EGF-promoted germinative cell proliferation ( 2 . 9‰ of the total cells were EdU+ and the EdU+BrdU+ cells were hardly detected ) ( Fig 5F ) . Our further investigation shows that U0126 can significantly inhibit the EGF-stimulated vesicle growth ( Fig 5G ) . Taken together , these results suggest that the EGF-mediated EGFR/MEK/ERK signaling contributes to germinative cell proliferation during E . multilocularis larval growth and development .
Throughout the complex life cycle of E . multilocularis , the parasite always keeps a population of stem cell-like cells , the germinative cells , which are considered as one of the fundamental underpinnings of its growth and development in the host [15] . Stem cell maintenance and functions are strictly controlled by signals from the local tissue microenvironments known as “niches” , which have been widely characterized and elucidated for mammals and invertebrate model animals [18] . Proliferation and differentiation of neoblasts , the stem cells in the free-living flatworm planarian , are also regulated by signals from the surrounding cells [11] . However , the situation is somewhat different for the host liver tissue-dwelling metacestode larvae of E . multilocularis . Due to the intimate parasite-host contact , the parasite is believed to be able to sense signals derived not only from its own tissue but also from the host-derived hormones and cytokines , as these signaling receptors and downstream signaling cascades are evolutionarily conserved between the mammalian hosts and E . multilocularis [27] . Thus it is tempting to suggest that E . multilocularis germinative cells should be regulated by the host-derived factors [30] . In the present study , we show that germinative cell proliferation of the in vitro-cultivated E . multilocularis larvae is promoted upon the addition of human EGF , which in turn drives vesicle growth and vesicle formation from protoscolex ( Figs 1 and 2 ) . These results suggest that the germinative cells are regulated by the signaling pathways within the parasite that could sense the host-derived EGF signal . A physiologically relevant concentration of EGF ( 1 ng/mL ) showed a modest effect on the growth of in vitro-cultivated vesicles , while 10 ng/mL or higher concentrations of EGF greatly stimulated the growth ( S1 Fig ) . We then used 100 ng/mL of EGF , which exhibited the most significant effect on parasite’s growth , for further in vitro studies . This concentration of EGF is considerable and widely used in human cancer cell research , and is relevant to those used in the in vitro studies of E . multilocularis and S . mansoni [32 , 37] . However , it could exceed the physiological concentrations in liver . Considering that the larval development of E . multilocularis causes host liver tissue destruction and regeneration while EGF is continually made available to the liver and strongly produced during regeneration processes [41–42] , it will be interesting to investigate the effects of host EGF observed in this study on the parasite in future using in vivo infection models . The promoted proliferation of germinative cells upon EGF stimulation was supported by our EdU-BrdU dual labeling experiments ( Fig 2B and 2C and S3 Fig ) , which also suggest that the cell-cycle time for most of the actively proliferating germinative cells is less than three days ( S2 Fig and Fig 2B ) , similar to that for S . mansoni adult somatic stem cells [14] . It has been suggested that there are subpopulations of the germinative cells capable of maintaining their pluripotency and self-renewing like stem cells [15] . Although our data suggest that the increased number of EdU+BrdU+ cells ( as well as the increased ratio of EdU+BrdU+ cells to EdU+ cells ) may also result from the promoted self-renewal of these stem-like cell populations , due to the limitations of the dual labeling experiments we still could not distinguish the self-renewing cells from the transit amplifying cells . Specific molecular markers of the stem-like cell populations that work independently of proliferation would be needed for further explorations to clarify the contribution of EGF to germinative cell self-renewal in E . multilocularis . Addition of host EGF promotes germinative cell proliferation , which then promotes in vitro-cultivated protoscoleces to form metacestode vesicles ( Fig 1C ) . The formation of vesicle from protoscolex occurs in vivo following the rupture of parasite cysts and distribution of protoscoleces , and is thought to contribute to prolonged parasite survival in the intermediate host [19 , 28] , which would result in a poor prognosis after surgery-induced rupture of parasite cysts in human echinococcosis , at least in cystic echinococcosis ( CE ) . Interestingly , we found that addition of host EGF may not only promote this formation process but also initiate it by triggering activation of the germinative cells from a quiescent state in the developed protoscoleces ( S6 Fig ) . Koziol et al . [15] indicated that there is a large population of germinative cells capable of proliferation in the developed protoscolex , but they remain in a quiescent state or with slow cell-cycle kinetics for as long as the protoscolex remains resting within the metacestode . These quiescent germinative cells were activated when the protoscoleces were activated by artificially mimicking the ingestion by the definitive host or when the protoscoleces were in vitro cultured in serum-containing DMEM . Thus it is tempting to suggest that host factors activate the quiescent germinative cells to re-enter the cell-cycle for proliferation and self-renewal , which may further stimulate the protoscoleces to mature into adults within the definitive host’s intestine or to form metacestode vesicles in the intermediate host’s liver and other tissues . It is still unclear how protoscoleces alternate between developmental fates: the adult or metacestode vesicle . In any case , this unique development potential for protoscoleces is based on the germinative cells , which may response to different host-derived signals from different host tissue microenvironments . Our data suggest that host factors may play a vital role in host-parasite interaction via mediating the relative signaling pathways in the parasite to regulate germinative cell functions . It has been suggested that host EGF could activate the highly conserved Ras/Raf/MEK/ERK signaling cascade in E . multilocularis , which is probably mediated by the parasite’s EGF-receptor-like kinase [31–33] . Besides EmER , the first EGFR homologue identified in E . multilocularis [31] , two additional members of the EGFR family could be identified , which display significant homologies with human EGFR and S . mansoni EGFR in the functional domains ( S4A–S4D Fig ) . Like EmER , these two EGFR homologues are continually expressed in E . multilocularis metacestode vesicles and protoscoleces ( S4E Fig ) . It will be interesting to clarify their roles as the EGF receptor kinase in the parasite’s development within the host in the future work . In this study , using the Xenopus oocyte expression system we show that one of the EGFR homologues EmER can be activated by host EGF ( Fig 3 ) . We also show here that inhibition of the MEK/ERK signaling activation by either the EGFR inhibitors CI-1033 and BIBW2992 or the MEK inhibitor U0126 significantly impaired E . multilocularis germinative cell proliferation , larval growth and development ( Figs 4 and 5 and S5 Fig ) . In mammals , the MEK/ERK pathway plays a critical role in regulating stem cells . For example , it is required for maintenance of stemness and self-renewal of mouse neural stem/precursor cells [43–44] . The role of MEK/ERK pathway in mammalian embryonic stem cells ( ESCs ) is much more complex . The MEK/ERK signaling plays a functional role in promoting differentiation of mouse ESCs , while it promotes self-renewal of human ESCs ( reviewed in [20] and [45] ) . In invertebrates , much of what is known about the role of the MEK/ERK signaling in regulation of stem cells derives from the studies of Drosophila . It has been extensively documented that the EGFR-dependent activation of the MEK/ERK signaling pathway is essential for promoting the maintenance and self-renewal of various types of adult somatic stem cells in Drosophila [23–24 , 46] . Our data also show that EGF-promoted germinative cell proliferation and larval growth rely on the activation of the parasite’s EGFR/ERK signaling . However , it still remains unclear that the contribution of EGFR/ERK signaling to the promoted proliferation is attributed to the direct response in the germinative cells or to the indirect response to a second signal produced by their surrounding differentiated cells upon EGF stimulation , or to both . Further analysis of the EGFR activation in proliferating germinative cells would be needed to clarify this issue . While downstream of EGFR lies the PI3K/AKT , MEK/ERK and STAT3 pathways , our findings define here that the MEK/ERK pathway contributes to the role of EGFR signaling in regulating E . multilocularis germinative cell proliferation . Considering that the PI3K/AKT pathway in E . multilocularis was recently suggested to be involved in the host insulin-stimulated germinative cell proliferation [28] , we also treated metacestode vesicles with the PI3K inhibitor LY294002 , which was shown to effectively inhibit E . multilocularis PI3K activity [28] . We found that LY294002 did not exhibit as obvious an inhibitory effect as U0126 on the proliferation of germinative cells but slightly decreased the number of EdU+ cells in metacestode vesicles . Future work would be required for evaluating the contribution of EGFR/AKT/PI3K signaling to the regulation of E . multilocularis germinative cells . Increasing evidence has shown that the inhibitors originally designed against the human kinases can effectively inhibit the activity of related kinases in E . multilocularis [28 , 47–49] . Based on the evolutionary conservation among the kinases of vertebrates and invertebrates ( including invertebrate parasites ) , it has been widely considered that small molecules that target human kinases are promising drug candidates for treating human helminthiasis , including echinococcosis [50–51] . Considering our observations that both basal ERK phosphorylation and ERK phosphorylation induced by administered EGF were effectively suppressed by either CI-1033 or U0126 ( Fig 5A–5D ) , it is therefore conceivable that the kinase inhibitors used in this study could impair EGFR/ERK signaling in E . multilocularis . Although these inhibitors were used within the range concentrations required for these compounds to specifically inhibit their respective targets in humans , it is possible that they may also have cellular targets other than EGFR/ERK signaling in E . multilocularis . Our findings suggest that exogenous EGF-activated EGFR/ERK signaling in the parasite was inhibited by CI-1033 . Given that the long-term in vitro maintenance of E . multilocularis larvae and primary cells requires continuous presence of host cell-derived growth factors [29] , it is reasonable to assume that EGFR inhibitors could impair the activation of parasite’s EGFR upon host EGF stimulation , which might be the main reason for the diminished germinative cell proliferation and the impaired larval growth and development ( Fig 4 and S5 Fig ) . However , our data could not exclude the possibility that EGFR inhibitors may also impair the parasite’s EGFR activation mediated by its own EGF molecules . It has been recently shown that neoblast clonal expansion in the free-living flatworm planarian is regulated by its own EGF-mediated EGFR signaling [52] . Since the parasite also possesses a putative EGF homologue [53] , this endogenous EGFR ligand-mediated signaling might also be involved in regulating E . multilocularis germinative cells . Improved in vitro cultivation systems and methods that could avoid/reduce the impacts of host factors will be helpful to investigate the roles of this endogenous signaling in germinative cell regulation . Although stem cell-like germinative cells has been widely described in tapeworms and their roles in the parasite’s development within the host are thought to be of fundamental importance , there are still long standing gaps in our knowledge of mechanisms controlling the behavior of these cells . This study defines an essential role for the EGF-mediated EGFR/ERK signaling in promoting germinative cell proliferation in E . multilocularis . It makes an effort to unravel the mechanisms of regulation of tapeworm germinative cells in response to host-derived growth factors , and helps in understanding the delicate developmental strategies of these parasites within the host . Targeting the signaling pathways involved in regulating germinative cells may provide a novel therapeutic strategy against echinococcosis and other human cestodiasis .
|
E . multilocularis is considered as one of the most lethal parasitic helminth in humans . It grows like tumors mainly in human liver and infiltrates other tissues , and even metastasizes . It is believed that the parasite possesses a population of stem cell-like cells , the germinative cells . These cells are totipotent , have the ability for extensive self-renewal , and drive the parasite’s development and growth in the host . However , mechanisms controlling the behavior of germinative cells are poorly understood . Here , we show that the highly conserved EGFR/ERK signaling pathway in the parasite promoted germinative cell proliferation upon addition of human EGF ( epidermal growth factor ) in vitro , resulting in stimulated growth and development of the parasite . Our study provides information important for understanding this mechanism regulating germinative cells and the complex host-parasite interaction , and we hope it will help in developing new therapeutic strategies for the treatment of human helminthic infections .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"medicine",
"and",
"health",
"sciences",
"vesicles",
"pathology",
"and",
"laboratory",
"medicine",
"cell",
"processes",
"egfr",
"signaling",
"vertebrates",
"animals",
"xenopus",
"animal",
"models",
"developmental",
"biology",
"model",
"organisms",
"amphibians",
"developmental",
"signaling",
"stem",
"cells",
"experimental",
"organism",
"systems",
"cellular",
"structures",
"and",
"organelles",
"research",
"and",
"analysis",
"methods",
"cell",
"proliferation",
"animal",
"cells",
"life",
"cycles",
"pathogenesis",
"xenopus",
"oocytes",
"signal",
"transduction",
"cell",
"biology",
"host-pathogen",
"interactions",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"cell",
"signaling",
"larvae",
"frogs",
"organisms"
] |
2017
|
EGF-mediated EGFR/ERK signaling pathway promotes germinative cell proliferation in Echinococcus multilocularis that contributes to larval growth and development
|
Paracoccidioidomycosis ( PCM ) is an infectious disease endemic to South America , caused by the thermally dimorphic fungi Paracoccidioides . Currently , there is no effective human vaccine that can be used in prophylactic or therapeutic regimes . We tested the hypothesis that the immunogenicity of the immunodominant CD4+ T-cell epitope ( P10 ) of Paracoccidioides brasiliensis gp43 antigen might be significantly enhanced by using a hepatitis B virus-derived particle ( VLP ) as an antigen carrier . This chimera was administered to mice as a ( His ) 6-purified protein ( rPbT ) or a replication-deficient human type 5 adenoviral vector ( rAdPbT ) in an immunoprophylaxis assay . The highly virulent Pb18 yeast strain was used to challenge our vaccine candidates . Fungal challenge evoked robust P10-specific memory CD4+ T cells secreting protective Th-1 cytokines in most groups of immunized mice . Furthermore , the highest level of fungal burden control was achieved when rAdPbT was inoculated in a homologous prime-boost regimen , with 10-fold less CFU recovering than in non-vaccinated mice . Systemic Pb18 spreading was only prevented when rAdPbT was previously inoculated . In summary , we present here VLP/P10 formulations as vaccine candidates against PCM , some of which have demonstrated for the first time their ability to prevent progression of this pernicious fungal disease , which represents a significant social burden in developing countries .
Paracoccidioidomycosis ( PCM ) is a neglected dermal , mucosal , and respiratory disease endemic to South America , which is caused by the thermally dimorphic fungi Paracoccidioides spp . [1] . The acquisition of Paracoccidioides spp . occurs mainly by inhalation of air-borne fungal propagules [2] . In the lungs , morphological changes occur from mycelia to the yeast form , which leads to alveolar injuries , and lympho-haematogenous spreading may occur , depending upon host susceptibility . Inflammatory responses sustained by interferon gamma ( IFNγ ) and interleukin 2 ( IL-2 ) are desirable to enhance microbial killing by macrophages and polymorphonuclear cells ( PMN ) , whereas high levels of IL-4 and IL-10 are associated with the spread of fungus to other organs and worse prognosis [3 , 4] . Due to the limitations of antifungal therapies [3] , immunogenic antigens , and in particular peptide P10 ( amino acid sequence QTLIAIHTLAIRYAN ) , the immunodominant CD4+ T-cell epitope of the 43 kDa glycoprotein ( gp43 ) from P . brasiliensis , were selected for vaccine purposes . P10 has the capacity to bind to major histocompatibility complex class II molecules from mice and humans [5 , 6] . Amongst P10-based vaccines , DNA vaccines containing P10 and IL-12 encoding sequences , and a P10 mixture with a TLR-5-engaging-molecule , i . e . flagellin , were the most effective formulations for controlling the fungal burden in the host [7 , 8] . Other vaccine strategies were also developed , such as recombinant proteins and radio-attenuated Pb18 yeast cells , which were also shown to be effective against P . brasiliensis infection , displaying a safe profile [9 , 10] . Chimeric virus-like particles ( VLPs ) are artificial structures in which relevant protective epitopes from pathogen antigens may be included . They behave as harmless but potent immune-stimulating scaffolds , representing a new strategy for vaccine development [11 , 12] . Hepatitis B virus core antigen ( HBcAg ) is considered an excellent vaccine carrier that self-assembles into VLPs [13] . Vaccines based on purified proteins or peptides , which are very efficient at inducing antibody responses , present some challenges at inducing T-cell responses , and frequently induce weak protection against intracellular infectious agents . However , proteins that self-assemble into VLPs can induce potent T-cell responses because they are rapidly engulfed and processed by macrophages and dendritic cells , the most efficient T-cell response stimulators [14] . Additionally , regulatory agencies already approved vaccines based on VLP for human use [11] . Live , replication-deficient recombinant viruses are also interesting platforms for delivering vaccines due to their safety profile [15–18] and the fact that some vectors deliver large amounts of foreign antigens directly into antigen-presenting cells , significantly inducing , in particular , cellular immunity [19 , 20] . Among recombinant viruses , adenoviral vectors are considered one of the most efficient platforms for vaccination , due to their capacity to elicit memory responses mediated by T lymphocytes . In recent years , and despite widely-spread concerns about pre-existing immunity blockade of heterologous antigen-specific responses , recombinant human type 5 adenovirus ( HuAd5 ) vaccine candidates have shown high-efficiency in the induction of memory T cells and protection against infectious agents such as Ebola virus [15] , Trypanosoma cruzi [19] , Leishmania infantum [20] , or Mycobacterium tuberculosis [21] . In this context , we developed two vaccine candidates based on the same VLP construct , both containing the main CD4+ T-cell-specific epitope from P . brasiliensis ( P10 ) within the C-terminal portion of HBcAg ( fused as alanine-flanking sequences to amino acid 179 of HBcAg ) . First candidate was a P10/ ( His ) 6-purified protein ( rPbT ) , and the second was a recombinant P10/human type 5 adenoviral vector ( rAdPbT ) . We show here the immunogenicity of these vaccine candidates when administered individually or combined in a prime-boost protocol , as well as the protection induced in mice against a highly pathogenic strain of fungus that can cause a severe disease in humans .
Initially , a synthetic , sequence verified , HASS-HBcAg-P10 cassette , contained in pUC57 plasmid , was unidirectionally cloned into the adenoviral shuttle vector pAdCMV-Link to form the pAdPbT plasmid ( pAd-HASS-HBcAg-P10 ) , using BamHI/BglII and HindIII restriction sites . A recombinant adenovirus ( rAdPbT ) was generated by co-transfection of this pAdPbT shuttle vector together with a plasmid containing the human type 5 adenoviral E1/E3-deleted genome ( pJM17 ) into HEK-293 cells [22] . rAdPbT-infected 293 cells were used for western blotting to verify expression of HBcAg/P10 chimera from the recombinant adenovirus sequences . A subsequent overnight CsCl gradient centrifugation allowed for the recovery of a high titer recombinant adenovirus [23] . In parallel , HBcAg-P10 and HBcAg-pp89 encoding sequences contained in pUC57 plasmids were separately cloned into the BamHI and HindIII restriction sites of pET28a system ( Novagen ) to form pET28aPbT and pET28aCMV plasmids , respectively , then transformed into E . coli Rosetta ( DE3 ) lineage ( Novagen ) for protein expression ( rPbT and rCMV , respectively ) under IPTG induction . Purification was performed using Ni2+ columns under denaturing conditions ( Probond Purification Kit , Life Technologies ) , according to the manufacturer’s instructions . Each protein was dialyzed against refolding buffer ( 10 mM Tris; 0 . 5 mM DTT; 20% glycerol; pH 7 . 2 ) at 4°C for 16 h . SDS-PAGE and western blotting were performed to verify expression of recombinant HBcAgs . P10 peptide sequence QTLIAIHTLAIRYAN was synthesized by F-moc solid-phase method by GenScript . The purity of the peptide was at least 95% as judged by HPLC . Animal experiments were carried out with strict adherence to the Ethics Commission on Animal Use ( CEUA ) from Federal University of Minas Gerais ( UFMG ) , Brazil , ( Protocol 206/11 ) and the Brazilian Federal Law 11 , 794 ( October 8 , 2008 ) . Male BALB/c mice ( 6–8 weeks old ) were purchased from UFMG´s central animal facilities , and housed in clean micro-isolator cages with food and water ad libitum for 12h light/dark cycles . Nine animals were assigned to each group . A detailed description of the prime-boost protocols used , vaccine candidates combinations and time frames ( immunization intervals , challenge and sacrifice ) are shown in Fig 1 . Ten micrograms of recombinant proteins ( rPbT or rCMV ) or peptide ( P10 ) were separately emulsified in Montanide ISA 720 adjuvant ( Seppic , France ) in a final volume of 100 μL . Recombinant adenovirus ( rAdPbT ) at 3×108 plaque-forming units ( PFU ) was prepared in apyrogenic sterile phosphate buffered saline ( PBS ) supplemented with 1% normal mouse serum in a final volume of 100 μL . Experimental groups were immunized with rAdPbT and rPbT in prime-boost homologous and heterologous protocols as shown in Fig 1 . Control groups were immunized with P10 or rCMV or PBS emulsified in adjuvant in a homologous prime-boost protocol . All formulations were administrated subcutaneously in the tail base at the first and 30th experimental days . Mice were anesthetized with a solution containing ketamine hydrochloride ( 80 mg . kg-1 ) and xylazine ( 10 mg . kg-1 ) in sterile PBS , then inoculated intratracheally with 3×105 viable yeast cells of virulent Pb18 strain [8] . Non-infected animals were inoculated with sterile PBS . Mice that were challenged with the Pb18 strain but previously non-immunized were called mock-immunized . Animals were monitored daily until euthanasia . Inflammatory mediators such as IFNγ , TNF-α , IL-12 , IL-10 , and IL-4 were detected in 100 mg of lung homogenates using a BD Opteia ELISA Kit ( BD Bioscience , USA ) , according the manufacturer’s instructions . Spleens were aseptically removed from mice and red blood cells were disrupted by osmotic lysis . Splenocytes ( 1 . 5×106 viable cells ) were labeled with 5 μM CFDA-SE ( carboxyfluorescein succinimidyl ester , Sigma Aldrich , USA ) [24] , then incubated in RPMI 1640 medium supplemented with 10% of FBS and 40 mg/L of gentamycin for 80h at 37°C and 5% CO2 . Positive ( cell stimulation with ConA at 4 μg/mL ) and negative ( non-stimulated cells ) controls of proliferation were performed using non-immunized and non-infected ( NI ) mice . Intratracheal Pb18 infection itself was the stimulus for evocation of memory CD4+ T-cell responses . After 80h of restimulation , splenocytes were harvested and labeled with a panel of antibodies ( anti-CD3-APC-Cy7 , anti-CD4-PerCP-Cy5 . 5 , anti-CD44-PE , and anti-CD62L-APC; BD Bioscience , USA ) . All samples were acquired in a BD FACSCanto II instrument ( BD Bioscience , USA ) and results were analyzed using FlowJo software v7 . 5 ( TreeStar ) . Mice were euthanized for aseptic removal of the lungs , liver , and spleen . Each organ homogenate was plated onto Brain Heart Infusion agar plates ( Difco , USA ) supplemented with 4% FBS , 5% spent P . brasiliensis 192 strain culture supernatant , and 40 mg/L of gentamicin . Plates were incubated at 36°C for 20 days for yeast growth , and the colony-forming unit ( CFU ) counts were determined per gram of tissue . For evaluation of histopathological alterations , each organ was stained with hematoxylin-eosin ( HE ) . All data were subjected to ANOVA ( Tukey’s post hoc test ) and Student’s t-test to analyze significant differences between groups . Results are shown as means ± standard deviation .
A genetic chimera constructed with sequences encoding hepatitis B virus core antigen ( HBcAg ) fused to those of the promiscuous and protective P . brasiliensis gp43 CD4+ T-cell epitope ( P10 ) was used to generate two vaccine candidates , a recombinant adenovirus ( rAdPbT ) and a recombinant protein purified from E . coli ( rPbT ) . An experimental control , containing a murine cytomegalovirus ( MCMV ) epitope from immediate early protein 1 ( pp89 ) was also constructed and expressed in E . coli . An overview of the cloning steps and expression analyses are shown in Fig 2A and 2B . Amino acid sequences of HBcAg-chimeric VLP constructs are shown in Fig 2C . VLPs assembly of our vaccine candidate was demonstrated by transmission-electron microscopy , and shown in Fig 2D . Inflammatory responses induced by experimental PCM were quantified in lung homogenates at the 75th and 105th days . At the 75th day , lungs of mice previously immunized with rAdPbT/rAdPbT and rAdPbT/rPbT protocols secreted the highest levels of IL-12 compared with mock-immunized ( MI ) or non-immunized ( NI ) mice , while other P10-based formulations secreted lower levels of this cytokine ( Fig 3A ) . IFN-γ or TNF-α secretion was not statistically different after challenge among infected mice ( Fig 3B and 3C ) although differences were highly significant when compared to non-infected animals , suggesting that Pb18 infection strongly boosts production of these cytokines in all animals masking any previous differences due to vaccination . Interleukin-10 was less secreted in lungs of mice immunized with P10-based formulations ( Fig 3D ) , while IL-4 was marginally less secreted only in those groups that received any of the P10/VLP constructs , i . e . rAdPbT/rPbT , rPbT/rAdPbT and rPbT/rPbT , when compared to mock-immunized mice ( Fig 3E ) . As had been observed at day 75 , lungs of rAdPbT/rAdPbT- and rAdPbT/rPbT-immunized mice also secreted the highest levels of IL-12 compared to mock-immunized mice at day 105 ( Fig 4A ) , while all other mice immunized with P10-based formulations displayed levels of this cytokine that were not statistically different ( a borderline significance was observed in some cases for group rPbT/rAdPbT ) . Regarding IFNγ , the lungs of rAdPbT/rAdPbT secreted more cytokine than mock-immunized mice ( Fig 4B ) . As for TNF-α secretion , no group secreted this inflammatory cytokine at levels above those of mock-immunized mice ( Fig 4C ) . As previously noted , interleukin-10 was significantly less secreted in lungs of mice previously immunized with P10-based formulations in relation to mock-immunized mice ( Fig 4D ) . Finally , interleukin-4 was not statistical different between all infected mice ( Fig 4E ) . Proliferative CD4+ T cells ( Fig 5A ) were monitored in CFSE histograms , where CD3+CD4+CFSEHIGH dotted lines histogram represents non-specific proliferation and CD3+CD4+CFSELOW continuous line histogram represents specific proliferation after Pb18 infection , as shown in Fig 5B . As expected , CD4+ T lymphocytes of all infected mice previously immunized with P10-based formulations displayed proliferative responses . The highest proliferative response was displayed in the rAdPbT/rAdPbT group ( red line; ~32% ) followed by the rAdPbT/rPbT ( blue line; ~26% ) , rPbT/rAdPbT ( purple line; ~22% ) , P10/P10 ( green line; ~15% ) , and rPbT/rPbT groups ( black line; ~2 . 5% ) , according to Fig 5C and 5D , respectively . CD4+ T cells of rCMV/rCMV , PBS/PBS , and mock-immunized mice were unable to proliferate in CFSELOW region ( <1% ) . Next , we investigated memory CD4+ T-cell phenotypes that were evoked after Pb18-challenge and the influence of immunoprophylaxis previously performed , based on circulating cells expressing activation ( CD44 ) and homing ( CD62L ) cell-surface molecules ( Fig 6A and 6B ) . The central memory CD4+ T-lymphocyte ( TCM ) phenotype ( CD3+CD4+CD44HIGHCD62LHIGH ) was predominantly displayed in mice previously immunized with P10-based formulations . Amongst them , the rAdPbT/rAdPbT group displayed the most expressive TCM phenotype ( ~20% ) , followed by the rAdPbT/rPbT and rPbT/rPbT ( ~16% ) , then the rPbT/rAdPbT and P10/P10 ( ~14% ) groups . The TCM phenotype of mice vaccinated with P10-based formulations was percentually more expressive than that of rCMV/rCMV , PBS/PBS , and mock-immunized mice ( ~8% ) , as shown in Fig 6C ( black bars ) . We also verified the effector memory ( TEM ) phenotype ( CD3+CD4+CD44HIGHCD62LLOW ) , which was more expressed in the rAdPbT/rAdPbT group ( ~10% ) . Unexpectedly , the rAdPbT/rPbT , rPbT/rAdPbT , rPbT/rPbT , and P10/P10 groups displayed similar responses of effector T cells ( ~ 6% ) , which were higher than those of the rCMV/rCMV , PBS/PBS , and mock-immunized groups ( ~2 . 5% ) , as shown in Fig 6C ( white bars ) . To verify the protection induced by P10-based vaccines against P . brasiliensis infection , the fungal burden was recovered from the lungs , liver , and spleen at the 75th and 105th days . Unexpectedly , rPbT/rAdPbT , rPbT/rPbT , and P10/P10 were unable to reduce the fungal burden in the lungs at the 75th day . On the other hand , mice immunized with a homologous adenovirus prime-boost regimen ( rAdPbT/rAdPbT ) showed a pronounced reduction in the fungal burden in the lungs ( 8-fold less than mock-infected mice ) , and a heterologous protocol initiated with adenovirus inoculation ( rAdPbT/rPbT ) was able to reduce the fungal burden ( 4-fold less than mock-immunized mice ) , as shown in Fig 7A . Mice immunized with control protein ( rCMV/rCMV ) or with adjuvant in PBS ( PBS/PBS ) displayed a similar fungal burden to that of mock-immunized mice . Colony-forming units were not recovered in the liver or spleen at the 75th day . Fortunately , all P10-based formulations ( rAdPbT/rPbT , rPbT/rAdPbT , rPbT/rPbT , and P10/P10 ) were able to control fungal proliferation in the lungs compared with the mock-immunized group at the 105th day ( Fig 7B ) . As previously noted , rAdPbT/rAdPbT was again the most effective protocol to control the fungal burden in the lungs ( 10-fold less than mock-immunized mice ) , followed by rAdPbT/rPbT and rPbT/rAdPbT ( 5-fold less than mock-immunized mice ) , then P10/P10 and rPbT/rPbT-immunized mice ( 2-fold less than mock-immunized mice ) . Unexpectedly , the fungal burden in the lungs of all mice was higher at the 105th day than those observed at the 75th day , highlighting the efficacy of our virulent Pb18 strain in the immunoprophylaxis assay . However , mice immunized with rAdPbT in the prime and/or boost regimen were able to prevent the fungus spreading from the lungs to other organs ( Fig 7C and 7D ) . Tissue injuries caused by virulent Pb18 infection were observed in organs sectioned at the 105th day ( Fig 8 ) . Yeast cells were found associated with alveolar injuries in an independent manner in immunized and non-immunized mice . Furthermore , infiltration of foam macrophages and neutrophils were found in the bronchial lumen , and alveolar thickening was also observed . Granulomas containing yeast cells were observed in the lung parenchyma of all infected mice , as expected because is this organ which receives the challenge microorganisms in the first place . On the other hand , it is of enormous interest the fact that there were no yeast cells in liver or spleen sections of any animal that received at least one dose of rAdPbT , the adenovirus-immunized groups of mice , whereas some hepatic granulomas containing yeast cells and neutrophil infiltration and even some free individual yeast cells were found in the rPbT/rPbT , P10/P10 , rCMV/rCMV , PBS/PBS , and mock-immunized groups .
The main challenge in vaccination is how to induce a long-lasting protective memory against microbes and their antigens . Selection of candidate vaccine antigens , vectors , dosage , and vaccination protocols are the keys to improving the immunogenicity and efficacy of the formulation [25 , 26] . In paracoccidioidomycosis , the stimulation of Th-1 cells is desirable to activate macrophages associated with the clearance of microbes in the intracellular environment [4 , 27] . Recombinant VLP/P10 formulations were made in an attempt to improve P10 immunogenicity by presenting it to the immune system and to enhance microbial killing mediated by cellular immunity . P10/HBcAg chimeras were expressed either in cells infected with a recombinant adenoviral vector ( rAdPbT ) or as recombinant E . coli His ( 6 ) -purified proteins ( rCMV and rPbT ) and self-assembled into VLPs according to our transmission electron microscopy data . Cytokine profiles of the mock-immunized , PBS/PBS and rCMV/rCMV groups suggested the occurrence of tissue injury and modulation of the inflammatory response by a Th2-biased immune response . These data corroborate the poor CD4+ T lymphocyte response evoked by intratracheal Pb18 challenge in these groups . In a PCM immunoprophylaxis study , a DNA vaccine based on HSP65 from Mycobacterium leprae induced a protective immune response against P . brasiliensis infection via high-level secretion of cytokines IFNγ and IL-12 [28] . In our study , the recombinant VLP protein carrying a specific CD8 T-cell epitope from murine cytomegalovirus ( rCMV ) did not cross-protect mice against P . brasiliensis proliferation , demonstrating the specificity of the CD4+ T cell response for protection , and allowing the use of that chimera for actual evaluation of P10-based vaccine efficacy . Braun and colleagues [14] demonstrated that a VLP vaccine could elicit both humoral and cellular immune responses , as did Almeida and colleagues with Plasmodium CSP antigen expressed on the VLP surface [13] . The efficacy of VLP vaccines was also shown in clinical trials against human papilloma virus ( HPV ) where volunteers were entirely protected against a new infection by some viral serotypes [29] . In our study , VLP/P10 vaccines could efficiently induce protective memory P10-specific CD4+ T lymphocytes . Immunization of mice with synthetic P10 peptide or purified VLP/P10 protein led to a delayed reduction of the fungal burden in the lungs that could be related to the decreased spread of Pb18 in the host . Concerning synthetic P10 vaccination , a significant protective response against P . brasiliensis infection was expected , in view of previous results obtained by members of our group [8] . Thus , Taborda and colleagues [30] had shown the protective immune responses induced against murine PCM by P10 peptide , either by using different adjuvants ( Freund´s complete and incomplete or Flagellin ) , doses ( 1–20 μg ) , vaccination protocols ( prime-boost with repetitions ) and routes of inoculation ( intranasal , subcutaneous , and intraperitoneal ) [8 , 30 , 31] . However , purified VLP/P10 protein given by a homologous prime-boost protocol induced a low proliferative response of CD4+ T lymphocytes evoked by P18 infection and a limited , though significant , control of the fungal burden in mice . In the present study , synthetic P10 peptide and purified recombinant proteins were emulsified in Montanide ISA 720 adjuvant , composed of natural metabolizable oil and a highly refined emulsifier from the mannide monooleate family [32] , which is allowed for vaccine trial [33] and is able to induce the switch towards Th1 responses [33 , 34] . There is no need to add microorganisms or microbial products to this formulation , as opposed to Freund’s complete emulsion [8 , 30 , 31] and some Alum preparations [35 , 36] . In standardization experiments ( S1 Fig ) the subcutaneous route was the most immunogenic when considering both the protein chimera and the adenovirus vector . This had been also previously shown for rPbT [13] as well as for the adenovirus vector ( in all of our previous publications on this subject since 1986 ) . Regarding immunization with P10 peptide in adjuvant , we could not try to improve immunogenicity by increasing the frequency of inoculations , because maintaining the same time intervals among immunization would impede subsequent yeast challenge with Pb18 ( it has to be done in adolescent to young adults mice ) [7–10 , 28 , 31 , 35 , 36] , while if shortening the inoculation intervals , immunogenicity would have decreased due to lymphocyte Activation-Induced Cell Death ( AICD ) [37] . Recombinant adenovirus inoculation was substantially more immunogenic than purified recombinant proteins or synthetic P10 peptide , even when the viral vector was used in a homologous prime-boost protocol , something already observed in previous studies [21] , demonstrating that preexisting host immunity against the vector was not enough to affect its immunogenicity or its capacity to delivery foreign antigens directly into antigen presenting cells . The highest secretion of proinflammatory cytokines from memory CD4+ T cells ( central and effector phenotypes ) was achieved when mice were primed with recombinant adenovirus in prime-boost regimens . This immune response profile is desirable for the clearance of infectious agents in the intracellular environment [4] . Of interest too is the fact that all groups of mice immunized with a P10-derived construct secreted significantly less IL-10 cytokine in the lungs than those mice mock-immunized , suggesting the induction of a focused and favorable Th1 immune response , with little regulation , in the main organ were lymphocyte effector functions should be displayed . More importantly , all mice immunized at some point with the recombinant adenovirus were protected against Pb18 systemic dissemination . It has been shown that attenuated Pb18 yeast formulation , one of the most promising vaccine candidates against PCM , can evoke Th-1/Th-2 responses to control the host fungal burden [10] . In addition , recombinant Pb27 and Pb40 proteins based on Pb18-fractionated antigens have been shown to play an important role in defense against P . brasiliensis infection [35 , 36] . However , the mechanisms by which these formulations elicit a protective immune response are unclear . Immunophenotyping assays should be used to clarify the immunological memory elicited by whole and fractionated Pb18 antigens , which culminates in microbial killing . Among live vaccines , recombinant adenoviral vectors , in particular , human type 5 adenovirus ( Hu5Ad ) are considered the most powerful activators of T lymphocytes [38 , 39] . In experimental trypanosomiasis , HuAd5-based vaccines constructed by us strongly elicited memory T cells that led to a pronounced reduction in parasitemia , an augmentation of mouse survival , and a regression of chronic cardiomyopathy [19] . In M . tuberculosis , a recombinant HuAd5 vaccine elicited a robust immune response of T lymphocytes when the vector was inoculated alone or as a booster regimen in volunteers previously immunized with BCG formulation , including individuals already pre-sensitized with Hu5Ad . The vaccine was well tolerated , effective , and safe for human use [21] . In this context , we believe that our replication-deficient rAdPbT vector , also built from the Hu5Ad genome , may be effective in patients suffering from PCM , as well as for immunization of individuals whose occupation may predispose them to P . brasiliensis infection . The main advantages of recombinant adenoviruses in relation to subunit vaccines are: i ) they are powerful innate immune activators of Toll-like receptor and MyD88 pathways , and there is no need to use adjuvant with them , ii ) they have the capacity to deliver high amounts of foreign antigens to the host , iii ) they mimic intracellular infection by microbes , improving antigen presentation to the immune system iv ) they are activators of T lymphocytes in both innate and adaptive immune responses . Furthermore , recombinant adenoviral vectors have some advantages over other live vaccine candidates: i ) non-replicative ability in the host; ii ) non-integrative ability in the host genome; iii ) very low oncogenic potential; iv ) facility of scaling up production; and v ) target antigen expression [21 , 40 , 41] . In contrast with other mouse strains in which P . brasiliensis either leads to an acute and fatal outcome , e . g . in B10 . A mice , or to a chronic and regressive self-healing disease , i . e . the disease observed in A/Sn or A/J mice [42 , 43] , in the BALB/c strain of mice used in our study , Pb18 displayed an intermediate behavior [44] , more similar to the human PCM for which we used those mice as model . Thus , albeit fungal burden didn’t lead mice to death until the day of sacrifice , something that would require immunosuppression with dexamethasone [31] thus invalidating any of our current observations and conclusions , lymphohaematogenous dissemination , related to clinical worsening [45] , was significantly suppressed by the use of some of our vaccine candidates . Regarding tissue injuries , granulomas surrounding yeast cells and cellular infiltrate were found in the lungs of all mice with PCM as expected . However , the containment of yeast cells inside granulomas in lung parenchyma seemed to be more efficient in mice that received at least one adenovirus dosage in the prime-boost protocol , which was enough to prevent the fungus from spreading to other organs . These results corroborate with the CFU numbers recovered from organs of those mice that displayed resistance against PCM [46–47] . Eventually , our recombinant vaccines may be used in a PCM therapeutic regimen [36] or , in a challenging approach , as tools in cancer therapy [48–51] due to the parallel antitumor properties that have been shown for P10 [52] . In conclusion , some of our recombinant VLP/P10 formulations , inoculated in homologous and heterologous prime-boost protocols , elicited a strong , long-lasting cell-mediated immune response that led to intense control of local infection and prevented the systemic spread of the fungus in the host .
|
Human paracoccidioidomycosis ( PCM ) represents a serious public health issue due to its disabling sequelae in working-age people and mortality rates ( 8th among chronic infectious parasitic diseases in endemic countries and first among mycoses in Brazil ) . Although antifungal drugs have been widely used and provide clinical improvement for patients , the long duration of treatments ( commonly from 6 to 12 months ) has contributed to non-compliance and worsening of the disease in many cases . Induction of protective immune responses ( either prophylactic or immune-therapeutic ) remains the most cost-effective approach against most infectious agents . Members of our group have previously reported the protective properties of P10 , a peptide contained in Paracoccidioides brasiliensis gp43 antigen , against PCM . However , the magnitude of the CD4+ T-cell responses elicited lacked the capacity of completely protecting experimental animals . We demonstrate here that immunogenic virus-like particles ( VLP ) carrying multiple copies of P10 peptide substantially improve P10-related cellular immunity in mice . This was particularly true when an adenoviral immunization vector expressed the chimeric VLP . Moreover , VLP/P10 formulations were capable of controlling the host fungal burden and prevented fungal systemic dissemination . The efficacy of diverse VLP/P10 formulations in murine PCM showed at least one promising candidate vaccine against human PCM .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"blood",
"cells",
"medicine",
"and",
"health",
"sciences",
"immune",
"cells",
"immune",
"physiology",
"pathology",
"and",
"laboratory",
"medicine",
"yeast",
"infections",
"pathogens",
"immunology",
"microbiology",
"vaccines",
"viruses",
"dna",
"viruses",
"fungal",
"diseases",
"infectious",
"disease",
"control",
"immune",
"system",
"proteins",
"infectious",
"diseases",
"white",
"blood",
"cells",
"animal",
"cells",
"proteins",
"medical",
"microbiology",
"antigens",
"t",
"cells",
"microbial",
"pathogens",
"recombinant",
"proteins",
"immune",
"response",
"biochemistry",
"cell",
"biology",
"viral",
"pathogens",
"physiology",
"adenoviruses",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"organisms",
"recombinant",
"vaccines"
] |
2017
|
Recombinant vaccines of a CD4+ T-cell epitope promote efficient control of Paracoccidioides brasiliensis burden by restraining primary organ infection
|
The domestication and development of cattle has considerably impacted human societies , but the histories of cattle breeds and populations have been poorly understood especially for African , Asian , and American breeds . Using genotypes from 43 , 043 autosomal single nucleotide polymorphism markers scored in 1 , 543 animals , we evaluate the population structure of 134 domesticated bovid breeds . Regardless of the analytical method or sample subset , the three major groups of Asian indicine , Eurasian taurine , and African taurine were consistently observed . Patterns of geographic dispersal resulting from co-migration with humans and exportation are recognizable in phylogenetic networks . All analytical methods reveal patterns of hybridization which occurred after divergence . Using 19 breeds , we map the cline of indicine introgression into Africa . We infer that African taurine possess a large portion of wild African auroch ancestry , causing their divergence from Eurasian taurine . We detect exportation patterns in Asia and identify a cline of Eurasian taurine/indicine hybridization in Asia . We also identify the influence of species other than Bos taurus taurus and B . t . indicus in the formation of Asian breeds . We detect the pronounced influence of Shorthorn cattle in the formation of European breeds . Iberian and Italian cattle possess introgression from African taurine . American Criollo cattle originate from Iberia , and not directly from Africa with African ancestry inherited via Iberian ancestors . Indicine introgression into American cattle occurred in the Americas , and not Europe . We argue that cattle migration , movement and trading followed by admixture have been important forces in shaping modern bovine genomic variation .
High-throughput genotyping assays have allowed population geneticists to use genome-wide marker sets to analyze the histories of many species , including human [1] , cattle [2]–[4] , sheep [5] , dog [6] , horse [7] , yeast [8] , mouse [9] , [10] , rice [11] , [12] , maize [13]–[16] , grape [17] , and wheat [18] . We previously described the phylogeny of domesticated bovine populations using their genetic variation inferred from a sample of 40 , 843 single-nucleotide polymorphisms ( SNPs ) [3] . Although we had sampled 48 cattle breeds , we did not have samples from key geographic regions including China and Southeast Asia , Anatolia , the Baltic States , southern and eastern Africa , and the Iberian Peninsula . As a consequence of those gaps in geographic sampling , we were unable to address the origins of cattle in these regions and the extent to which these cattle influenced the population structure of regions such as the New World . We have now assembled a genomic data set which represents the largest population sampling of any mammalian species . This allows for an extremely detailed description of the population structure of domesticated cattle worldwide . Using this data set , we accurately establish the patterns of exportation , divergence , and admixture for domesticated cattle .
We used principal component analysis ( PCA ) [19] , ancestry graphs implemented in TreeMix [20] , and ancestry models implemented in ADMIXTURE [21] to analyze the relationships between 134 breeds of domesticated bovids ( Table S1 ) . These breeds arose from three domesticated ( sub ) species: Bos javanicus , Bos taurus indicus and Bos taurus taurus ( we use the terms breed and population interchangeably , due to the different definitions of breed worldwide ) . The principal source of SNP genotype variation was between B . t . taurus and B . t . indicus breeds ( Figure 1 ) . This split corresponds to the cattle which originated from the two separate major centers of domestication in the Fertile Crescent and Indus Valley [22] . Although Bos javanicus has a more distant common ancestor compared with Bos t . indicus and Bos t . taurus [3] , the uneven sample sizes and ascertainment of SNPs common in Bos t . taurus in the design of the BovineSNP50 assay [23] caused the Bos t . indicus/Bos t . taurus split to be the main source of variation in these data . The second principal component split African taurine cattle from Eurasian taurine , indicine , and Bali cattle . Early farmers were able to expand their habitat range because of the availability of a reliable supply of food and likely displaced indigenous hunter-gatherer populations by introducing new diseases [24] . The genomes of modern cattle reflect the history of animal movements by migratory farmers out of the ancient centers of cattle domestication . We first ran TreeMix with all 134 populations to identify patterns of divergence ( Figure 2 ) . We next ran TreeMix with 74 representative populations ( Figure 3 , residuals presented in Figure S1 ) and began to add migration edges to the phylogenetic model ( Figure 4 , residuals presented in Figure S2 , see Methods for an explanation of TreeMix ) . The proportion of the variance in relatedness between populations explained by the model began to asymptote at 0 . 998 ( a value also obtained by simulations [20] ) when 17 migration edges were fit ( Figure S3 ) . The consistency of these migration edges was evaluated using 5 independent runs of TreeMix with 17 migration edges ( Figure S4 ) . In addition to the migratory routes previously described from the Fertile Crescent to Europe [3] , we now find strong evidence of exportations from the Indian subcontinent to China and southeast Asia , India to Africa , Africa to the Iberian Peninsula and Mediterranean Europe , India to the Americas , and Europe to the Americas ( Figures 4 and 5 , discussed in detail in the following subsections ) . Subsequent to these initial exportations , there have been countless exportations and importations of cattle worldwide . When domesticated cattle were present and new germplasm was imported , the introduced cattle were frequently crossed with the local cattle resulting in an admixed population . Admixed populations were most readily identified when Bos t . indicus and Bos t . taurus animals were hybridized , which occurred in China , Africa , and the Americas ( crosses in Figure 1 ) . In the late 18th and 19th centuries , European cattlemen began forming closed herds which they developed into breeds [25] . Because breeds are typically reproductively isolated with little or no interbreeding , we found that the cross-validation error estimates continued to decrease as we increased the number of ancestral populations K modeled in the admixture analysis ( Table S2 ) . This reflects the large differences in allele frequencies that exist between breeds resulting from separate domestication events , geographic dispersal and isolation , breed formation , and the use of artificial insemination . The method of Evanno et al . [26] , which evaluates the second order rate of change of the likelihood function with respect to K ( ΔK ) , identified K = 2 as the optimum level of K ( Figure S5 ) . This method was overwhelmed by the early divergence between indicine and taurine cattle , and was not sensitive to the hierarchical relationships of populations and breeds [27] . As we increased the value of K , we recapitulated the increasingly fine structure represented in the branches of the phylogeny ( Figures 6 , S6 , S7 , S8 , S9 , S10 ) . Anatolian breeds ( AB , EAR , TG , ASY , and SAR ) are admixed between blue Fertile Crescent , grey African-like , and green indicine-like cattle ( Figures 5 and 6 ) , and we infer that they do not represent the taurine populations originally domesticated in this region due to a history of admixture . Zavot ( ZVT ) , a crossbred breed [25] , has a different history with a large portion of ancestry similar to Holsteins ( Figures 2 and S8 , S9 , S10 ) . The placement of Anatolian breeds along principal components 1 and 2 in Figure 1 [23] , the ancestry estimates in Figure 6 , their extremely short branch lengths in Figures 2–4 , and significant f3 statistics confirm that modern Anatolian breeds are admixed ( see Methods for explanation of f-statistics ) . For example , the Anatolian Southern Yellow ( ASY ) has 3 , 003 significant f3 tests , the most extreme of which has Vosgienne ( VOS , a taurine breed ) and Achai ( ACH , an indicine breed ) as sister groups with a Z-score of −43 . 69 . Our results support previous work using microsatellite loci [28] which inferred Anatolian cattle to possess indicine introgression . We further demonstrate that Anatolian breeds have introgression from African taurine . We calculated f4 statistics with East Anatolian Red , Anatolian Southern Yellow , and Anatolian Black as sister , and N'Dama , Somba , Lagune , Baole , Simmental , Holstein , Hereford , and Shorthorn as the opposing sister group . From Figure 2 , we would expect these relationships to be tree-like . But 45 of the possible 84 f4 tests indicated significant levels of admixture . The most significant was f4 ( East Anatolian Red , Anatolian Southern Yellow; Somba , Shorthorn ) = −0 . 0026±0 . 0003 ( Z-score = −8 . 10 , alternative trees have Z-scores of 9 . 88 and 5 . 20 ) . If African and Asian taurines were both exported from the Fertile Crescent in similar numbers at about the same time , we would expect them to be approximately equally diverged from European taurines . However , African taurines were consistently revealed to be more diverged from European and Asian taurines ( Figures 1 , 2 , 3 , and 5 , Anatolian breeds are not considered in this comparison because of their admixed history ) . Two factors appear to influence this divergence . First , European cattle were exported into Asia and admixed with Asian taurines . In the admixture models in which K = 15 or 20 ( Figures S9 and S10 ) , there was evidence of European taurine admixture in the Mongolian ( MG ) , Hanwoo ( HANW ) , and Wagyu ( WAGY ) breeds . We ran TreeMix with 14 representative populations and estimated Wagyu to have 0 . 188±0 . 069 ( p-value = 0 . 003 ) of their genome originating from northwestern European ancestry ( Figure 7 ) . We also see some runs of TreeMix placing a migration edge from Chianina cattle to Asian taurines ( Figure S4 ) . We ran f4 tests with Mongolian , Hanwoo , Wagyu , Tharparkar ( THA ) , or Kankraj ( KAN ) as sister populations , and Piedmontese ( PIED ) , Simmental ( SIM ) , Brown Swiss ( BSW ) , Braunvieh ( BRVH ) , Devon ( DEV ) , Angus ( AN ) , Shorthorn ( SH ) , or Holstein ( HO ) as the opposing pair of sister groups . From previous research [3] and Figures 2 and 3 , these relationships should be tree-like if there were no admixture . For 53 of the possible 280 tests , the Z-score was more extreme than ±2 . 575829 . The most extreme test statistics were f4 ( Wagyu , Mongolian; Simmental , Shorthorn ) = −0 . 003 ( Z-score = −5 . 21 , other rearrangements of these groups had Z-scores of 7 . 32 and 16 . 55 ) and f4 ( Hanwoo , Wagyu; Piedmontese , Shorthorn ) = 0 . 002 ( Z-score = 4 . 90 , other rearrangements of these groups had Z-scores of 21 . 79 and 27 . 77 ) . When K = 20 , Hanwoo appear to have a Mediterranean influence , whereas Wagyu have a northwestern European , including British , influence ( Figure S10 ) . We conclude that there were two waves of European introgression into Far East Asian cattle , first with Mediterranean cattle ( which carried African taurine and indicine alleles ) brought along the Silk Road [29] and later from 1868 to 1918 when Japanese cattle were crossed with British and Northwest European cattle [25] . The second factor that we believe underlies the divergence of African taurine is a high level of wild African auroch [30] , [31] introgression . Principal component ( Figure 1 ) , phylogenetic trees ( Figures 2 and 3 ) , and admixture ( Figure 6 ) analyses all reveal the African taurines as being the most diverged of the taurine populations . Because of this divergence , it has been hypothesized that there was a third domestication of cattle in Africa [32]–[36] . If there was a third domestication , African taurine would be sister to the European and Asian clade . When no migration events were fit in the TreeMix analyses , African cattle were the most diverged of the taurine populations ( Figures 2 and 3 ) , but when admixture was modeled to include 17 migrations , all African cattle , except for East African Shorthorn Zebu and Zebu from Madagascar which have high indicine ancestry , were sister to European cattle and were less diverged than Asian or Anatolian cattle ( Figure 4 ) , thus ruling out a separate domestication . Our phylogenetic network ( Figure 4 ) shows that there was not a third domestication process , rather there was a single origin of domesticated taurine ( Asian , African , and European all share a recent common ancestor denoted by an asterisk in Figure 4 , with Asian cattle sister to the rest of the taurine lineage ) , followed by admixture with an ancestral population in Africa ( migration edge a in Figure 4 , which is consistent across 6 separate TreeMix runs , Figure S4 ) . This ancestral population ( origin of migration edge a in Figure 4 ) was approximately halfway between the common ancestor of indicine and the common ancestor of taurine . We conclude that African taurines received as much as 26% ( estimated as 0 . 263 in the network , p-value<2 . 2e-308 ) of their ancestry from admixture with wild African auroch , with the rest being Fertile Crescent domesticate in origin . Although three other migration edges originate from the branch between indicine and taurine ( such as edge b ) , all of the receiving populations show indicine ancestry in the ADMIXTURE models . But African auroch are extinct and samples were not available for the ADMIXTURE model , thus the admixed auroch ancestry of African taurines cannot specifically be discovered by this model [27] , [37] and African taurine , especially Lagune , are depicted as having a single ancestry without indicine influence ( Figures 5 and 6 , see f3 and f4 statistics reported later ) . Unlike ADMIXTURE , TreeMix can model admixture from an unsampled population by placing a migration edge more basal along a branch of the phylogeny , in this case African auroch . Others have observed distinct patterns of linkage disequilibrium in African taurines , resulting in larger estimates of ancestral effective population size than for either Bos t . taurus or Bos t . indicus breeds [2] consistent with greater levels of admixture from wild aurochs . Just as Near Eastern domesticated pig mitochondrial lineages were replaced by mitochondria from indigenous wild populations [38] , we infer that the divergent T1d African mitochondrial subgroup [39] previously observed originated either from Fertile Crescent domesticates or admixture with wild African auroch . Similar patterns of admixture from wild forebears have been observed in other species [38] , such as pig [40]–[42] , chicken [43] , and corn [14] , and this conclusion represents the most parsimonious explanation of our results . We hypothesize that the auroch introgression in Africa may have been driven by trypanosomiasis resistance in African auroch which may be the source of resistance in modern African taurine populations [44] . Admixture with distant relatives has had an important impact on the immune system of other species , such as human [45] and possibly chicken [46] . More sophisticated demographic models and unbiased whole-genome sequence data will be needed to further test these hypotheses . African cattle also demonstrate a geographical gradient of indicine ancestry [47] . Taurine cattle in western Africa possess from 0% to 19 . 9% indicine ancestry ( Figures 5 and 6 , LAG , ND1 , ND2 , NDAM , BAO , OUL , SOM ) , with an average of 3 . 3% . Moving from west to east and from south to central Africa , the percent of indicine ancestry increases from 22 . 7% to 74 . 1% ( Figures 5 and 6 , ZFU , ZBO , ZMA , BORG , TULI , BOR , SHK , ZEB , ANKW , LAMB , an AFR ) , with an average of 56 . 9% . As we increased values of K to 10 , 15 , and 20 ( Figures S8 , S9 , S10 ) , we revealed two clusters of indicine ancestry possibly resulting from the previously suggested two waves of indicine importation into Africa , the first occurring in the second millennium BC and the second during and after the Islamic conquests [25] , [34] , [48] . The presence of two separate clades of African cattle in Figure 4 also supports the idea of two waves of indicine introgression . Asian cattle breeds were derived from cattle domesticated in the Indian subcontinent or imported from the Fertile Crescent and Europe . Cattle in the north and northeast are primarily of Bos t . taurus ancestry ( Figures 5 and 6; HANW , WAGY , and MG ) , but Hanwoo and Mongolian also have Bos t . indicus ancestry ( Figures 5 , 6 , S9 , and S10 ) . Cattle in Pakistan , India , southern China and Indonesia are predominantly Bos t . indicus ( Figures 5 and 6; ONG , MAD , BRE , HN , ACE , PES , ACH , HAR , BAG , GUZ , SAHW , GBI , CHO , GIR , KAN , THA , RSIN , HIS , LOH , ROJ , DHA , and DAJ ) . Cattle located between these two geographical regions are Bos t . taurus×Bos t . indicus hybrids ( Figures 1 , 4 , 5 , and 6; QC and LX ) . Our results suggest an additional source for increased indicine diversity—admixture with domesticated cattle from other species . In addition to cattle domesticated from aurochs ( Bos primigenius ) , bovids were also domesticated from water buffalo ( Bubalus bubalis ) , yak ( Bos grunniens ) , gaur ( Bos gaurus ) , and banteng ( Bos javanicus ) , represented in our sample by the Bali breed [25] , [49] . We find that the Indonesian Brebes ( BRE ) and Madura ( MAD ) breeds have significant Bos javanicus ( BALI ) ancestry demonstrated by the short branch lengths in Figures 2–4 , shared ancestry with Bali in ADMIXTURE analyses ( light green in Figures S8 , S9 , S10 ) , and significant f3 statistics ( Table S3 ) . The Indonesian Pesisir and Aceh and the Chinese Hainan and Luxi breeds also have Bali ancestry ( migration edge c in Figure 4 , migration edges in Figure S4 , and light green in Figures S8 and S9 ) . Cattle were imported into Europe from the southeast to the northwest . The descendants of Durham Shorthorns ( the ancestral Shorthorn breed [25] ) were the most distinct group of European cattle as they clustered at the extremes of principal component 2 ( lower left hand corner of Figure 1 ) , and they formed a distinct cluster in the ADMIXTURE analyses whenever K was greater than 4 ( Figures S6 , S7 , S8 , S9 , S10 ) . As shown in Figures S6 through S10 , f3 statistics in Table S4 , and from their breed histories [25] , many breeds share ancestry with Shorthorn cattle , including Milking Shorthorn , Beef Shorthorn , Lincoln Red , Maine-Anjou , Belgian Blue , Santa Gertrudis , and Beefmaster . From the previous placement of the American Criollo breeds including Romosinuano , Texas Longhorn , and Corriente , it has been posited that Iberian cattle became admixed as a result of an introgression of cattle from Africa into the local European cattle [3] , [50] , [51] . Our genotyping of individuals from 11 Spanish breeds supported , but clarified , this hypothesis . On average , Spanish cattle had 19 . 3% of African ancestry when K = 3 , with a minimum of 8 . 8% and a maximum of 23 . 4% , which supports previous analyses of mitochondrial DNA [52] , [53] . Migration edge d in the phylogenetic network ( Figure 4 , and consistently seen in Figure S4 ) estimates that Iberian cattle , Texas Longhorn , and Romosinuano derive 7 . 5% of their ancestry from African taurine introgression , similar to the ancestry estimates from the models with larger K values ( Figures S8 , S9 , S10 ) . The Oulmès Zaer ( OUL ) breed from Morocco also shows that cattle were transported from Iberia and France to Africa ( tan and red in Figure S10 , and short branch length in Figure 4 ) . However , the 11 Spanish breeds had no more indicine ancestry than all other European taurine breeds ( essentially none for the majority of breeds , see Figures 5 and 6 ) . Maraichine ( MAR ) , Gascon ( GAS ) , Limousin ( LIM ) , and other breeds from France , and Piedmontese cattle ( PIED ) from northwest Italy have a similar ancestry . These data indicate that the reason that the American Criollo breeds were found to be sister to European cattle in our previous work [3] was because of their higher proportion of indicine ancestry . The 5 sampled American Criollo breeds had , on average , 14 . 7% African ancestry ( minimum of 6 . 2% and maximum of 20 . 4% ) and 8 . 0% indicine ancestry ( minimum of 0 . 6% and maximum of 20 . 3% ) . Other Italian breeds ( MCHI , CHIA , and RMG ) share ancestry with both African taurine and indicine cattle ( Figures 6 , S6 , S7 , S8 ) . This introgression may have come from Anatolian or East African cattle that carried both African taurine and indicine ancestry , which is modeled as migration edge b in Figure 4 . The placement of Italian breeds is not consistent across independent TreeMix runs ( Figure S4 ) , likely due to their complicated history of admixture . We also used f-statistics to explore the evidence for African taurine introgression into Spain and Italy . We did not see any significant f3 statistics , but this test may be underpowered because of the low-level of introgression . With Italian and Spanish breeds as a sister group and African breeds , including Oulmès Zaer , as the other sister group , we see 321 significant tests out of 1 , 911 possible tests . Of these 321 significant tests , 218 contained Oulmès Zaer . We also calculated f4 statistics with the Spanish breeds as sister and the African taurine breeds as sister ( excluding Oulmès Zaer ) . With this setup , out of the possible 675 tests we saw only 1 significant test , f4 ( Berrenda en Negro , Pirenaica;Lagune , N'Dama ( ND2 ) ) = 0 . 0007 , Z-score = 3 . 064 . With Italian cattle as sister and African taurine as sister ( excluding Oulmès Zaer ) , we saw 17 significant tests out of the 90 possible . Patterson et al . [54] defined the f4-ratio as f4 ( A , O; X , C ) /f4 ( A , O; B , C ) , where A and B are a sister group , C is sister to ( A , B ) , X is a mixture of B and C , and O is the outgroup . This ratio estimates the ancestry from B , denoted as , and the ancestry from C , as . We calculated this ratio using Shorthorn as A , Montbeliard as B , Lagune as C , Morucha as X , and Hariana as O . We choose Shorthorn , Montbeliard , Lagune , and Hariana as they appeared the least admixed in the ADMIXTURE analyses . We choose Morucha because it appears as solid red with African ancestry in Figure S10 . This statistic estimated that Morucha is 91 . 23% European ( = 0 . 0180993/0 . 0198386 ) and 8 . 77% African , which is similar to the proportion estimated by TreeMix . The multiple f4 statistics with Italian breeds as sister and African breeds as the opposing sister support African admixture into Italy . The f4-ratio test with Morucha also supports our conclusion of African admixture into Spain . It has recently been concluded that indicine ancestry is a common feature of European cattle genomes [55] . However , our data refute this conclusion . McTavish et al . relied on the Evanno test to arrive at an optimal number of ancestral populations of K = 2 , which masks the fact that there are cattle breeds in Africa with 100% African taurine ancestry ( Figure 6 ) . Although our K = 2 ADMIXTURE results suggested that most African breeds had at least 20% indicine ancestry ( Figure S5 ) , when we increased K to 3 , Lagune ( LAG ) revealed no indicine ancestry , and Baoule ( BAO ) and N'Dama ( NDAM ) possess very little indicine ancestry . If the K = 2 model was correct , we would expect to see numerous significant f3 and f4 tests with Eurasian taurine and indicine as sister groups . Whereas , if the K = 3 model more accurately reflected the heritage of European and African taurines , we would not observe any significant f3 or f4 tests showing admixture of taurine and indicine in the ancestry of African taurine . For the Lagune , Baoule and N'Dama ( NDAM and ND2 ) breeds we found no significant f3 statistics . Among the 225 f4 statistics calculated with NDAM , LAG , BAO , ND2 , SH , and MONT as sisters and BALI , GIR , HAR , SAHW , PES , and ACE as the opposing sister group , only 36 were significantly different from 0 ( Table 1 ) . When ND2 was excluded from the results , only 4 tests were significant ( Table 1 ) , and we have no evidence that the Lagune breed harbors indicine alleles . Thus , we conclude that contrary to the assumptions and conclusions of [55] cattle with pure taurine ancestry do exist in Africa . Further , we conclude that indicine ancestry in European taurine cattle is extremely rare , and that some breeds , especially those prevalent near the Mediterranean , possess African taurine introgression—but with the exception of the Charolais , Marchigiana , Chianina and Romagnola breeds—not African hybrid or African indicine introgression . We concur that Texas Longhorn and other American Criollo breeds possess indicine ancestry , but infer that this introgression occurred after the arrival of Spanish cattle in the New World and likely originated from Brahman cattle ( migration edges e and f in Figure 4 ) . In TreeMix replicates , Texas Longhorn and Romosinuano are either sister to admixed Anatolian breeds or they receive a migration edge that originates near Brahman ( Figure S4 ) . To reiterate , Iberian cattle do not have indicine ancestry , American Criollo breeds originated from exportations from Iberia , Brahman cattle were developed in the United States in the 1880's [25] , American Criollo breeds carry indicine ancestry , and the introgression likely occurred from Brahman cattle . Domestication , exportation , admixture , and breed formation have had tremendous impacts on the variation present within and between cattle breeds . In Asia , Africa , North and South America , cattle breeders have crossbred Bos t . taurus and Bos t . indicus cattle to produce hybrids which were well suited to the environment and endemic production systems . In this study , we clarify the relationships between breeds of cattle worldwide , and present the most accurate cattle “Tree of Life” to date in Figure 4 . We elucidate the complicated history of Asian cattle involving the domestication and subsequent admixture of several bovid species . We provide evidence for admixture between domesticated Fertile Crescent taurine and wild African auroch in Africa to form the extant African taurine breeds . We also observe African taurine content within the genomes of European Mediterranean taurine breeds . The absence of indicine content within the majority of European taurine breeds , but the presence of indicine within three Italian breeds is consistent with two separate introductions , one from the Middle East potentially by the Romans which captured African taurines in which indicine introgression had already occurred and the second from western Africa into Spain which included African taurines with no indicine introgression . It was this second group of cattle which likely radiated from Spain into Southern France and the Alps . The prevalence of admixture further convolutes the cryptic history of cattle domestication .
We used 1 , 543 samples in total , including 234 samples from [3] and 425 samples from [4] , see Table S1 . We selected samples that had fewer than 10% missing genotypes , and for breeds with fewer than 20 genotyped samples , we used all available samples which passed the missing genotype data threshold . When pedigree data were absent for a breed , the 20 samples with the highest genotype call rates were selected . For breeds which had pedigree information , we filtered any animals whose sire or dam was also genotyped . For identified half-siblings , we sampled only the sibling with the highest genotype call rate . After removing genotyped animals known to be closely related , we selected the 20 animals with the highest genotype call rate to represent the breed . All DNA samples were collected in an ethical manner under University of Missouri ACUC approved protocol 7505 . Samples were genotyped with the Illumina BovineSNP50 BeadChip [56] . Autosomal SNPs and a single pseudo-autosomal SNP were analyzed , because the data set from Gautier et al . [4] excluded SNPs located exclusively on the X chromosome . We also filtered all SNPs which mapped to “chromosome unknown” of the UMD3 . 1 assembly [57] . In PLINK [58] , [59] , we removed SNPs with greater than 10% missing genotypes and with minor allele frequencies less than 0 . 0005 ( 1/[2*Number of Samples] = 0 . 000324 , thus the minor allele had to be observed at least once in our data set ) . The average total genotype call rate in the remaining individuals was 0 . 993 . Genotype data were deposited at DRYAD ( doi:10 . 5061/dryad . th092 ) [60] . The sample genotype covariance matrix was decomposed using SMARTPCA , part of EIGENSOFT 4 . 2 [19] . To limit the effects of linkage disequilibrium on the estimation of principal components , for each SNP the residual of a regression on the previous two SNPs was input to the principal component analysis ( see EIGENSOFT POPGEN README ) . TreeMix [20] models the genetic drift at genome-wide polymorphisms to infer relationships between populations . It first estimates a dendrogram of the relationships between sampled populations . Next it compares the covariance structure modeled by this dendrogram to the observed covariance between populations . When populations are more closely related than modeled by a bifurcating tree it suggests that there has been admixture in the history of those populations . TreeMix then adds an edge to the phylogeny , now making it a phylogenetic network . The position and direction of these edges are informative; if an edge originates more basally in the phylogenetic network it indicates that this admixture occurred earlier in time or from a more diverged population . TreeMix was used to create a maximum likelihood phylogeny of the 134 breeds . Because TreeMix was slow to add migration events ( modeled as “edges” ) to the complete data set of 134 breeds , we also analyzed subsets of the data containing considerably fewer breeds . For these subsets , breeds with fewer than 4 samples were removed . To speed up the analysis , we iteratively used the previous graph with m-1 migrations as the starting graph and added one migration edge for a total of m migrations . We rooted the graphs with Bali cattle , used blocks of 1000 SNPs , and used the -se option to calculate standard errors of migration proportions . Migration edges were added until 99 . 8% of the variance in ancestry between populations was explained by the model . We also ensured that the incorporated migration edges were statistically significant . To further evaluate the consistency of migration edges , we ran TreeMix five separate times with -m set to 17 . ADMIXTURE 1 . 21 was used to evaluate ancestry proportions for K ancestral populations [21] . We ran ADMIXTURE with cross-validation for values of K from 1 through 20 to examine patterns of ancestry and admixture in our data set . Map figure was generated in R using rworldmap ( http://cran . r-project . org/web/packages/rworldmap/index . html ) . The f3 and f4 statistics are used to detect correlations in allele frequencies that are not compatible with population evolution following a bifurcating tree; these statistics provide support for admixture in the history of the tested populations [54] , [61] . The THREEPOP program from TreeMix was used to calculate f3 statistics [54] for all possible triplets from the 134 breeds . The FOURPOP program of TreeMix was used to calculate f4 statistics for subsets of the breeds .
|
The DNA of domesticated plants and animals contains information about how species were domesticated , exported , and bred by early farmers . Modern breeds were developed by lengthy and complex processes; however , our use of 134 breeds and new analytical models enabled us to reveal some of the processes that created modern cattle diversity . In Asia , Africa , North and South America , humpless ( Bos t . taurus or taurine ) and humped ( Bos t . indicus or indicine ) cattle were crossbred to produce hybrids adapted to the environment and local production systems . The history of Asian cattle involves the domestication and admixture of several species whereas African taurines arose through the introduction of domesticated Fertile Crescent taurines and their hybridization with wild African aurochs . African taurine genetic background is commonly observed among European Mediterranean breeds . The absence of indicine introgression within most European taurine breeds , but presence within three Italian breeds is consistent with at least two separate migration waves of cattle to Europe , one from the Middle East which captured taurines in which indicine introgression had already occurred and the second from western Africa into Spain with no indicine introgression . This second group seems to have radiated from Spain into the Mediterranean resulting in a cline of African taurine introgression into European taurines .
|
[
"Abstract",
"Introduction",
"Results",
"and",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"phylogenetics",
"animal",
"genetics",
"genetics",
"introgression",
"biology",
"and",
"life",
"sciences",
"animal",
"management",
"animal",
"breeding",
"hybridization",
"evolutionary",
"biology",
"population",
"genetics",
"evolutionary",
"systematics",
"evolutionary",
"processes",
"gene",
"flow",
"population",
"biology",
"agriculture"
] |
2014
|
Worldwide Patterns of Ancestry, Divergence, and Admixture in Domesticated Cattle
|
The life cycle of HPV is tied to the differentiation status of its host cell , with productive replication , late gene expression and virion production restricted to the uppermost layers of the stratified epithelium . HPV DNA is histone-associated , exhibiting a chromatin structure similar to that of the host chromosome . Although HPV chromatin is subject to histone post-translational modifications , how the viral life cycle is epigenetically regulated is not well understood . SETD2 is a histone methyltransferase that places the trimethyl mark on H3K36 ( H3K36me3 ) , a mark of active transcription . Here , we define a role for SETD2 and H3K36me3 in the viral life cycle . We have found that HPV positive cells exhibit increased levels of SETD2 , with SETD2 depletion leading to defects in productive viral replication and splicing of late viral RNAs . Reducing H3K36me3 by overexpression of KDM4A , an H3K36me3 demethylase , or an H3 . 3K36M transgene also blocks productive viral replication , indicating a significant role for this histone modification in facilitating viral processes . H3K36me3 is enriched on the 3’ end of the early region of the high-risk HPV31 genome in a SETD2-dependent manner , suggesting that SETD2 may regulate the viral life cycle through the recruitment of H3K36me3 readers to viral DNA . Intriguingly , we have found that activation of the ATM DNA damage kinase , which is required for productive viral replication , is necessary for the maintenance of H3K36me3 on viral chromatin and for processing of late viral RNAs . Additionally , we have found that the HPV31 E7 protein maintains the increased SETD2 levels in infected cells through an extension of protein half-life . Collectively , our findings highlight the importance of epigenetic modifications in driving the viral life cycle and identify a novel role for E7 as well as the DNA damage response in the regulation of viral processes through epigenetic modifications .
Human papillomaviruses ( HPVs ) are small , circular , double-stranded DNA viruses that have a genome of approximately 8 kilobases . More than 200 types of HPV have been identified and are categorized based on their dominant site of infection; either the cutaneous or mucosal epithelium [1] . Mucosal HPVs are further divided into high and low-risk types based on their potential to induce transformation [2] . The high-risk types , of which there are ~15 , are the causative agents of cervical cancer [3] . In addition , high-risk HPV types are also associated with other anogenital cancers , as well as an increasing number of head and neck cancers [4] . HPV-associated cancers are driven by the two main oncoproteins E6 and E7 , which do not exhibit enzymatic function , but work primarily through protein-protein interactions , affecting cellular signaling pathways to provide a replication-competent environment [5 , 6] . The life cycle of HPV is intimately linked to the differentiation of its host cell , the keratinocyte , and consists of three phases of replication [7 , 8] . HPV infects the actively dividing basal cells of the stratified epithelium upon exposure through a microwound . Upon entry into the nucleus , the viral genome undergoes a transient amplification to 50–100 episomal copies per cell in a process termed establishment replication . In undifferentiated basal cells , the viral genome is maintained at a low copy number by replicating once per cell cycle along with cellular DNA [9 , 10] . In these cells , early viral genes ( E1 , E2 , E6 , E7 ) are expressed at a low level from the early promoter ( p97 for HPV31 ) , which is located upstream of the E6 open reading frame in the upstream regulatory region ( URR ) [11 , 12] . Epithelial differentiation triggers the productive phase of the viral life cycle , which results in activation of the late promoter ( p742 for HPV31 ) and high levels of the replication proteins E1 and E2 to drive viral genome amplification to hundreds to thousands of copies per cell [11–16] . E4 and E5 are also expressed at high levels upon differentiation and contribute to productive replication through mechanisms that are not well understood [17 , 18] . Transcripts encoding the capsid proteins L1 and L2 are only generated upon differentiation , such that virion assembly is restricted to the uppermost layers of the epithelium [19 , 20] . Normally , epithelial differentiation results in an exit from the cell cycle . However , the limited coding capacity of the viral genome renders HPV reliant on cellular factors for replication . HPV supports productive replication by subverting key pathways that regulate host cell replication , in turn maintaining differentiating cells active in the cell cycle [6] . This is accomplished in large part through the E7 protein , which pushes differentiating cells back into the cell cycle through its ability to target members of the retinoblastoma family of tumor suppressors ( pRb , p107 , p130 ) for degradation [21 , 22] . In addition , E7 sustains a replication-competent environment upon differentiation by maintaining activation of the ATM- and ATR-dependent DNA damage response ( DDR ) pathways , which are essential for productive viral replication [23–29] . The HPV genome is histone-associated in infected cells as well as in viral particles [30–32] . The chromosomal organization of HPV genomes is similar to that of cellular chromatin and is thought to be regulated by histone-based modifications [33] . Histone tails are subject to extensive post-translational modifications , including acetylation , phosphorylation , and methylation , which can occur at particular genic regions ( e . g . enhancer , promoter , gene body ) [34] . Histone marks are dynamic and occur as a balance between enzymes that deposit the mark ( writers ) and other enzymes that remove the mark ( erasers ) . Epigenetic readers , which are often part of larger , multi-subunit protein complexes , bind directly to the histone mark through a particular domain , serving as effector proteins [35] . It has become clear that these histone modifications play fundamental roles in most cellular processes that require access to DNA [36 , 37] . During the viral life cycle , the early and late promoters of HPV31 exhibit an active conformation , characterized by acetylated H3 and H4 as well as dimethyl H3K4 , suggesting that viral transcription is coordinated by histone modifications [38] . In support of this , the Tip60 histone acetyltransferase , the SIRT1 deacetylase and the chromatin-binding protein Brd4 have been implicated in viral transcription as well as replication [39–41] . In addition , histone modifications associated with DNA repair ( e . g . γH2AX ) are found on HPV31 genomes [42] . The E6 and E7 oncoproteins are well established to induce epigenetic changes in cellular chromatin by affecting the expression or activity of numerous epigenetic modifiers , including histone acetyltransferases , histone deacetylases , histone methyltransferases and histone demethylases [40 , 43] . However , how E6/E7 modulation of host epigenetic machinery regulates viral chromatin and the impact of these epigenetic changes on the viral life cycle remains largely uncharacterized . SETD2 is a methyltransferase that interacts with the Ser2 phosphorylated C-terminal domain of RNA polymerase II ( RNA pol II ) and places the trimethyl mark on H3K36 ( H3K36me3 ) during transcription elongation [44 , 45] . H3K36me3 is thus a mark of active transcription and increases along gene bodies peaking at the 3’ end [46 , 47] . Through the recruitment of numerous readers of the H3K36me3 mark , SETD2 regulates multiple cellular processes , including modulation of chromatin structure and maintenance of transcription initiation through nucleosome remodeling , as well as alternative splicing through recruitment of splicing factors [48–52] . In addition , SETD2 is associated with DNA replication and repair through the recruitment of factors to H3K36me3 involved in homologous recombination ( HR ) as well as mismatch repair [53–55] . Alternative splicing is a key mechanism by which HPV regulates viral gene expression and ensures maximal protein production from a compact genome [56] . In addition , previous studies from our lab and others have demonstrated that HR repair factors are bound to HPV DNA and are required for productive viral replication [41 , 57] . Since SETD2-regulated cellular processes through H3K36me3 are also central to successful completion of the HPV life cycle , we wanted to determine if SETD2-mediated H3K36me3 contributes to epigenetic regulation of HPV replication . In this study , we have found that high-risk HPV positive cells exhibit high levels of SETD2 protein in an E7-dependent manner that are required for productive replication upon differentiation and also contribute to episomal maintenance in undifferentiated cells . H3K36me3 is enriched on the early region of the HPV31 genome in a SETD2-dependent manner . Interestingly , we have found that ATM kinase activity contributes to maintenance of H3K36me3 on the viral genome , identifying a novel role for DDR activation in epigenetic regulation of the viral life cycle . Depletion of H3K36me3 by SETD2 knockdown or inhibition of ATM blocks productive viral replication and alters splicing of late RNAs , suggesting that the recruitment of H3K36me3 readers to HPV chromatin is critical to the viral life cycle . Intriguingly , we have found that HPV31 increases the stability of SETD2 protein in a manner dependent on E7’s Rb binding domain . These studies not only provide insight into how chromatin modifications affect viral processes , but also identify a novel role for E7 in the epigenetic regulation of the HPV life cycle through increasing levels of the epigenetic modifier SETD2 .
To determine if SETD2 could play a role in the HPV life cycle , we first examined SETD2 levels in HPV positive cells . For these studies , we used human foreskin keratinocytes ( HFKs ) stably transfected with HPV31 ( HFK-31 ) or HPV16 ( HFK-16 ) genomes . We also used CIN612 9E ( CIN612 ) cells , which are derived from a CIN1 cervical lesion and maintain HPV31 genomes episomally [11] . As shown in Fig 1A , HPV31 ( HFK-31 , CIN612 ) as well as HPV16 positive cells ( HFK-16 ) exhibited higher levels of SETD2 protein compared to uninfected HFKs . While we did observe some variation in SETD2 levels based on the HFK background , the levels of SETD2 in the HFK-31 , HFK-16 and CIN612 cells were consistently higher than that found in HFKs . Interestingly , in contrast to substantial differences in protein levels , the transcript levels of SETD2 were similar between the HFKs , HFK-31 and HFK-16 cells ( Fig 1B ) . Surprisingly , the SETD2 mRNA levels were significantly lower in the CIN612 cells compared to the HFKs , despite high levels of protein . These results indicate that SETD2 protein levels are increased post-transcriptionally in HPV positive cells . To determine if the increased levels of SETD2 protein are maintained upon differentiation , we examined SETD2 protein levels in HFKs and CIN612 cells grown in high calcium medium for 48hr and 96hr , which is a commonly used method to induce the productive phase of the viral life cycle . As shown in Fig 1C , although SETD2 levels decreased in CIN612 cells upon differentiation , they were still consistently present at higher levels than in HFKs . Involucrin and keratin 10 ( K10 ) were examined as markers of epithelial differentiation . The HPV oncoprotein E7 plays a crucial role in maintaining higher levels of many cellular factors required for the viral life cycle , several of which require E7’s Rb binding domain [25 , 58] . Furthermore , E7 induces the expression and/or affects the activity of several DNA/chromatin modifying enzymes [40 , 43] . To determine if SETD2 levels are increased in an E7-dependent manner , we examined SETD2 protein levels in HFKs retrovirally transduced and expressing either wild-type HPV31 E7 , or E7 containing a mutation in the Rb binding domain ( ΔLHCYE ) , both of which are stably expressed [59] . As shown in Fig 2A , E7-expressing cells had substantially increased levels of SETD2 protein compared to HFKs , and this phenotype was lost in cells containing the ΔLHCYE mutant . We found that co-expression of E6 with E7 did not alter the levels of SETD2 , indicating that E7 is primarily responsible for the increase in SETD2 ( S1 Fig ) . To determine if E7 is necessary for the elevated levels of SETD2 protein observed in the context of HPV infection , we generated HFKs that maintain either wild type HPV31 genomes or genomes containing the E7 ΔLHCYE Rb binding mutation . We , and others , have shown that HFKs containing ΔLHCYE mutant viral genomes are maintained episomally , but exhibit a defect in productive viral replication as well as reduced episome copy number over time [25 , 59] . As shown in Fig 2B , similar to E7 expression alone , the increase in SETD2 levels we observed in HFK-31 cells compared to HFKs was lost in HFK-31 ΔLHCYE cells . Overall , these studies demonstrate that the increase in SETD2 protein in HPV positive cells occurs in an E7-dependent manner , requiring E7’s Rb binding domain . SETD2 is normally present at low levels due to rapid turnover by proteasome-dependent degradation by the SPOP ubiquitin ligase [60] . Previous studies from our lab demonstrated that E7 increases the protein half-life of several DNA repair factors required for viral replication [25] . To determine if E7 increases SETD2 at the level of protein stability , we examined the half-life of SETD2 in HFKs and HFK-31 cells using cycloheximide to block protein synthesis . As shown in Fig 2C , while the half-life of SETD2 protein in HFKs was approximately 3hr , the half-life was extended to greater than 8hrs ( the longest time point measured ) in HFK-31 cells . The increase in SETD2 protein half-life was lost in HFK-31 ΔLHCYE cells , indicating that E7 increases the protein stability of SETD2 in a manner dependent on its Rb binding domain . CIN612 cells exhibited a similar increase in SETD2 protein half-life as the lab-generated HFK-31 lines ( Fig 2D ) . Taken together , these results demonstrate that E7 , through its Rb binding domain , post-transcriptionally regulates the levels of SETD2 in HPV positive cells through an increase in protein stability . To determine if the increased SETD2 levels are important for the HPV life cycle , we examined the impact of SETD2 depletion on viral replication in undifferentiated and differentiated cells using small hairpin RNAs ( shRNA ) . CIN612 cells were transduced with lentiviruses expressing a control shRNA ( shScram ) or two different SETD2-specific shRNAs ( shSETD2 #1 or shSETD2 #2 ) . 72 hours post-transduction , undifferentiated cells were either harvested ( T0 ) , or differentiated in high calcium medium for 72hr . SETD2 is the sole methyltransferase that places the trimethyl mark on H3K36me3 [45] , and as expected SETD2 knockdown resulted in a global decrease in H3K36me3 ( Fig 3A ) . In addition , SETD2 depletion resulted in a dose-dependent decrease in episomal copy number in undifferentiated cells that correlated with the efficiency of SETD2 knockdown , with shRNA #1 and #2 decreasing episome copy number by ~12% and ~60% , respectively . SETD2 knockdown with both shRNAs resulted in a block in productive viral replication upon differentiation ( Fig 3A ) . Importantly , levels of the differentiation specific markers involucrin and K10 were not affected by loss of SETD2 , indicating that the defect in viral genome amplification was not an indirect effect of blocking cellular differentiation . In addition , using SETD2 shRNA #2 , we found that SETD2 depletion minimally altered the levels of cellular factors involved in cell cycle regulation in undifferentiated or differentiated cells , including cyclin A and cyclin E ( S-phase cyclins ) and CDK2 ( cyclin-dependent kinase 2 ) , mitotic cyclin B and CDK1 as well as the Cdc25c phosphatase ( S2A Fig ) . We also observed minimal impact on levels of the cellular replication protein RPA32 ( S2A Fig ) . SETD2 knockdown also did not substantially affect cell number compared to the scramble control 72hr post-transduction with lentivirus particles ( S2B Fig ) . These results suggest that the defect in viral replication observed upon SETD2 knockdown is not due to alterations in cell cycle control . To confirm the importance of SETD2 in productive replication , we utilized suspension in methylcellulose; a commonly used method in the HPV field to induce epithelial differentiation and activate late viral events . Again , we found that transient knockdown of SETD2 using shRNA #2 resulted in reduction in episome copy number in undifferentiated cells ( T0 ) ( S3 Fig ) . Similar to calcium-induced differentiation , SETD2 knockdown resulted in a block in productive viral replication upon differentiation in methylcellulose , without affecting the levels of the differentiation-specific markers involucrin and K10 ( S3 Fig ) . As a final means to examine the importance of SETD2 in viral replication , we utilized a CRISPR/Cas9 genomic editing approach to knockdown SETD2 in CIN612 cells . We designed two single guide RNAs ( sgRNA ) and used a lentiviral system to establish a heterogenous population of cells exhibiting depleted SETD2 , as previously described [61] . These experiments were performed ten days to three weeks after selection when all living cells were puromycin resistant . As shown in Fig 3B , sgRNA #1 and sgRNA #2 resulted in a partial depletion of SETD2 ( ~70% knockdown ) compared to the non-targeting control . Similar to shRNA-mediated knockdown of SETD2 , depletion of SETD2 using guide RNAs blocked productive replication upon differentiation ( Fig 3B ) . In addition , sgRNA #2 resulted in a significant decrease ( ~45% ) in episome copy number in undifferentiated cells . Taken together , these data demonstrate that SETD2 is required for productive viral replication and may also contribute to maintenance of episomal copy number in undifferentiated cells . Since SETD2 facilitates cellular processes through recruitment of readers to H3K36me3 , we next determined if SETD2 is active on viral chromatin . Chromatin immunoprecipitation ( ChIP ) for H3K36me3 as well as H3 . 1 was performed on chromatin harvested from CIN612 cells that were undifferentiated ( T0 ) or differentiated in high calcium medium for 72hr . Using 17 primer pairs across the HPV31 genome ( S1 Table ) , we found by quantitative PCR that H3K36me3 progressively increases across the HPV31 genome , peaking at the 3’ end of the early region , centered over the E2 , E4 and E5 open reading frames ( ORF ) ( Fig 4 ) . In contrast , lower levels of the H3K36me3 modification was found in the URR , the E6 and E7 ORFs as well as the early ( p97 ) and late ( p742 ) promoters . Increased H3K36me3 was not due to increased nucleosome density , as H3 . 1 was fairly constant across the viral genome ( Fig 4 ) . Interestingly , the placement of H3K36me3 on viral chromatin was similar in cells undergoing maintenance replication in undifferentiated cells or productive replication upon differentiation . To determine if SETD2 is required for the maintenance of the H3K36me3 mark on the HPV31 genome , we performed ChIP for H3K36me3 as well as H3 . 1 on chromatin harvested from undifferentiated or differentiated CIN612 cells transiently transduced with either the control shRNA ( shScram ) or the SETD2 shRNA #2 . qPCR was performed using seven primer sets that amplify regions of the HPV31 genome corresponding with either low H3K36me3 ( #3 , #5 , #12 , #15 ) or high H3K36me3 ( #7 , #8 , #9 ) ( Fig 5A , S1 Table ) . We found that SETD2 knockdown resulted in a significant decrease in H3K36me3 across the HPV31 genome in undifferentiated as well as differentiated cells at almost every region examined ( Fig 5B ) . In contrast , SETD2 knockdown had minimal effect on levels H3 . 1 ( Fig 5C ) . The results indicate H3K36me3 enrichment on viral chromatin requires SETD2 activity . Our finding that SETD2 is necessary for HPV31 replication suggests that the H3K36me3 modification may also be required . To examine this , we utilized two methods to reduce H3K36me3 in a SETD2-independent manner: ( 1 ) expression of a dominant negative H3 . 3K36 containing a K36M ( lysine to methionine ) mutation ( H3 . 3K36M ) and ( 2 ) expression of the KDM4A demethylase . H3 . 3K36M cannot be trimethylated but results in a global loss of H3K36me3 without affecting other histone methylations [62] . KDM4A specifically removes the trimethyl mark from H3K36 as well as H3K9 [63] . CIN612 cells were transduced with lentivirus to express either wild-type H3 . 3 or the mutant H3 . 3K36M transgene . After 72hr , CIN612 cells were harvested as a T0 ( undifferentiated ) or differentiated in high calcium medium for 72hr . As previously shown , expression of the H3 . 3K36M mutant resulted in a global decrease in H3K36me3 compared to the wild-type H3 . 3 control ( Fig 6A ) [62] . Similar to knockdown of SETD2 , we found that expression of H3 . 3K36M resulted in approximately a 70% reduction in episomal copies in undifferentiated cells and blocked productive viral replication upon differentiation , while cells expressing the wild-type H3 . 3 exhibited a replication phenotype similar to untreated ( UT ) cells , with a moderate effect on productive replication ( Fig 6A ) . To examine the effect of KDM4A overexpression on viral replication , CIN612 cells were stably transduced with either lentivirus harboring empty vector or FLAG-KDM4A . Cells were harvested either prior to or after inducing differentiation in high calcium medium for 72hr . As shown in Fig 6B , CIN612 cells expressing FLAG-KDM4A had lower levels of H3K36me3 compared to the control , as previously demonstrated [53 , 63] . In addition , KDM4A overexpression resulted in a defect in viral genome amplification upon differentiation ( Fig 6B ) . Collectively , our studies using SETD2 depletion , expression of the H3 . 3K36M mutant as well as KDM4A , which share the common feature of reducing H3K36me3 , indicate that H3K36me3 is required for viral replication . Furthermore , given that H3K36me3 is enriched on the HPV31 genome , our data suggest that SETD2 may regulate viral processes through the recruitment of cellular readers to the H3K36me3 mark . Transcription of HPV31 mRNAs is regulated by the early promoter ( p97 ) in undifferentiated cells and by the late promoter ( p742 ) upon differentiation ( Fig 7A ) . HPV genes are transcribed as polycistronic transcripts that are alternatively spliced to yield individual gene products [56] . Alternative splicing is thus a key control mechanism of HPV gene expression and is accomplished through multiple splice donor and splice acceptor sites ( Fig 7A ) . SETD2-mediated H3K36me3 provides a docking site for chromatin adapter proteins ( e . g . MRG15 , p52 ) that in turn recruit splicing factors to drive splice site selection [49 , 52] . We have found that H3K36me3 is enriched over the most commonly used 3’ splice acceptor at 3295 ( SA3295 ) , located at the 5’ end of the E4 exon , the splice donor at 3590 ( SD3590 ) , located at the 3’ end of the E4 exon , the putative p3320 promoter located in the E4 ORF as well as the early polyadenylation site located downstream of the E5 ORF . Use of SA3295 produces transcripts encoding E6 , E7 , E4 and E5 as well as the capsid proteins L1 and L2 [6 , 19 , 20 , 64] , while SD3590 is utilized for the production of L1 RNAs upon differentiation [19 , 20] . To determine if SETD2 activity is important for viral splicing , we extracted RNA from undifferentiated or differentiated CIN612 cells transiently transduced with either control shRNA or SETD2 shRNA #2 . We first determined if splicing in the late region is affected by SETD2 knockdown . To ensure detection of low-level transcripts , we performed end-point PCR ( 35 cycles ) using a 5’ primer that anneals at nucleotide 766 in the E7 open reading frame ( E7F ) that is upstream of the splice donor site SD877 , and a 3’ primer that anneals at nucleotide 6595 in the L1 open reading frame ( L1R ) , downstream of SA3295 and SA5552 . As shown in Fig 7B , amplification of RNA from CIN612 cells containing the control shRNA resulted in two major products upon differentiation , one of which was also present in undifferentiated cells . Some minor products were also detected . The two major splice products of 1 . 5kb and 1 . 2kb were identified by sequencing and shown to be two alternatively spliced L1 RNAs , spliced at 877^3295 and 3590^5552 ( E1^E4^L1 RNA ) ( L1a ) and 877^5552 ( E1^L1 RNA ) ( L1b ) , respectively . Upon SETD2 knockdown , levels of the E1^E4^L1 RNA ( L1a ) were substantially reduced ( Fig 7B ) . Furthermore , we found that the ratio of the two L1 splice variants ( L1a/L1b ) was altered upon SETD2 knockdown in differentiating cells , suggesting that the H3K36me3 modification is necessary for efficient expression of the E1^E4^L1 mRNA . A similar change in the L1 splice variant ratio was observed using semi-quantitative RT-PCR ( S4A Fig ) . Furthermore , using the E4F/L1R primer pair to PCR amplify across the E4^L1 splice junction ( 3590^5552 ) , we found that SETD2 knockdown resulted in a substantial decrease in generation of this spliced product ( S4A Fig ) , suggesting that the presence of the H3K36me3 mark over SD3590 may be required for efficient use of this splice site upon differentiation . To determine if splicing to SA3295 is also affected by SETD2 knockdown , we first utilized a 5’ primer that anneals at nucleotide ( nt ) 121 ( 121F ) , upstream of the first splice donor ( SD ) site ( nt 210 ) , and a 3’ primer that anneals at nt 3452 ( E4R ) that is downstream of the four splice acceptor ( SA ) sites in the early region of HPV31 . Amplification of RNA from CIN612 cells containing either the control or SETD2 shRNA #2 resulted in two major products in undifferentiated and differentiated cells ( Fig 7C ) . The two splice products of 0 . 7kb and 0 . 25kb were identified by sequencing and shown to be spliced at 210^413 and 877^3295 ( E6* , E7 , E1^E4 RNA ) and 210^3295 ( E6^E4 RNA ) , respectively . As shown in Fig 7C , both splice products were readily detected upon SETD2 knockdown in undifferentiated cells , with a slight decrease observed upon differentiation , suggesting that splicing to SA3295 is not severely compromised . To include transcripts produced from the late promoter ( p742 ) , which is located in the E7 ORF , we utilized the E7F primer ( nt 766 ) and a reverse primer located in the E4 exon ( E4R , nt 3452 ) to amplify across the E1^E4 splice junction ( 877^3295 ) . In contrast to splicing across the E4^L1 junction , we found that there was not a defect in splicing across the E1^E4 junction upon SETD2 knockdown , with spliced E1^E4 RNAs detected in undifferentiated and differentiated cells ( Fig 7C ) . Similarly , using a 5’ primer that anneals at nt 1270 and the 3’ E4R primer , we found that splicing still occurred across the 1296^3295 junction to generate the E8^E2C RNA upon SETD2 depletion ( S4A Fig ) . In addition , using the E7F and E2R ( nt 2807 ) primer pair we observed minimal effect of SETD2 knockdown on splicing across the 877^2646 junction to generate the spliced E2 product in either undifferentiated or differentiated cells ( S4B Fig ) . While SETD2 depletion did prevent the differentiation-dependent increase in levels of spliced E1^E4 , E8^E2C and E2 , this could stem from the inability of HPV to productively replicate in the absence of SETD2 activity , resulting in fewer templates for transcription . The decrease in E1^E4^L1 transcripts also likely contributes to the reduction in E1^E4 observed upon SETD2 knockdown upon differentiation . Not surprisingly , the levels of E5 , which is present on polycistronic transcripts containing E4 , were also lower upon differentiation in SETD2 knockdown cells compared to scramble control ( S4B Fig ) . Further characterization of splicing in the early region revealed minimal impact of SETD2 knockdown on the relative levels of unspliced E6E7 transcripts produced from the early promoter p97 using the 121F/295R primer pair ( Fig 7C ) . Overall , these results indicate that SETD2-mediated H3K36me3 on viral chromatin does not markedly affect splicing in the early region , but does influence splice site selection for the production of late L1 RNAs , which may occur through the recruitment of splicing factors . ATM has been shown to inhibit the demethylase KDM2A , which removes di-methyl groups from H3K36 , and also negatively impacts KDM4A levels through proteasome-dependent degradation [65 , 66] . Interestingly , a recent study using integrated subgenomic HPV16 reporters demonstrated that exogenously-induced DNA damage leads to production of spliced late L1 mRNAs , specifically the E1^E4^L1RNA , in an ATM-dependent manner [67] . ATM activation is required for productive replication of HPV31 [27] , though whether ATM contributes to the viral life cycle through epigenetic modifications on viral chromatin is unknown . To determine if ATM activity is required for maintenance of H3K36me3 on viral chromatin , we performed ChIP on chromatin harvested from undifferentiated ( T0 ) and differentiated ( 72hr Ca ) CIN612 cells that were treated with DMSO or 10uM of the ATM inhibitor KU55933 ( Fig 8A ) . qPCR was performed using primer pairs to areas of the HPV31 genome exhibiting high and low levels of H3K36me3 . The location of the primers is shown in Fig 5A and the sequences are listed in S1 Table . Similar to SETD2 knockdown , we found that ATM inhibition results in a significant decrease in H3K36me3 across most regions of the HPV31 genome ( Fig 8A and 8B ) , without affecting levels of H3 . 1 ( S5 Fig ) . Using the same primer pairs described above , we found that ATM inhibition , similar to that of SETD2 knockdown , resulted in decreased levels of the E1^E4^L1 RNA ( L1a ) upon differentiation and altered the ratio of L1a to L1b ( Fig 8C and S6A Fig ) . Furthermore , inhibition of ATM activity resulted in decreased splicing across the E4^L1 junction upon differentiation ( S6A Fig ) . In contrast , splicing in the early region was minimally affected , including the generation of the E6* , E7 , E1^E4 and E6^E4 RNAs ( Fig 8D ) , splicing across the 877^3295 junction ( E1^E4 ) ( Fig 8D ) and the 1296^3295 junction ( E8^E2C ) ( S6A Fig ) as well as generation of the spliced E2 product ( 877^2646 ) ( S6B Fig ) . In addition , the relative levels of unspliced E6E7 transcripts from p97 were minimally affected by ATM inhibition ( Fid 8D ) . While ATM inhibition did result in decreased levels of E1^E4 , E8^E2C , and E5 upon differentiation compared to the DMSO control , this again likely reflects the defect in productive viral replication . The decreased levels of E1^E4 may also be due to the defect in generating the E1^E4^L1 spliced product . Interestingly , in contrast to SETD2 knockdown , we found that inhibition of ATM activity did not result in a global loss of H3K36me3 as detected by western blot analysis ( Fig 8E ) , suggesting that ATM’s affects on H3K36me3 may be largely restricted to viral chromatin . Importantly , these studies indicate that ATM activation is important for maintenance of H3K36me3 on viral chromatin and may regulate viral processes through the recruitment of H3K36me3 readers .
In this study , we demonstrate that the HPV life cycle is epigenetically regulated by a cellular methyltransferase and the H3K36me3 mark . Our data demonstrate that SETD2 is active on HPV31 DNA and that placement of the H3K36me3 mark on viral chromatin is necessary to support viral replication and promote processing of late viral RNAs . In addition , our studies identify a novel role for the ATM DNA damage kinase in the epigenetic regulation of viral processes by regulating the levels of H3K36me3 on viral DNA . Furthermore , we have found that E7 increases SETD2 levels post-transcriptionally in a manner dependent on its Rb binding domain . These studies identify a mechanism by which HPV manipulates epigenetic pathways for viral processes and also reveal an important role for E7 in the epigenetic regulation of the viral life cycle . The distribution of H3K36me3 on the HPV31 genome resembles that of a cellular gene body: progressively increasing across the early region and peaking at the 3’ end . A previous study using subgenomic HPV16 reporter plasmids reported a similar increase in H3K36me3 over the early polyadenylation site ( pAE ) , located just downstream of E5 [68] . This result , coupled with our finding that HPV16 positive cells have increased levels of SETD2 , suggests that regulation of viral processes by SETD2 through H3K36me3 may extend to other HPV types . H3K36me3 is associated with active transcription , and the distribution of H3K36me3 could potentially result from transcription from the early promoter ( p97 ) and the generation of transcripts polyadenylated at the pAE [64] . However , upon differentiation we did not observe a similar increase in H3K36me3 over the late region when the late promoter is active and transcripts are polyadenylated at the late site ( pAL ) , located downstream of L1 . These results suggest that increased transcription through the late region is insufficient to increase H3K36me3 . However , several studies have shown that splicing is a determinant of H3K36me3 placement , with formation of the spliceosome enhancing the recruitment of SETD2 to RNA pol II [69 , 70] . As mentioned , the H3K36me3 peak encompasses the most commonly used 3’ splice site on the HPV genome ( SA3295 for HPV31 ) located at the 5’ end of the E4 ORF , which is important for the generation of transcripts encoding E6 , E7 , E4 and E5 as well as the capsid proteins L1 and L2 [11 , 19 , 20] . The frequent use of SA3295 may increase the recruitment of SETD2 to RNA pol II leading to enhanced placement of H3K36me3 at the end of the early region in both undifferentiated and differentiated cells . In addition , H3K36me3 is enriched over the 5’ splice site SD3590 , which is located at the 3’ end of E4 and is activated upon differentiation to allow for expression of L1 [64] . These results suggest that splicing , rather than transcription , may be the major driver in H3K36me3 placement on HPV chromatin . While splicing may affect the placement of H3K36me3 on the HPV genome , our results indicate that H3K36me3 is important for splice site selection on late viral RNAs . Depletion of SETD2 as well as inhibition of ATM kinase activity reduced H3K36me3 on the viral genome and resulted in exclusion of the E4 ORF from a late E1^E4^L1 mRNA upon differentiation . While SETD2 knockdown and ATM inhibition had limited effect on the use of the 3’ splice site SA3295 , we found that splicing from SD3950 at the 3’ end of E4 to SA5552 at the 5’ end of L1 ( E4^L1 ) was remarkably reduced . Furthermore , the ratio of E1^E4^L1 to E1^L1 was substantially altered by SETD2 knockdown and ATM inhibition , indicating that H3K36me3 on viral DNA influences splice site selection to ensure efficient E4^L1 splicing upon differentiation . H3K36me3 facilitates splice site selection by creating a docking site for the chromatin adapter proteins MRG15 and Psip1/p52 , which in turn recruit splicing factors [49 , 52] . MRG15 recruits PTB ( polypyrimidine tract binding protein ) to alternatively spliced exons , while Psip1/p52 recruits SRSF1 , a member of the splicing enhancing serine-arginine rich ( SR ) protein family . Interestingly , both PTB and SRSF1 have been reported to play a role in the regulation of HPV splicing [56] . In addition , SRSF3 , which also binds Psip1/p52 , has recently been shown to regulate expression of HPV31 and HPV16 L1 mRNAs , specifically expression of the E4^L1 mRNA [49 , 71] . Furthermore , PTB has been shown to play a role in the expression of L1 by relieving suppression of the late splice sites SD3950 [72] . PTB may be recruited to H3K36me3 on viral chromatin through MRG15 specifically during the late stages of the viral life cycle to promote L1 expression . Future studies will determine if SRSF1 , SRSF3 and PTB recruitment occurs in an H3K36me3-dependent manner . ATM has been recently shown to affect the splicing of late mRNAs expressed from integrated HPV16 subgenomic reporter plasmids , which is postulated to occur through phosphorylation of BRCA1 and the recruitment of splicing factors to viral RNAs [67] . Our studies suggest that ATM activity regulates splicing of late L1 RNAs expressed from HPV31 episomes through the maintenance of H3K36me3 on viral chromatin . Whether this also involves BRCA1 is currently unknown , though we have previously shown that BRCA1 is required for productive viral replication [73] . The mechanism by which ATM regulates H3K36me3 on viral chromatin is currently unclear . In response to DNA damage , ATM inhibits the activity of the KDM2A demethylase , which removes dimethyl groups from H3K36 and could in turn prevent trimethylation by SETD2 . Intriguingly , ATM activity has also been shown to promote the proteasome-dependent degradation of KDM4A [65] , which we have found blocks productive replication upon overexpression . KDM4A stability is regulated by the RNF8 and RNF168 ubiquitin ligases , which are recruited to sites of DNA damage in an ATM-dependent manner [65] . While multiple effectors of ATM have been shown to localize to sites of HPV replication [23] , whether RNF8 and RN168 are also recruited to HPV genomes has not been examined . The regulation of H3K36me3 by ATM will be a focus of future investigation . In this study , we have found that H3K36me3 depletion by SETD2 knockdown , expression of the H3 . 3K36M mutant or the KDM4A demethylase blocks productive viral replication . Importantly , previous studies have shown that SETD2 knockdown , as well as H3 . 3K36M and KDM4A expression do not affect cell cycle progression [53] , and we have found similar results , indicating that the effect of H3K36me3 deficiency on viral replication is not due to an inability to remain active in the cell cycle . How SETD2 contributes to HPV31 replication is unclear , but could occur through the recruitment of several readers of the H3K36me3 mark [74] . SETD2 promotes nucleosome reassembly behind RNA pol II through recruitment of the FACT ( Facilitates Chromatin Transcription ) complex to H3K36me3 [50] . Loss of H3K36me3 leads to reduced FACT loading and a decrease in nucleosome density , which could impact viral replication as well as transcription . In addition , previous studies from our lab and others demonstrated that DNA repair processes regulated by SETD2-mediated H3K36me3 are required for HPV replication [23] . SETD2 promotes error-free homologous recombination ( HR ) repair within transcriptionally active regions in response to double-strand DNA breaks ( DSB ) as well as replication stress through the constitutive binding of LEDGF and PALB2 , respectively , to H3K36me3 [53 , 55] . In response to DSBs , LEDGF allows for the recruitment of the CtIP nuclease , resulting in DSB resection and formation of single-strand DNA that is bound by the HR recombinase Rad51 [53] . PALB2 associates with H3K36me3 through MRG15 and facilitates swift linkage of the HR repair factors BRCA2 and BRCA1 , which recruit Rad51 to protect and/or repair nearby replication forks suffering from replication stress [55] . In addition to BRCA1 , we have found that Rad51 is also required for productive viral replication [57] . Rad51 binds to HPV31 genomes , but whether this occurs in an H3K36me3-dependent manner and protects viral genomes from DNA damage and replication stress is unclear . ATM activation promotes HR repair and is required for productive viral replication [27 , 75] . Future studies will determine if SETD2 and ATM-dependent regulation of H3K36me3 on viral chromatin ensures the recruitment of HR factors to viral genomes . Our studies indicate that E7 increases SETD2 levels post-transcriptionally through an increase in protein half-life , in a process that requires E7’s Rb binding domain . How E7 increases the SETD2 protein half-life is currently unknown . SETD2 stability is regulated by the ubiquitin ligase SPOP , which forms an ubiquitin E3 ligase complex with cullin 3 ( CUL3 ) and ring-box 1 ( ROC1/RBX1 ) [60] . E7 proteins bind CUL3 and may in turn block formation of the SPOP/CUL3 complex [76] . SETD2 contains a conserved C-terminal SRI domain for interaction with RNA pol II and a SET domain responsible for catalyzing substrate methylation [77] . Mutations in the SET domain of SETD2 and the conserved SRI domain of yeast Set2 have been shown to influence protein stability through loss of binding to histone H3 and RNA pol II , respectively [78 , 79] . E7 significantly affects host gene expression and may protect SETD2 from degradation by promoting transcription and interaction with RNA pol II or histone H3 [80] . Interestingly , recent studies have shown that SPOP and SETD2 are targeted for degradation by the APC/Ccdh1 complex in G1 [81 , 82] . HPV16 E7 interferes with the degradation of APC/Ccdh1 substrates , though whether this occurs in a manner dependent on the Rb binding domain is unclear [83] . E7 may therefore have multiple mechanisms to counteract SETD2 protein degradation . SETD2 is typically associated with tumor suppressor activity and is often mutated in several cancer types [77] . In contrast , we have found that high-risk HPV positive cells exhibit increased SETD2 levels that are necessary for viral replication and viral RNA processing . Our studies suggest that the E7-mediated increase in SETD2 plays a role in regulating viral processes through placement of the H3K36me3 mark on viral chromatin . Understanding how SETD2 regulates the viral life cycle through binding of alternative H3K36me3 effector proteins to viral chromatin will be a focus of future investigation . However , an equally important area will be to understand the impact of increased SETD2 levels on the cellular landscape and how these epigenetic alterations may promote viral persistence , a major risk factor for the development of cancer . Our results provide further support that the effects of E7 and E6 on host epigenetic modifiers have consequences on viral chromatin and viral processes . A more complete understanding of how HPV manipulates epigenetic pathways to facilitate the viral life cycle will provide further insight into how these pathways can be exploited for the treatment of HPV-associated diseases .
Human foreskin keratinocytes ( HFKs ) were isolated from neonatal foreskin tissue and were maintained in Dermalife keratinocyte growth medium ( KGM; Lifeline Cell Technology ) , as described previously . [84] . HFKs containing wild-type HPV31 , HFK-31 ΔLHYCE and HFK-16 were generated by co-transfecting HFKs with re-circularized HPV31 ( or ΔLHYCE ) genome ( excised from pBR-322min ) or HPV16 ( excised from p1203 PML2d ) along with a pSV2-neo resistance plasmid using PolyJet transfection reagent ( Signagen Laboratories ) , followed by eight days of selection in G418 ( Sigma ) , as described [59] . E7-expressing HFKs were made using pLXSN encoding wild-type HPV31 E7 and the E7 ΔLHCYE mutant , along with G418 ( Sigma ) selection , as described [41] . surviving populations were pooled and expanded for analysis . All lines were cultured in E medium supplemented with 5 ng/ml mouse epidermal growth factor ( EGF; BD Biosciences ) in the presence of mitomycin C-treated NIH J2 3T3 murine fibroblast feeder cells ( obtained from Lou Laimins , Northwestern University ) [85] . When necessary , J2 feeders were removed from HPV-positive cells by incubation with 1 mM EDTA in phosphate-buffered saline ( PBS ) . Human embryonic kidney 293T cells and NIH 3T3-derived PT67 cells were obtained from Lou Laimins ( Northwestern University ) and cultured in Dulbecco’s modified Eagle’s medium ( DMEM; Life Technologies ) supplemented with 10% bovine growth serum ( BGS; ThermoFisher Scientific ) . Calcium-induced differentiation was performed as previously described [27] . Where indicated , cells were treated with 10uM of the ATM kinase inhibitor KU55933 ( Tocris Bioscience ) or DMSO for the designated amount of time . Human keratinocytes were isolated from discarded , de-identified foreskins obtained from routine circumcisions performed at UNC hospital in Chapel Hill , NC . Because these tissues were anonymous ( therefore not requiring consent ) , not collected specifically for our research , and would have been discarded otherwise , the Office of Human Research Ethics at UNC-Chapel Hill has determined that our use of human foreskin keratinocytes does not constitute human subjects research as defined under federal regulations [45 CFR 46 . 102 ( d or f ) and 21 CFR 56 . 102 ( c ) ( e ) ( l ) ] and does not require IRB approval . Study number 18–0950 . CIN612 cells were suspended in 1 . 5% methylcellulose as previously described [86] . Cells were harvested as an undifferentiated sample ( T0 ) or after 48hr differentiation in methylcellulose . At each time point , cells were harvested for DNA and protein as described below . pBR322min containing the HPV31 and–HPV31 ΔLHCYE genome has been described [87] . p1203 PML2D containing HPV16 was obtained from Addgene . pLXSN-HPV31 E7 , pLXSN-HPV31 E6/E7 and pLXSN-HPV31 ΔLHCYE plasmids were described previously [59] . The KDM4A ORF was cloned into pLenti-C-Myc-DDK-IRES-Puro ( Origene ) within the SgfI/MluI cloning site using T4 DNA ligase ( Invitrogen ) . The H3 . 3 and H3 . 3K36M lentiviral plasmids were kind gifts from Dr . David Allis and were previously described [62] . Pre-validated lentiviral shRNAs specific to SETD2 were purchased from Sigma Aldrich ( #1 TRCN0000237839 , #2 TRCN0000237837 ) . A scramble control shRNA cloned into the pLKO . 1-puro background was obtained from the UNC Lentiviral Core Facility ( Chapel Hill , NC ) . Lentivirus particles expressing control or SETD2 shRNAs , FLAG-KDM4A , H3 . 3 or H3 . 3K36M were prepared as previously described [88] . Each plasmid was transiently transfected into 293T cells , along with Gag-Pol-Tet-Rev plasmid DNA and vesicular stomatitis virus G ( VSV-G ) plasmid DNA using polyethyleneimine ( PEI ) ( VWR ) . Supernatants containing lentivirus were harvested 72 hours post-transfection . CIN612 cells were transduced with 5ml viral supernatant in the presence of 4 ug/ml Polybrene ( Sigma ) . The medium was changed 24hr post-transduction and the cells were allowed to grow for another 48hr . At this point , cells were harvested or differentiated , or stably selected in puromycin . Two different sgRNAs targeting SETD2 were designed using CRISPRdirect ( https://crispr . dbcls . jp/ ) and cloned into pLentiCRISPRv2 ( kind gift from Dr . Gaurov Gupta ) , as described previously [89] . Lentivirus particles expressing each sgRNA or vector control were generated as previously described [88] . CIN612 cells were transduced with 5ml viral supernatant in the presence of 4 ug/ml Polybrene ( Sigma ) . 48hr post-transduction , cells were selected in medium containing 5 μg/ml of puromycin for six days . The target guide sequences were sgRNA #1 ACTCTGATCGTCGCTACCATAGG and sgRNA #2 GAGAGAGGACGCGCTATTCTCGG . The primers utilized to generate the SETD2 sgRNAs are listed in the S1 Table . SETD2 depletion was confirmed by western blotting . Lysate harvesting and western blot analysis was performed as previously described [57] . To quantify the levels of keratin 10 ( K10 ) , insoluble cell pellets obtained from RIPA-SDS lysis were resuspended in 8 M urea , 10% β-mercaptoethanol , 2 mM PMSF , and incubated at room temperature while rotating for 30 min , as described previously [90] . The insoluble debris was removed by centrifugation at 14 , 000 rpm at 4’C . Primary antibodies used: anti-SETD2 ( Kind gift from Dr . Brian Strahl , Epicypher ) , anti-H3K36me3 , anti-cyclin B , anti-CDK2 ( Abcam ) , anti-H3 . 1 ( Active Motif ) , p84 ( GeneTex ) , anti-Involucrin , anti-keratin 10 , anti-CDC25C , anti-cyclin A , anti-cyclin E and anti-GAPDH ( Santa Cruz ) , anti-CDK1 , anti-RPA32 ( Bethyl laboratories ) and anti-Flag ( Sigma ) . Secondary antibodies used were horseradish peroxidase ( HRP ) -conjugated anti-rabbit ( Cell Signaling Technology ) and HRP-conjugated anti-mouse ( GE Life Sciences ) . Western blots were developed using Clarity Western ECL blotting substrate ( Bio-Rad ) . Images were captured on either autoradiography film or Biorad ChemidocMP imaging system . Blots were analyzed with Biorad Imagelab 5 . 0 software . DNA isolation and Southern blotting were performed as previously described [91] . Briefly , cells were harvested in DNA lysis buffer ( 400mM NaCl , 10mM Tis pH 7 . 5 and 10mM EDTA ) , then lysed by the addition of 30uL 20% SDS . Samples were subsequently treated with 15ul of 10mg/mL proteinase K overnight at 37°C . DNA was extracted using phenol chloroform , followed by ethanol precipitation in the presence of sodium acetate . 5ug of DNA was digested with either BamHI ( New England Biolabs ) ( does not cut the genome ) , or HindIII ( New England Biolabs ) ( cuts the genome once ) . DNAs were resolved on a 0 . 8% agarose gel for 15 h at 40 V and were then transferred to a positively charged nylon membrane ( Immobilon-Ny+; EMD Millipore ) . The DNA was fixed to the membrane via UV irradiation and then hybridized to a radioactive DNA probe consisting of 32P-labeled linearized HPV31 genome . HFKs , HFK-31 , HFK-31 ΔLHCYE and CIN612 cells were grown in 10cm dishes until approximately 80% confluent . Cells were treated with 50 ug/ml cycloheximide and whole cell lysates were harvested at the indicated time points . Western blot analysis was performed using 50ug of total protein as described previously [25] . Westerns were digitally imaged using the Bio-Rad Chemidoc MP system , and densitometry was performed with the Biorad ImageLab 5 . 0 software . Total RNA was extracted using RNA Stat 60 ( Tel-Test ) followed by DNase digestion ( Promega ) and was reverse transcribed using the Superscript Vilo reverse transcription kit ( Invitrogen ) . To analyze spliced transcripts , 50 ng of cDNA was amplified using 500 nM primers and Q5 High-Fidelity 2X Master Mix ( NEB ) . The primers are listed in the S1 Table . PCR products were separated on a 1% percent gel and post-stained with ethidium bromide . Where indicated , PCR products were gel purified and splice junctions were determined by sequencing . 50ng cDNA was analyzed in triplicate reactions for three different experiments using an Applied Biosystems QuantStudio 6 Flex real-time PCR thermal cycler ( Life Technologies ) . qPCR was performed using 375 nM primers and SsoAdvanced Universal SYBR Supermix ( Bio-Rad ) . Reaction profiles were setup as follows: initial denature at 95°C for 10 min followed by 40 cycles of 95°C for 15 sec , 63°C for 1 min , 72°C for 30 sec . Melt curves were subsequently performed to ensure proper primer annealing . Relative transcript levels were determined using the threshold cycle method ( ΔΔCT ) with GAPDH as an endogenous control gene . SETD2 primers are listed in the S1 Table . ChIP assays were on chromatin prepared from CIN612 cells using 2 μg of anti-H3K36me3 ( Abcam ) , 10 μg of anti-H3 . 1 ( Active Motif ) or normal rabbit or mouse IgG , as previously described [57] . Input and immunoprecipitated DNA was quantified in triplicate using the HPV31 qPCR primers listed in the S1 Table . qPCR was performed using an Applied Biosystems QuantStudio 6 Flex real time PCR thermal cycler ( Life Technologies ) . Reaction profiles were setup as follows: initial denature at 95°C for 10 min followed by 40 cycles of 95°C for 15 sec , 63°C for 1 min , 72°C for 30 sec . Melt curves were subsequently performed to ensure proper primer annealing .
|
High-risk HPVs are associated with multiple human cancers , most notably cervical cancer . Understanding mechanisms by which HPV co-opts cellular pathways to replicate could identify potential therapeutic targets . The HPV genome is associated with histones in a manner similar to that of cellular DNA , but how histone modifications influence the viral life cycle is not well understood . Here , we demonstrate that high-risk HPV positive cells exhibit elevated levels of SETD2 in an E7-dependent manner . SETD2 places the trimethyl mark on H3K36 ( H3K36me3 ) and we have found that SETD2 as well as H3K36me3 are necessary for productive viral replication and splicing of late viral RNAs upon epithelial differentiation . H3K36me3 is present on the HPV31 genome suggesting that SETD2 regulates viral processes through the recruitment of effector proteins to H3K36me3 on viral DNA . In addition , we have found that HPV31 maintains H3K36me3 on viral chromatin and regulates splicing of viral RNAs through activation of the ATM DNA damage response , which is also required for required productive viral replication . Overall , these findings advance our understanding of how the viral life cycle is epigenetically regulated and identify a novel role for the DNA damage response in facilitating viral processes .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"medicine",
"and",
"health",
"sciences",
"pathology",
"and",
"laboratory",
"medicine",
"pathogens",
"microbiology",
"cell",
"differentiation",
"viruses",
"developmental",
"biology",
"dna",
"viruses",
"hpv-31",
"epigenetics",
"molecular",
"biology",
"techniques",
"chromatin",
"research",
"and",
"analysis",
"methods",
"papillomaviruses",
"artificial",
"gene",
"amplification",
"and",
"extension",
"chromosome",
"biology",
"medical",
"microbiology",
"gene",
"expression",
"microbial",
"pathogens",
"life",
"cycles",
"rna",
"splicing",
"viral",
"replication",
"molecular",
"biology",
"biochemistry",
"rna",
"rna",
"processing",
"cell",
"biology",
"nucleic",
"acids",
"polymerase",
"chain",
"reaction",
"virology",
"viral",
"pathogens",
"genetics",
"biology",
"and",
"life",
"sciences",
"organisms"
] |
2018
|
SETD2-dependent H3K36me3 plays a critical role in epigenetic regulation of the HPV31 life cycle
|
The plantation sector in Sri Lanka lags behind the rest of the country in terms of living conditions and health . In 1992 , a sector-wide survey of children aged 3–12 years and women of reproductive age showed >90% prevalence of soil-transmitted helminth infections . Biannual mass de-worming targeting children aged 3–18 years started in 1994 and was continued until 2005 . The present study was carried out to assess the status of infection four years after cessation of mass de-worming . A school-based cross-sectional survey was carried out . Faecal samples from approximately 20 children from each of 114 schools in five districts were examined using the modified Kato-Katz technique . Data regarding the school , the child's family and household sanitation were recorded after inspection of schools and households . Multivariate analysis was carried out using logistic regression , to identify risk factors for infection . Faecal samples were obtained from 1890 children . In 4/5 districts , >20% were infected with one or more helminth species . Overall combined prevalence was 29 . 0%; 11 . 6% had infections of moderate-heavy intensity . The commonest infection was Ascaris lumbricoides , present in all five districts , as was Trichuris trichiura . Hookworm was not detected in two districts . Multivariate analysis identified low altitude and maternal under-education as risk factors for all three infections . Poor household sanitation was identified as a risk factor for A . lumbricoides and hookworm , but not T . trichiura infections . The results indicate that regular mass de-worming of plantation sector children should be resumed along with more emphasis on better sanitation and health education . They show that even after 10 years of mass chemotherapy , prevalence can bounce back after cessation of preventive chemotherapy , if the initial force of transmission is strong and other long-term control measures are not concomitantly implemented .
Sri Lanka has already achieved several Millennium Development Goals such as universal primary school enrolment , gender parity in school enrolment , and is on track to achieve the desired reduction in under-five and infant mortality . Poverty has markedly reduced in the urban and rural sectors between 1990 and 2006 , from 26 . 1% to 15 . 2% [1] . However , not all parts of the country have benefited equally from these gains . In the plantation sector , which has a resident population of about 939 , 000 living and working on tea and rubber plantations , there is widespread child malnutrition , maternal mortality rates are exceptionally high and poverty has increased by over 50% in the same period [1] . Soil-transmitted helminth ( STH ) infections are well-known accompaniments of poverty in the developing world [2] . A survey that covered the entire plantation sector in Sri Lanka in 1992 found over 90% of children to be infected [3] . A major de-worming program , offering bi-annual treatment with 500 mg mebendazole to children aged 3–18 years , was launched in 1994 [4] . However , this programme was discontinued after about ten years , due to lack of funds , without proper reassessment of the epidemiological situation . A national survey of the health of school children , carried out in 2003 , found only 6 . 9% to be infected with any of the three major STH infections [5] . This is well below the threshold of 20% prevalence recommended by the WHO for implementation of mass de-worming of school children in endemic areas [6] . Despite ten years of mass de-worming between 1994 and 2005 in the plantation sector , the extremely high prevalence of STH infection at the outset of the de-worming programme , and the actual increase in poverty in the intervening period , suggested that the results of the national survey should not be extrapolated to this sector . This study was designed to estimate the current prevalence and intensity of STH infections among primary school children in the plantation sector and to describe the factors associated with infection , in order to provide data for rational design and targeting of school-based health and nutrition programmes .
Approval was obtained from the Ethics Review Committee of the Faculty of Medicine , University of Kelaniya ( application no P103/08/2009 ) . After selection of children in each school , their homes were visited and their parent ( s ) interviewed in order to obtain written , informed consent for the child's participation in the study , and to obtain information on the socio-economic status of the family . Only children whose parents gave consent were included in the study . All children were offered treatment with mebendazole 500 mg at the end of the study . This was a school-based , cross-sectional survey . The 2007 School Census of the Ministry of Education identifies 830 ‘plantation sector schools’ , almost all of which provide instruction in the Tamil language ( also referred to as ‘Tamil-medium schools’ ) . These schools are divided into two categories: inside or outside a plantation . Tamil-medium schools located inside the plantations formed the setting of the study . The study was performed in five administrative districts , namely Nuwara Eliya , Badulla , Kegalle , Ratnapura and Kandy . These five districts are centrally located in the southern half of Sri Lanka ( Figure 1 ) . The study region occupies a total land area of 11 , 500 km2 representing 17 . 5% of the total land area of Sri Lanka . The 2010 estimated midyear population indicates the region's population as about 3 . 5 million people , representing 19 . 1% of the country's population . The study population was selected from among students who were registered in grade 4 classes in 2009 , in schools lying within plantations in the five selected districts . Grade 4 classes were selected because this study was part of another larger study which assessed children's cognitive abilities with a test specifically designed for this grade . The schools within this study area include nearly 90% of the primary school population in plantation schools . Schools with less than 60 students were excluded since it was unlikely that they would have the 20 or 27 students in Grade 4 needed for a cluster . Schools with more than 400 students were also excluded since they draw students from a wide geographical area and different society other than from the plantation sector . The eligible schools were considered as clusters for sampling . Sampling was carried out based on cluster sampling procedure [7] . This involved two stages: firstly , to select the primary units ( i . e . schools ) within each district , and secondly , to select the elementary sampling units ( classes ) within the primary units . In Nuwara Eliya district , the sample size calculation was based on the need to determine if the overall prevalence of infection was over the 20% threshold recommended by the WHO for introduction of mass deworming , and the need to ensure that an actual prevalence of 15% or less would not be misclassified as 20% . On this basis , the required sample size for the district was 196; with an anticipated response rate of 70% , this increased to 280 . As cluster sampling was to be used , a design effect of 1 . 5 was also taken into consideration , thus giving a final sample size of 420 . For logistical reasons , it was decided that the sample of 420 children would be drawn from16 schools from among 297 schools in the district . In each selected school , 27 children were selected using the lottery method . If a school had less than 27 children in Grade 4 , the remainder was made up by selection of children in other grades . In the other four districts , children included in the study were part of another larger study on the impact of school-based de-worming and iron supplementation on the cognitive abilities of school children . Thus , while the desired minimum sample size per district remained the same as for Nuwara Eliya , the size of a cluster was limited to 20 children per school because of constraints with regard to assessment of cognitive ability . An overall sample size of 100 schools with 20 children per school was selected; if the school had less than 20 children in Grade 4 , all children registered in the grade were selected , but numbers were not made up from other classes . Data collection in the field was conducted in July 2009 in Nuwara Eliya District; and between September and November 2009 in the other four districts . Faecal examination for helminth infections was carried out by experienced medical laboratory technicians . A single smear per faecal sample was examined using the modified Kato-Katz technique as recommended by WHO [8] . Samples were left to clear for 20–60 minutes before reading . Kits for the Kato-Katz test were purchased from Vestergaard-Frandsen , India . According to the manufacturers' instructions , the egg count recorded in each positive sample was multiplied by a factor of 24 to obtain the number of eggs per gram ( epg ) faeces . Intensity of infection was categorized using cut-off values recommended by WHO [9] . Accordingly , A . lumbricoides , T . trichiura and hookworm infections with egg counts less than 5000 , 1000 , and 2000 epg faeces respectively were categorized as light; those with egg counts of 5000–49999 , 1000–9999 , and 2000–3999 epg faeces respectively were categorized as moderately heavy; and those with egg counts of 50000 or over , 10000 or over , and 4000 epg faeces or over , respectively , were categorized as heavy infections . Faecal samples were examined at hospital/medical faculty laboratories located at a central , easily accessible point within each district ( see Figure 1 ) , in order to minimize sample transport time . Data regarding the school , the child's family and household sanitation were recorded in pre-tested questionnaires by trained , Tamil-speaking data collectors . The schools were also inspected for availability of water on tap in the school premises for flushing the toilet or for washing hands; availability of soap and water in the latrines; and cleanliness of the latrines . Adequacy in the number of toilets was calculated according to the standards adopted by the Education Ministry , which works out to roughly one latrine for every 50 children ( Ministry of Education circular no 2007/21 of 8 October 2007 , available at http://www . moe . gov . lk/web/images/stories/circulars/2007-21e . pdf ) . A household latrine facility and usage score was constructed based on direct observation of the latrine by data collectors who were trained in this task . Their observations were recorded on a pre-tested form with categorical variables . This included the type of latrine used by the family ( water-sealed/ pit/ bucket/ none ) , the availability of water on tap , soap , a door and a roof in the latrine , and the observation of faecal contamination in and around the latrine at time of inspection . Poor sanitation resulted in a low score ( 0 being the worst ) while good sanitation ( water-sealed latrine with roof and door , water on tap in latrine and soap available , no faecal contamination observed ) was awarded a high score ( maximum of 83 ) . A cut-off of 73 was used for the logistic regression since this score divided the study population for which scores were available , into approximately equal halves . The score was not calculated for Nuwara Eliya district , as data collection was carried out by a different team , and records were not consistent . The geographical coordinates and altitude of every school was recorded using a hand held global positioning system ( GPS ) monitors ( Trimble Juno SB from Trimble Navigation , Sunnyvale , CA USA; Garmin eTrex® H from Garmin International , Inc . Olathe , KS USA; or Magellan eXplorist 500 from Magellan Navigation Inc San Dimas , CA USA ) . Data recorded on GPS monitors were uploaded into a Geographic Information System ( GIS ) ( ArcGIS , ESRI , Redlands , CA USA ) , and associated with attribute data on STH infection prevalence . A predicted prevalence map was created using ordinary kriging based on the prevalence of infection in schools . Kriging is a class of geo-statistical techniques used for optimal spatial prediction . They are statistically unbiased techniques ( i . e . , on average , the predicted value and the true value coincide ) that minimize prediction mean-squared error , and provide a measure of uncertainty or variability in the predicted values . Kriging uses the semivariogram , a function of the distance and direction separating two locations , to quantify the spatial autocorrelation in the data . The semivariogram is then used to define the weights that determine the contribution of each data point to the prediction of new values at the unsampled locations . Ordinary kriging is a linear predictor , meaning that prediction at any location is obtained as a weighted average of neighbouring data . It assumes a constant but unknown mean , and estimates the mean value as a constant in the searching neighbourhood [10] . Double entry was carried out using a database developed in EpiInfo Version 3 . 5 . 1 and data cleaned where there were discordant entries . Data analysis was done on SPSS Version 16 . 0 and R version 2 . 10 . Multivariate analysis was carried out using logistic regression for each helminth infection , with the outcome variable categorized as presence or absence of infection . Taking into account the multistage sampling method used in the study design , the explanatory variables were grouped into district level ( district , type of plantation and altitude ) , school level ( duration since the last school medical inspection , total number of children in the school and availability of telephone at school ) , family level ( mother's education beyond primary school , father's education beyond primary school , latrine score and family size ) and individual level ( sex , worm treatment within the preceding six months , use of footwear and condition of fingernails on inspection ) variables . The assumptions for a logistic regression were checked with all the models . Variables were added to the model in blocks by level in the following order: district , school , family and individual level . The significant variables in a block ( p≤0 . 05 ) were retained in the model and the variables from the next lower level were added to the model . Continuous variables and categorical variables with more than two categories were dichotomized prior to being used as predictor variables . The selection of a model from among the models having the same or different number of explanatory variables was done based on the stability of model using the Akaike Information Criterion , degree of multi-co-linearity using variance inflation factor , significance of the odds ratio and relevance of the variable considering the existing knowledge . Father's education and mother's education had co-linearity and only mother's education was retained in the model considering its greater relevance to childcare .
Faecal samples were obtained from 1 , 890 children ( compliance rate of 93% ) . Overall , 1072 children ( 52 . 5% of the study population ) consisted of boys . Table 1 shows the prevalence of infection by district , while Figure 1 shows the proportion of children found infected in each study school . In four of the five districts surveyed , over 20% of the children were infected with one or more STH . The overall combined prevalence of infection was 29 . 0% . The commonest infection was A . lumbricoides , which was present in all five districts , as was T . trichiura . Hookworm infections ( usually Necator americanus in Sri Lanka ) were not detected in Nuwara Eliya and Badulla districts . Figure 3 presents a continuous surface map , which was created using ordinary kriging based on the prevalence of any STH infection at each school . Only 20/462 children ( 4 . 3% ) with A . lumbricoides infection had infections of heavy intensity; 40 . 7% had moderately heavy infections; the rest ( 55 . 0% ) had light infections . In contrast , nearly 90% of the T . trichiura infections were of light intensity and the remainder was of moderate intensity; no heavy T . trichiura infections were seen at all . In two of the three districts where hookworm was detected , all infections were of light intensity . In Ratnapura district only , a few heavy infections were seen ( 6 . 3% of infected children ) ; while the large majority were light infections ( 40/48 , 83 . 3% ) . Overall , as shown in Table 2 , 219 children ( 11 . 6% ) had moderate or heavy infections of one or more of the three infections . Overall , 22 . 4% of children ( 423/1884 ) were found to have overgrown , dirty fingernails on inspection during the survey . Among boys , this proportion was 25 . 5% ( 252/990 ) , while amongst girls it was only 18 . 9% ( 166/877 ) . Most of the children wore shoes to school ( 1417/1867 , 75 . 9% ) , while another 21% ( 387/1867 ) wore slippers . Only a very small proportion of children ( 63/1867 , 3 . 4% ) had come barefoot to school . Boys were more likely to be barefoot than girls [46/991 , ( 4 . 6% ) vs 17/876 , ( 1 . 9% ) respectively]; whereas a higher proportion of girls were in shoes [685/876 , ( 78 . 2% ) vs 732/991 ( 73 . 9% ) respectively] or slippers [174/876 ( 19 . 9% ) vs 213/991 ( 21 . 5% ) respectively] when compared with boys ( Pearson chi-square = 11 . 8 , d . f . = 2 , p = 0 . 003 ) . On direct questioning , parents of 56 . 8% of children ( 1067/1880 ) claimed that the child had been de-wormed within the last 6 months , either through the school , or by the parents . The percentage of de-wormed children varied between districts: 74 . 5% ( 277/372 ) of children in Badulla had been de-wormed , whereas in Nuwara Eliya , only 46 . 7% ( 176/377 ) children had been de-wormed . The annual School Medical Inspection ( SMI ) had yet not been conducted in 27 of the 114 schools at the time of the survey . Since anthelmintics are usually administered during the School Medical Inspection , this indicated that the children in these 27 schools were unlikely to have received anthelmintics through school in the last 6 months . The number of students per school latrine ranged from 6 to 209 with a mean of 66 . 9 students per latrine . Overall , only 61/113 schools ( 54 . 0% ) met the Education Ministry norm regarding the number of latrines in relation to student population . When comparing districts , Kandy had the lowest numbers of students per school latrine ( 45 . 3 ) and 18/26 schools ( 69 . 2% ) met the Education Ministry norm , whereas Nuwara Eliya had the highest number of students per school latrine ( 89 . 0 ) and only 25% of schools met the Education Ministry norm . Ratnapura and Badulla Districts had 79 . 7 and 62 . 2 students per school latrine , while 12/23 ( 52 . 2% ) and 14/25 ( 56 . 0% ) of schools respectively , met the Education Ministry norm . Almost all schools ( 98/114 , 86 . 0% ) had water on tap , either from a local water collection or from the public water supply . However , more than a third of schools ( 44/114 , 38 . 6% ) did not have water on tap in the toilets . Furthermore , just two schools in the Ratnapura district and one school in Nuwara Eliya district had soap available for hand-washing in the latrines; none of the other schools had soap for use by students . Information was obtained on the educational attainment of 1516 fathers and 1469 mothers . Amongst the fathers , 11 . 9% had not attended school at all; this proportion was a little higher ( 15 . 2% ) among mothers . Another 40 . 0% of fathers and 42 . 3% of mothers had attended only primary school . With regard to the latrine score in the districts where household latrines were assessed , the score ranged from 0 to 83 , with an overall mean of 64 . 8 ( SD 21 . 3 ) . There was no significant difference in mean scores between districts . The results of the multi-variate analysis identified several risk factors for STH infection in this study population ( see Table 3 ) . At district level , altitude was a significant determinant of all three infections , with lower prevalence of infection at higher altitudes ( >500 m above sea level ) . The effect was most marked for hookworm [Odds Ratio ( OR ) 0 . 08 , 95% Confidence Interval ( CI ) 0 . 04–0 . 17] , and least so for A . lumbricoides infection ( OR 0 . 55 , 95% CI 0 . 40–0 . 76 ) . At school level , the duration since the last School Medical Inspection ( >180 days ) increased the risk of A . lumbricoides infection ( OR 1 . 77 , 95% CI 1 . 29–2 . 46 ) but not infection with either T . trichiura or hookworm . At household level , higher maternal school attainment ( Grade 6 or higher ) significantly reduced risk of all three infections ( A . lumbricoides OR 0 . 40 , 95% CI 0 . 28–0 . 57; T . trichiura OR 0 . 28 , 95% CI 0 . 14–0 . 52; hookworm OR 0 . 47 , 95% CI 0 . 25–0 . 84 ) . Better household sanitation , as reflected by a latrine score of 74 or more , also significantly reduced the risk of A . lumbricoides ( OR 0 . 70 , 95% CI 0 . 51–0 . 97 ) and hookworm infection ( OR 0 . 50 , 95% CI 0 . 28–0 . 86 ) , but not T . trichiura . At individual level , sex was a risk factor for hookworm , with girls being less at risk of infection than boys ( OR 0 . 53 , 95% CI 0 . 30–0 . 93 ) . Although nail hygiene and use of footwear are generally considered protective factors , these variables were not identified as predictors in these models .
These results suggest that STH infections are still highly prevalent among schoolchildren in the plantation sector . At 29 . 0% , the overall prevalence of infection is well above the level at which the WHO recommends introduction of annual mass de-worming . A significant proportion of infected children ( 11 . 6% ) had infections of moderate or heavy intensity , indicating that they are at particular risk of morbidity . It is likely that observed prevalence rates are an underestimate of true prevalence , since only a single faecal sample was examined , using only the Kato-Katz technique for detection of eggs . Use of serial faecal samples and multiple diagnostic techniques have been reported to increase detection rates , especially with regard to hookworm and light infections [11] , [12] . Extremely high prevalence rates , such as those observed in the plantation sector of Sri Lanka in the 1990s , are necessarily accompanied by very high levels of environmental contamination with the eggs and larvae of STH , consequent on pollution of the soil with human faeces . Under favorable conditions , these stages remain viable in soil for many months , sometimes years , and so re-infection after anthelmintic treatment is virtually inevitable , and prevalence rates rebound rapidly [2] . For example , in a study conducted on school children in Pemba Island , Tanzania , when prevalence was well over 90% , infection intensities reached pre-treatment levels by 6 months after treatment with single dose albendazole or mebendazole [13] . Two recent studies from Zanzibar , Tanzania , which also had helminth prevalence rates >90% among school children in the 1990s , found that despite implementation of mass deworming for about 15 years , overall helminth prevalence ( which included infection with A . lumbricoides , hookworm , T . trichiura , Strongyloides stercoralis and Schistosoma haematobium ) remained high . A cross-sectional study was carried out in 2 schools in Unguja Island , Zanzibar , about 6 months after the last school-based anthelmintic treatment was carried out , and compared with data obtained from the same schools in 1994 [14] . Overall prevalence of STH infection had dropped significantly from 98 . 9% in 1994 , but was still high at 59 . 7% . Another study that compared prevalence in one rural and one peri-urban community in Unguja Island found overall helminth prevalences of 73 . 7% and 48 . 9% respectively [15] . It is possible that the greater decline in prevalence seen in the Sri Lankan plantation sector over a similar period of about 15 years , starting from similar prevalences of over 90% , may be attributed to the fact that the mass deworming programme was biannual in Sri Lanka , thus reducing rebound in infection between treatment rounds . Concomitant efforts to improve sanitary facilities in the estates is also likely to have contributed to the less than expected increase in the prevalence of STH infections following cessation of the mass deworming programme . China has also seen a significant decline in the prevalence of soil-transmitted helminth infections , between its first national survey conducted in 1990 and the second in 2003 , with the standardized rates of hookworm , A . lumbricoides and T . trichiura infections declining by 60 . 7% , 71 . 3% and 73 . 6% respectively [16] . This decline has been attributed to the conduct of mass deworming programmes targeting school children in many parts of China , along with the introduction of appropriate health education in schools , improvements in water supplies and sanitation , and the increasing use of chemical fertilizer instead of night soil in agriculture . The Ministry of Health has now adopted a three-pronged approach in its National Control Programme on Important Parasitic Diseases for 2006–2015 . They include large scale deworming with benzimidazoles; providing clean water and adequate sanitation; and health education programmes [17] . The Republic of Korea also had a major problem with STH infections , with an overall prevalence of 84 . 3% in 1971 . Nationwide mass de-worming targeting schoolchildren was conducted twice a year from 1969 to 1995 , and successive quinquennial nationwide surveys showed prevalence to decline gradually , reaching a nadir of 2 . 4% in 1997 [18] . However , the Korean national economy also grew very rapidly during this period , and living standards improved remarkably , along with sanitation and agricultural technology . These factors would have undoubtedly contributed to the reduction in STH prevalence . Since it is apparent that mass deworming must be resumed in the plantation sector of Sri Lanka , the question arises as to how this should be done . There are two ways by which anthelmintics could be delivered to children in the plantation sector . One would be the adoption of a mass de-worming day , when all children in plantation sector school are given anthelmintics on a single , designated day , preceded by adequate preparation in the form of teacher training and advance publicity to raise awareness regarding the issue . This is the usual practice in many countries where school-based de-worming is carried out on a national scale [2] . The other approach would be to strengthen the existing School Medical Inspection programme , where a team from the nearest Medical Officer of Health office visits the school on an annual basis . Traditionally , de-worming of schoolchildren has been a part of the SMI programme . However , de-worming may not occur if the inspection is not conducted for some reason or sufficient stocks of anthelmintics are not available when the SMI is conducted . Also , while all children are examined during the SMI in schools where the total student population is <200 , in the larger schools , only children in Grades 1 , 4 and 7 are inspected and treated . This policy will need to be changed , so that all children in all plantation sector schools are treated regardless of school size . The choice of anthelmintic to be used in mass treatment needs to be given some consideration . At present , single dose mebendazole ( 500 mg ) tablets are given to children during the SMI . However , our results suggest that a history of recent anthelmintic treatment significantly reduced only the risk of infection with A . lumbricoides , and had no long-term impact on T . trichiura or hookworm infections . This is not surprising given that single dose mebendazole is known to be much less effective against hookworm and T . trichiura than against A . lumbricoides [19] . It may be advisable to consider the possibility of using either single-dose albendazole or a 3-day course of mebendazole , particularly in Ratnapura and Kegalle Districts , where T . trichiura and hookworm infections show a relatively high prevalence . It should be noted here that although mebendazole 500 mg was used in much of the mass deworming program in the plantation sector , when a survey carried out 2 years after commencement of the program showed only a very modest decrease in hookworm infections , mebendazole was replaced with albendazole 400 mg in the low altitude plantations where hookworm was most prevalent [20] . A re-evaluation carried out two years after the introduction of albendazole showed that the prevalence of hookworm had reduced sharply on the three plantations that were surveyed [20] . Our findings also strongly suggest that de-worming alone will not eliminate the problem of soil-transmitted helminth infections . After the sector-wide survey conducted in 1994 showed extremely high prevalence rates , biannual mass de-worming of school aged children was carried out for nearly 10 years before its cessation in 2005 . During this period , however , other measures that could also reduce transmission , such as improved sanitation and health education , were not greatly emphasized . Given that transmission of all three infections is dependent upon contamination of soil with human faeces , good sanitation plays an extremely important role in breaking the cycle of transmission and preventing infection , as evidenced by identification of the household latrine score as a significant risk factor . Household sanitation is one of the areas in which the plantation sector in Sri Lanka lags behind the rest of the country . It has been estimated that only 85 . 1% of plantation sector households had sustainable access to improved sanitation in 2006/07 , whereas in the rural and urban sectors , this proportion was 94 . 8% and 91 . 5% respectively [1] . Thus in order to sustain gains achieved by the re-introduction of preventive chemotherapy , other strategies such as information education and communication strategies , behavior change interventions and community led total sanitation projects will need to be pursued more vigorously . Poor maternal education was also identified as a risk factor for infection , as has been found in other studies [21]-[23] . Given that 15 . 2% of mothers in this study had not attended school at all , and 42 . 3% had only completed primary school , this is not surprising . Unfortunately this is part of a vicious cycle where poverty and general lack of education among parents , and mothers in particular , increases the risk of worm infections among children , which in turn has a negative impact on the children's cognitive ability [2] . Poor performance and achievement levels in school results in lack of job opportunities and low income in adult life and so the cycle continues . We also found that the risk of infection with any of the three soil-transmitted helminth infections decreased with increasing altitude . This association was most marked in relation to hookworm , and least evident with A . lumbricoides infections . It is known that free-living infective stages in the environment develop and die at temperature-dependent rates [24] . It is likely that larval survival declines at higher altitudes , where ambient temperatures are much lower than at lower altitudes . Badulla District had the lowest observed prevalence rates among the five districts surveyed . It is possible that this may be partly due to altitude ( since it had schools with the second highest mean altitude ) , but it could also be due to children being given anthelmintics by their parents , which was observed to be highest ( 74 . 5% ) in Badulla . The mean years of school attended by both fathers and mothers were also highest in Badulla ( 6 . 34 and 6 . 04 years respectively ) , but household latrine scores , sanitation in schools , and SMI coverage was not significantly different from other districts . Thus it is possible that the low prevalence in this district is maintained by the high frequency of deworming by parents . In conclusion , the results of this study indicate that there is a need to resume mass de-worming on an annual basis in the plantation sector in Sri Lanka , along with stronger emphasis on other long-term control measures such as improved sanitation and water supplies , and better health education programmes . The results highlight the fact that even after 10 years of biannual mass chemotherapy , when the initial force of transmission is strong and living conditions remain poor , the prevalence of soil-transmitted infections can bounce back after cessation of mass deworming .
|
Mass de-worming of pre-school and school-age children was introduced in Sri Lanka's plantation sector in 1994 after a survey showed that >90% of children and women of reproductive age were infected with intestinal worms . The present study was carried out to assess the status of infection four years after mass de-worming was stopped in 2005 due to lack of funds . Approximately 20 children from each of 114 schools in five districts were examined . Data regarding the school , the child's family and household sanitation were recorded . Faecal samples from 1890 children were examined for worm eggs . In 4/5 districts , >20% were found infected with one or more intestinal worm . Overall , 29 . 0% of children were infected and 11 . 6% of these had moderate–heavy worm burdens . The commonest infection was roundworm , followed by whipworm . Hookworm was not detected in two districts . Statistical analysis identified low altitude , maternal under-education and poor household sanitation as risk factors for infection . The results indicate that when initial infection rates are very high , in the absence of marked improvements in sanitation and health education , even 10 years of biannual mass de-worming may not be enough to prevent resurgence of infection after cessation of mass de-worming .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"medicine",
"infectious",
"diseases",
"public",
"health",
"and",
"epidemiology",
"soil-transmitted",
"helminths",
"trichuriasis",
"epidemiology",
"neglected",
"tropical",
"diseases",
"hookworm",
"ascariasis"
] |
2011
|
Soil-Transmitted Helminth Infections among Plantation Sector Schoolchildren in Sri Lanka: Prevalence after Ten Years of Preventive Chemotherapy
|
Recent experimental and computational studies suggest that linearly correlated sets of parameters ( intrinsic and synaptic properties of neurons ) allow central pattern-generating networks to produce and maintain their rhythmic activity regardless of changing internal and external conditions . To determine the role of correlated conductances in the robust maintenance of functional bursting activity , we used our existing database of half-center oscillator ( HCO ) model instances of the leech heartbeat CPG . From the database , we identified functional activity groups of burster ( isolated neuron ) and half-center oscillator model instances and realistic subgroups of each that showed burst characteristics ( principally period and spike frequency ) similar to the animal . To find linear correlations among the conductance parameters maintaining functional leech bursting activity , we applied Principal Component Analysis ( PCA ) to each of these four groups . PCA identified a set of three maximal conductances ( leak current , Leak; a persistent K current , K2; and of a persistent Na+ current , P ) that correlate linearly for the two groups of burster instances but not for the HCO groups . Visualizations of HCO instances in a reduced space suggested that there might be non-linear relationships between these parameters for these instances . Experimental studies have shown that period is a key attribute influenced by modulatory inputs and temperature variations in heart interneurons . Thus , we explored the sensitivity of period to changes in maximal conductances of Leak , K2 , and P , and we found that for our realistic bursters the effect of these parameters on period could not be assessed because when varied individually bursting activity was not maintained .
Vital adaptive rhythmic behaviors such as breathing and heartbeat in invertebrates are produced by central pattern-generating networks ( CPGs ) . Beside their inherent importance in pacing rhythmic movements , CPGs represent fertile test beds for understanding neuronal network dynamics because of the robustness of their activity patterns even in reduced experimental preparations [1] , [2] , [3] . The combination of the intrinsic electrical properties of the component neurons and their synaptic interactions within a CPG produces their rhythmic activity [1] . To maintain functional rhythmic activity , the CPG neurons and networks must be remarkably robust regardless of changing internal and external conditions . Recent experimental evidence suggests that animals show robust responses to modulation and environmental perturbations ( e . g . , large temperature changes [4] , [5] , [6] ) . Modeling studies have begun to address the mechanisms underlying the robustness in activity type . For example , Goldman et al . [7] tested a model neuron over a wide range of parameters and found that activity type was robust to certain changes in parameters but very sensitive to other changes . Bursting activity in CPGs [1] , [8] , [9] , [10] is characterized by intervals of repetitive spiking separated by intervals of quiescence . Autonomously bursting neurons are common components of CPGs [3] . Half-center oscillators ( HCOs ) , which consist of reciprocally inhibitory neurons ( often autonomous bursters ) , are one of the most prevalent circuit building blocks of CPGs that are thought to assure robust alternating bursting [3] , [10] . Studies of HCOs show that they can display a wide range of bursting activity when the parameters controlling intrinsic membrane properties and synaptic interactions of the neurons are varied [1] , [8] , . The analysis of this basic circuit building block has helped researchers understand how bursting activity is generated and how motor patterns are controlled by the nervous system . One CPG that is particularly well understood controls heartbeat in leeches [2] . The heartbeat period is regulated by a variety of environmental ( e . g . changes in temperature ) and physiologic ( brought on by locomotor movements like swimming ) inputs . When temperature increases [14] , the burst period of the heartbeat CPG decreases [15] . Similarly , when the animals swim , the swim CPG is active and the heartbeat CPG burst period decreases [14] . Therefore period is an important regulated characteristic of this CPG . Recent experimental and modeling analyses of bursting activity indicate that the parameters ( specifically the maximal conductances of specific ion channels ) influencing bursting activity show 3–5 fold variation from animal to animal or model instance to model instance but that there are relationships ( linear or non-linear ) between parameters [16] . For example , electrophysiological and molecular studies in stomatogastric neurons [17] , [18] , [19] found pairwise and four-way linear correlations between the parameters . These studies suggest that the functional activity of a given neuron may reside in the set of parameter correlation rules it maintains rather than in the value of any particular parameter . In addition , such correlations were also found in model solution spaces obtained by parameter space exploration of biologically realistic models [20] , [21] , [22] . Many studies , both experimental [23] , [24] and computational [11] , have provided evidence that linearly correlated sets of parameters ( intrinsic and synaptic properties of neurons ) allow CPG neurons to produce and maintain their rhythmic activity . To establish parameters relationships , some studies have used new visualizations ( e . g . , NDVis , parameterscape ) [25] , [26] , [9] , while others have used mathematical methods ( e . g . , regression , discriminant analysis ) [27] , [28] . However , it is still unclear , how multiple parameters interact to produce and maintain the rhythmic single cell and network activity . Our study here focuses on how intrinsic membrane and synaptic parameters interact to maintain functional bursting activity in HCO and in burster model neurons from the leech heartbeat CPG . For our study , we used the HCO computational model of Hill et al . [29] , which was successfully developed to replicate the electrical activity ( rhythmic alternating bursting of mutually inhibitory neurons ) of the oscillator ( HN ) interneurons of the leech heartbeat CPG under a variety of experimental conditions . This HCO model consists of a two reciprocally inhibitory model HN interneurons , represented as single isopotential electrical compartments with Hodgkin and Huxley [30] type intrinsic and synaptic membrane conductances . Each compartment contains 8 voltage-dependent currents , five inward currents INa - a fast Na+ current , IP - a persistent Na+ current , ICaF - a rapidly inactivating low-threshold Ca current , ICaS - a slowly inactivating low-threshold Ca current , Ih - a hyperpolarization-activated cation current ) and three outward currents ( IK1 - a delayed rectifier-like K current , IK2 - a persistent K current , IKA - a fast transient K current ) . The model has two types of inhibitory synaptic transmission between the two interneurons: graded transmission ( ISynG ) and spike-mediated transmission ( ISynS ) . The maximal conductances ( ) of each of the membrane and synaptic currents and the leak reversal potential ( ELeak ) are free parameters in the model . A comprehensive analysis of parameter relationships in the complete , canonical HCO neuron model presents a computational and theoretic challenge . To systematically explore the parameter space of the HCO and corresponding burster models , in our previous work [31] , we simulated about 10 . 5 million model instances , whose characteristics we recorded into a database named HCO-db [31] , [32] . The simulations were obtained by co-varying a carefully selected set of parameters that single parameter variation analyses showed were crucial in establishing bursting and controlling burst period [29] . These parameters comprise the maximal conductances SynS , SynG , P , K2 , h , CaS and Leak ( varied across of 0% , 25% , 50% , 75% , 100% , 125% , 150% , and 175% of their canonical values ) , and ELeak ( varied across −70 , −65 , −60 , −55 , and −50 mV values ) in all possible combinations . All simulated instances were classified into separate groups showing the same electrical activity . Our HCO-db is a very efficient tool for querying the simulated HCO model instances for finding potential parameter relationships . In this study , we focused only on the four groups of instances from our HCO-db database showing functional leech bursting characteristics , HCOs , realistic HCOs , bursters , and realistic bursters . A HCO instance has two model interneurons each showing bursting activity with at least two bursts in a 40 s time interval , and has the following characteristics: each of its bursts has normal spikes ( See Definitions ) , a small variation of period , a relative phase in the range of ( 0 . 45–0 . 55 ) , and at least one synaptic component present ( either SynS≠0 , or SynG≠0 , or both SynS≠0 and SynG≠0 ) . A realistic HCO instance is a HCO that showed realistic bursting corresponding to that observed in leech oscillator heart interneurons ( period between 5–15 s , average spike frequency between 8–25 Hz , and duty cycle between 50–70% ) . An isolated neuron instance ( isolated neuron ) has two identical interneurons ( though started with different initial conditions , but otherwise identical ) , and no synaptic interaction ( i . e . , SynS = 0 and SynG = 0 ) . A burster instance is an isolated neuron instance for which both neurons had at least two bursts , each with normal spikes , and regular periods ( as defined above for the HCOs ) . Note that burster instances can be thought of as being HCOs with no synaptic connections . A realistic burster instance is a burster instance that showed realistic bursting corresponding to isolated leech oscillator heart interneurons ( period between 5–15 s , and average spike frequency between 8–25 Hz ) . Notice that realistic instances are a subgroup of either HCOs or bursters , and in our discussion here , unless specifically indicated , the HCO and burster groups include their subgroup of realistic instances . We applied Principal Component Analysis ( PCA ) to automatically find the potential existing linear correlations between the parameters maintaining functional activity . The results returned by PCA identified three maximal conductances ( P , K2 , and Leak ) that correlate linearly for the bursters and the realistic bursters , and showed that for the HCOs and realistic HCOs there were no linear correlations between the parameters , but visualizations in a reduced space suggested that non-linear relationships between parameters might exists for these instances . In addition , we found that the bursting activity of the burster instances was very sensitive to variations in P , and Leak and to a lesser extent K2 .
In previous work [31] , our classification algorithm identified 1 , 202 , 139 ( 11 . 6% ) HCO model instances as HCOs and 424 isolated neuron model instances as bursters ( 0 . 26% of the isolated neurons ) . To generate the realistic instance populations , we queried our database , HCO-db [32] with the criteria given in our definitions ( see Methods ) for the realistic HCO instances and realistic burster instances . We recorded the results of these queries into two separate views ( MySQL Views ) to facilitate and speed queries involving these groups in the future . We obtained 99 , 066 instances ( 8 . 2% of the HCOs ) in the group of realistic HCOs , and 307 instances ( 72 . 4% of the bursters ) in the group of realistic burster instances . Out of 424 bursters , 263 produce realistic HCOs ( 1 , 055 instances ) and 419 produce HCOs ( 21 , 303 instances ) when coupled with inhibitory synapses . The number of HCO instances exceeds the number of burster instances because multiple values of synaptic conductance ( SynS or SynG ) give rise to HCO instances for each burster instance . Of the 424 bursters , 307 are realistic bursters , and 238 of these produce 990 realistic HCOs ( instances ) when coupled with inhibitory synapses . That is , 25 bursters ( out of 117 bursters that are not realistic ) produced 65 realistic HCOs . All 307 realistic bursters produce HCOs when coupled with inhibitory synapses ( 16 , 805 HCO instances ) . The vast majority of HCOs in the database are not composed of bursters isolated neurons but of spiking isolated neurons . For example , among 99 , 066 total realistic HCOs , only 1 , 055 ( 1 . 06% ) were composed of bursters ( 263 instances ) ( including 990 realistic HCOs that were composed of realistic bursters ( 238 instances ) ) as stated above , but 94 , 487 ( 95 . 37% ) were composed of spiking isolated neurons ( 12 , 443 instances ) and 3 , 524 ( 3 . 56% ) were composed of neurons classified as either bistable isolated neurons ( 3 , 096 HCOs from 820 isolated neuron instances ) , as irregular isolated neurons ( 368 HCOs from isolated neuron 55 instances ) , as silent isolated neurons ( 58 HCOs from 28 isolated neuron instances ) or as plateau neurons ( 2 HCOs from 2 isolated neuron instances ) . Thus realistic HCOs could also consist of irregular ( irregular bursters or irregular tonic firers ) , silent , or even bistable neurons . Previous work from our group shows that our burster instances have a high propensity for multistability and that mutual inhibition makes multistability much less prevalent [33] . Although we have not tested this idea systematically , we suspect that such multistability is present in the other classes of isolated neuron instances . Figure 1 shows the activity of ten randomly selected instances from each of the four groups of interest , HCOs , realistic HCOs , bursters , and realistic bursters . The figure shows that the instances within each group display different combinations of parameter values despite having similar bursting activity . For example , the two instances shown in turquoise and orange from the realistic HCO group have the same period ( Figure 1A ) and yet their parameters combinations are very different ( colored connected lines Figure 1B ) . In addition , Figure 1 illustrates that the parameter values of the instances within each group have wide ranges for almost every parameter . Some patterns seem to emerge in the parameter ranges that support our four categories of bursting . For example , both HCO and realistic HCO instances are possible without h-current ( h = 0 ) , while both bursters and realistic bursters require at least 50% of the canonical level of h . Figure 2 shows the intrinsic currents and synaptic conductances of two realistic instances randomly chosen from the ones presented in Figure 1 ( the realistic HCO shown in orange and the realistic burster shown in black ) . Both instances replicate ( with respect to period and spike frequency , and for the realistic HCOs duty cycle ) the oscillatory activity of leech HN interneurons , when coupled ( Figure 2A ) and in isolation ( Figure 2B ) . For the realistic HCO instance the leak reversal potential is ELeak = −55 mV ( with minVm = −59 . 5529mV ) , and for the realistic burster is ELeak = −65 mV ( with minVm = −55 . 4954mV ) . During the inhibited phase ( interburst interval ) of the burst cycle in the HCO , the hyperpolarization-activated cation current , Ih , slowly activates , depolarizing the inhibited neuron toward a burst ( escape ) . The persistent Na+ current , IP , also helps in depolarizing the inhibited neuron . The burst is formed by the rapid activation of slowly inactivating low threshold Ca2+ current , ICaS ( ICaF is very small in most instances ) and the inactivation of ICaS leads to its gradual decline leading to a reduced spike frequency and less inhibition of the opposite neuron ( release ) . During the burst IP sustains depolarization and a baseline spike frequency , and the outward currents IK2 and ILeak oppose; during the inhibited phase IP is opposed by the ILeak and the synaptic currents . The balance between IP , ILeak , and IK2 appears crucial for maintaining the excitability of the system and setting the membrane potential about which the system oscillates ( n . b . the −50 mV line ) . The spike currents INa and IK2 ( IKA is small in most instances ) do not directly participate in burst formation but simply provide a baseline of excitability against which the more persistent currents act . When the neurons are isolated ( Figure 2B ) , done in the HCO model by setting the maximal conductances of both synapses to 0 ( SynS = 0 and SynG = 0 ) , basically the same interactions apply except that only ILeak can oppose IP during the interburst interval and hyperpolarize the membrane potential sufficiently to activate Ih ( n . b . −50 mV line ) . The lack of inhibition leads to the apparent requirement for a relatively hyperpolarized ELeak in burster instances [3] . Figures 1 and 2 illustrate the complex interactions of the membrane and synaptic currents and they also suggest potential interactions . For example , for both instances shown in Figure 2 , the maximal conductances , P and K2 ( P = 100% , K2 = 125% for the realistic HCO , and P = 125% , K2 = 100% for the realistic burster ) , have values close to each other , and Leak is large ( 175% and 150% , respectively ) . Is this anecdotal correlation a potential mechanism in the HCO model to maintain realistic ( similar to animal ) bursting activity ? Next , we considered the influence of two parameters at the same time on the activity type of our four groups of interest . To explore visually the relationships existing between two parameters and a group of instances , we plotted the number of instances within a group versus all the possible values for two parameters . Several different methods were tried to make these plots , e . g . , Supplemental Material Figure S2A and S2B , but we settled on the methods of Figures 3 and 4 . These plot the number of instances as the size of each point and the two parameters on the x and y axes and cover all parameter pairs for our realistic groups of HCOs and bursters ( similar but more populated plots for HCOs and bursters were obtained - data not shown ) . Figure 3 shows these two-parameter plots for the realistic HCOs and illustrates the problems with a pair-wise approach because very little structure is apparent in any of the plots , but there were some exceptions . The plot of Leak vs . P shows that there are exclusive zones of high Leak and low P and of low Leak and high P , which do not support realistic HCOs . Similarly the plot of K2 vs . P shows that there is a small exclusive zone of high K2 and zero P and a large exclusive zone of low K2 and high P , which do not support realistic HCOs . In general low P does not support realistic HCOs and middle values of P do support realistic HCOs and the absence of P did not appear to limit the number of realistic HCO instances . Figure 4 shows these two-parameters plots for the realistic bursters and reveal considerably more structure . First a non-zero h was required to produce realistic bursters and the next smallest values supported very few instances . The largest number of instances ( 28 ) was obtained for h = 150% and more negative values of ELeak ( −70 , −65 mV ) . There appears to be a positive correlation between Leak and P required to produce realistic bursters and similarly , but in a looser way , between K2 and P , and between Leak and K2 . Most notably , more positive values of ELeak , and low values of CaS greatly restrict the number of realistic burster instances . These pairwise parameter variation plots suggest potential parameter relationships between more than two parameters for us to investigate in our database using more rigorous mathematical methods to identify all potential linear relationships influencing activity type . To find interactions among the conductance parameters , we applied the Principal Component Analysis method ( PCA ) ( see Methods ) to each of our four groups of interest . For each group of interest , we plotted the percent of variability explained by each principal component ( plots in panel A of Figures 5 and 6 for the realistic groups ) . Then , we identified the main principal components for each group as the smallest number of PCs for which the sum of their variances was greater than 95% . For each of these principal components we plotted the coefficients of their parameters ( panel B of Figures 5 and 6 ) . Tables S1A–D from Suppl . Material show the values of all the coefficients of the main principal components and figures S4A and S4B from Suppl . Material S4 show similar PC plots for the bursters and HCOs . Figure 5A shows the PCA results for the realistic HCOs group . The first six principal components accounted for 96 . 6% of the variance . There was not much difference between the variance values of the main principal components ( the biggest difference of 4 . 6% was between PC 4 and PC 5 ) , which indicates that all main principal components have similar importance for this group . For the first four PCs , the coefficients found for the realistic HCO group ( Figure 5 ) differed from those of the HCO group ( figure not shown ) . For PC 5 and PC 6 , these coefficients were quite similar for the two groups of instances , meaning that these linear combinations of conductances have some similar small influence on both groups . However , since there is no major differences in the amount of variance accounted for by each component , we could not discriminate one or two of these sets of coefficients as being the most influential in the realistic group's activity . Therefore , we hypothesize that there are no linear relationships between any sets of parameters that characterize this group's activity ( same situation for HCOs ) . We also applied PCA analysis to the groups of bursters ( figure not shown ) and realistic bursters ( Figure 6 ) . For both these groups the main principal components and their coefficients were remarkably similar . The first four ( of the six total ) components were main principal components that explained 96 . 88% and 97 . 32% respectively of the total variance . In both groups , there was a large difference between the amount of variance accounted for by the first and second components , which means the first component is the most important for these groups . The first component by itself explained 61 . 2% and 59 . 66% respectively , of the variance , which is very close to two-thirds of the total variance , so this component can be considered as sufficient to characterize the group or , for a more precise characterization , one can use the first three components , which together account for >90 . 3% of the variance . The coefficients of conductances that generate each of the first four PCs for both groups ( Figure 6B ) have the same sign ( positive or negative ) and only small differences in their values . In the first principal component , P , K2 and Leak had large negative coefficients , while ELeak was the only parameter with a positive coefficient albeit small . In the second PC , which accounted for only 17 . 3% and 19 . 7% of the variances respectively for these two groups , CaS dominates followed by K2 with significant negative coefficients , while only h has a significant positive coefficient . However , since for both these groups the first PC is so large , we predicted that P , K2 and Leak , which dominate this component , should all show positive linear correlations in proportions to their weights ( coefficients ) and each should be negatively correlated with ELeak . Next , we explored visually these predictions of the PCA for our groups of interest . We developed a Matlab tool to visualize five characteristics of a data set at once: in the present case three parameters which form a 3D parameter space of the data , the number of instances projected onto each point in this space given by the size of each point , and a fourth parameter which becomes visible when a point in this space is clicked with the mouse button . Each point clicked unveils a pie chart of the 4th parameter showing all instances projected onto this point in the 3D space . The pie chart is split into slices according to the number of values possible for the 4th parameter: 5 slices if the 4th parameter is ELeak and in 8 slices each for the other parameters , with each slice having a different color . If there was no instance projected into the 3D space for a particular value of the 4th parameter , then its corresponding slice was not shown in the pie chart . For a better visualization of the points projected onto the 3D plot , their deepness ( i . e . , their z- axis parameter values ) was color coded with a colormap starting from dark blue shades for closest points ( at 0% values ) to light blue shades for farthest points ( at 175% values ) . Each projected point was depicted by a circle filled with the color according to this mapping . We used this 5D clickable tool to visualize the characteristics of our groups of interest . For each main principal component ( see previous section ) of each group , we selected its three parameters with the biggest coefficients that have the same sign ( either positive or negative ) and then we selected the parameter with the biggest coefficient of the opposite sign . The first 3 parameters selected were the parameters of the 3D space used by our clickable tool , and the last one was used for plotting the pie chart of each point plotted . Figure 7 shows the views obtained by applying our 5D clickable tool to the groups of bursters ( Figure 7 A ) and realistic bursters ( Figure 7 B ) . In both groups , the first principal component was the most important PC for each group ( with variances of approximately 60% ) . The three biggest coefficients of this PC were for P , Leak , and K2 ( negative ) . We used them as the three axes of the 3D data projection space in our visualization tool . The biggest positive coefficient was for ELeak , which we used as the 4th parameter for display as pie charts in the visualization . The group of bursters had 91 points in the 3D space given by ( P , K2 , Leak ) , and the realistic group had 83 points in this space . In both groups , these points cluster around the main diagonal indicating that the amounts of P , K2 , and Leak are positively correlated . Expansions of all the pie charts for both bursters and realistic bursters revealed that the range of permissible ELeak's diminished and ELeak had to be more negative as the values of P , K2 , Leak increased ( from all ELeak values for P = 0 to only ELeak = −0 . 07 for P = 175% ) . Figure 8 shows the views obtained by applying our 5D clickable tool to the groups of HCOs ( Figure 8 A ) and realistic ( Figure 8 B ) HCOs . In both groups , we used the 3D space given by P , K2 , and Leak . For the 4th parameter , we used ELeak . We chose this parameter space to compare the HCO groups with the burster groups in Figure 7 . However , no clustering around the main diagonal , similar to that observed in Figure 7 , emerged in these plots . Similar plots considering different parameters also failed to reveal such a relationship among parameters ( see Supplemental Material Figure S3 for these plots ) . The plots in Figure 8 showed similar 3D shapes for the HCO and realistic HCO instances . This 3D shape has a complicated contour , similar to a wedge , and it is biased in the number of instances toward low Leak . In contrast with burster instances ( Figure 7 ) expansions of all the pie charts of Figure 8B revealed no contraction in the range of permissible ELeak's until the highest ( 175% ) values of P , K2 , Leak were reached , at which the ELeak‘s range was reduced to the four most negative values . In Figure 9 we plotted the subset of instances from each group that have CaS , h , SynS , and SynG for the HCOs , and CaS and h for the bursters at their canonical ( 100% ) values ( see Section Half-center oscillator model ) . For HCOs we had 478 such instances ( Figure 9 A ) , for the realistic HCOs we had 43 instances ( Figure 9 C ) , for the bursters we had 15 ( Figure 9 B ) , and for the realistic bursters 10 ( Figure 9 D ) . The plots of these points in the 3D space of ( P , K2 , Leak ) for burster groups ( Figure 9 B and D ) showed an apparent linear correlation . The subset of the realistic HCOs ( Figure 9 C ) also showed an apparent linear correlation between Leak , P , and K2 . Note that the boundaries of this cluster of points are defined by 50%–150% for P , 25%–175% for Leak , and 75%–175% for K2 . These plots ( Figure 9 B , C , D ) show that , if for a group of instances one restricts the parameter space to an appropriate subset of parameters , then linear correlations between the remaining parameters may emerge . Next we investigated in more detail the apparent linear correlations observed between P , K2 , and Leak for the burster and realistic burster instances . Figure 10 shows a 3D plot of the instances of the burster groups into the space defined by P , K2 , Leak . In this plot burster instances that are not realistic are shown in blue shades and the realistic burster instances are shown in red shades . Each point in this 3D space is depicted by a rectangle in ( P , Leak ) space . The number of instances projected onto each point in the space is shown by the size of a colored rectangle . Color maps show the value of K2 going from 0 ( dark shades ) to 175% ( light shades ) . We fitted a least square fit regression line to each group of points ( 3D Orthogonal Distance Regression ( ODR ) line ) . The 3D ODR line is least distant from all the points and contains the centroid of points , which is the 3D point that has the mean of the points on the three axes . The direction vector that defines the line is given by the coefficients for the first principal component ( PC ) , i . e . , the first column of the coefficients returned by the PCA ( in our case Matlab's princomp function ) when applied to the set of 3D points ( see Supplemental Material Text S1 for detailed mathematics and the coefficients of each equation corresponding to the two ODR lines shown in Figure 10 , see [34] for Methods explanation , and [35] for an example ) . For the realistic bursters the line is shown in Figure 10 in magenta , and for the not realistic bursters in cyan . The two lines did not intersect in the 3D space illustrated . However , their projections in either Leak , K2 plane or the P , K2 plane did intersect with a small angle , but did not intersect in the Leak , P plane . That is , Leak and K2 ( and similarly P and K2 ) have slightly different influences on the two groups of instances , while Leak and P have the same effect on the two groups . Interestingly , the main diagonal of the 3D space of Leak , P , K2 passes though many points from either group . Both fitted lines are on the same side of the main diagonal ( toward higher values of Leak ) , but do not intersect it in the space illustrated . These two lines seem almost parallel in the Leak , P plane , with the magenta line further shifted toward higher values of Leak than the cyan line . The centroids and the two lines give insight into the characteristics of the two groups . The magenta line shows a tendency for the realistic instances to be at the high values on all axes . That is , large values of P , K2 , and Leak produced more realistic instances than the small values . The cyan line shows a tendency for the not realistic bursters to be at the low and middle values on all axes ( in other words , small and moderate values of P , K2 , and Leak produced more not realistic instances than the large values ) . The magenta line has a slightly less steep slope than the cyan line , and a moderate slope ( with angle less than 45o ) if compared with the main diagonal of the 3D space . Precisely , the starting point of the magenta line in the space shown is higher than the starting point of the cyan line , and the end point of the magenta line is lower than the end point of the cyan line . When there was no P and Leak , there were more realistic bursters for larger values of K2 and more not realistic bursters for the lower values of K2 ( 0% , 25% , and 50% ) . The larger the values of P , K2 , and Leak are , the larger the number of realistic instances and the smaller the number of the not realistic instances . We explored the overlap of not realistic and realistic bursters in the P , K2 , and Leak space of Figure 10 . Some of the points ( shaded rectangles ) in this linear relationship ( 50 out of the total of 91 ) represent only not realistic or only realistic instances . Eight points ( light blue circle ) , each corresponding to eight instances , were characterized as only not realistic ( Figure 11 A ) . They were separated from the 42 points ( red circle ) , corresponding to 134 instances , that were characterized as only realistic . However , 41 points ( purple circle ) represented 282 both realistic and not realistic instances . Thus there are a total of 424 instances represented in Figure 10 . Having both realistic and not realistic burster instances projected into the same point in ( P , K2 , Leak ) space means that these instances were influenced by additional parameters ( either CaS , h , or ELeak ) toward being realistic or not . We also explored overlap for the HCOs in in the same 3D space as Figure 10 ( data not shown ) and found that the three maximal conductances of P , K2 , and Leak were not sufficient to characterize uniquely realistic instances ( Figure 11 B ) . 114 points ( light blue circle ) , corresponding to 66 , 962 instances , were characterized as not realistic HCOs ( i . e . , they are HCO instances that do not satisfy the necessary criteria to be also characterized as realistic instances ) . 243 points ( purple circle ) included both realistic and not realistic HCO instances . Having both realistic and not realistic HCO instances projected into the same point in ( P , K2 , Leak ) space suggests that additional parameters are needed to separate these two types of instances ( either CaS , h , SynS , SynG , or ELeak ) . Thus , for HCOs there is no point in the 3D space that represents only realistic instances ( Figure 11 B ) , which means that these instances are characterized by more than the parameters which define the 3D space . We then analyzed how the three parameters ( P , K2 , Leak ) were distributed in the groups of bursters and HCOs . Figure 12 A shows the distributions of the realistic and of the not realistic bursters on each of the three axes of the 3D space given by P , K2 , and Leak ( Figure 10 ) . We performed a two-sample Kolmogorov-Smirnov test on the two distributions obtained on each axis . Each test rejected the null hypothesis that the two distributions were from the same continuous distribution at the 5% significance level ( h = 1 in all cases , pLeak = 0 . 0014 , kLeak = 0 . 875 , pP = 0 . 0014 , kP = 0 . 875 , pK2 = 0 . 0098 , kK2 = 0 . 75 ) . Having different distributions for the two groups indicates that the realistic instances showed parameter distributions which can be used to separate them from the not realistic instances . First , both groups show almost the same number of instances if any of P , K2 , or Leak is equal to 0 . Then , the distributions diverge . On the P axis , the number of not realistic instances decreased and stayed below 18 with an increase in P whereas for P = 0 both groups have a large number of instances . The distribution of the realistic instances on the P axis , showed no clear tendency with a decrease in the number of instances for P = 25% followed by an increase at P = 100% then again a decrease with a minimum at P = 175% . On the Leak axis , the two distributions seem to have the same tendencies as the distributions on the P axis , but here the peak in number of realistic instances occurred for Leak = 125% . On the K2 axis , the distribution of the not realistic bursters showed a peak for K2 = 25% , then the distribution showed a continuous decrease in the number of instances with increasing K2 . On this axis , the realistic distribution had a peak at K2 = 75% , followed by a slight decrease with increasing K2 . Summing up , the largest number of only not realistic bursters occurred for small values of these conductances ( 0–25% ) and the largest number of realistic bursters occurred for moderate to large values of these conductances ( 50–150% ) . Figure 12 B shows the distributions of the realistic and not realistic HCO instances on each of the three axes of the 3D space given by P , K2 , and Leak . Similar to burster instances shown above , the two-sample Kolmogorov-Smirnov test applied to these two distributions rejected the null hypothesis ( p>0 . 05 ) that they are from the same continuous distribution ( h = 1 in all cases , pLeak = 0 . 00015562 , kLeak = 1 , pP = 0 . 00015562 , kP = 1 , pK2 = 0 . 0014 , kK2 = 0 . 875 ) . On the P axis , both distributions showed a peak in their number of instances at P = 75% . On the K2 axis , the distribution of the not realistic HCO instances showed a continuous increase in number of instances with increasing K2 . The distribution of the realistic HCO instances showed a peak in the number of instances at K2 = 150% . On the Leak axis , the distribution of the not realistic HCO instances showed a continuous decrease in number of instances with increasing Leak . The distribution of the realistic HCO instances showed a peak in the number of instances at K2 = 25% , then a continuous decrease in the number of instances with increasing Leak . Summing up , the largest number of realistic HCO instances occurred for moderate values of P ( 50–125% ) and large values of K2 ( canonical and above , 100–175% ) . Leak above 125% reduced the number of realistic HCO instances . To assess the sensitivity of period to variation of Leak , K2 , and P , we queried the HCO-db database to build up the families existing within each group of interest . We define a family as being a sub-set ( of a group ) of instances that have all the same parameter values except one ( e . g . , all realistic bursters that vary only by P constitute a family ) . Note that one can partition a group into families according to the number of members in the family . Surprisingly , for the groups of bursters and realistic bursters , all the instances of each group were part of families of one member only when P was varied , which means that activity of these groups of instances is very sensitive ( i . e . not robust ) to changes in P and the effect of P on period could not be assessed . The activity of the realistic bursters group also turned out to be very sensitive also to changes in Leak , with all instances being part of families of only one member . The activity of the bursters was slightly less sensitive to changes in Leak with 423 instances as members of families of one member and just 2 instances as members of a family of 2 members , which also precluded an assessment of an effect Leak on period . From previous research [3] , [11] , [36] , we knew that all three of these parameters influence the activity type of model instances . However , while we did expect the activity of the burster instances to be very sensitive to Leak [3] , we did not expect such a big influence of P . Finally , the activity of the realistic bursters showed less sensitivity to K2 variations than to Leak and P variations including 26 families of 2 members and 2 families of 3 members . Burster instances showed similar sensitivity to changes in K2 variations as the realistic instances including families of 2 , 3 , and 4 members ( 45 , 2 , and 1 , respectively ) . For all families of both bursters and realistic bursters , an increase in K2 resulted in a monotonic decrease of the period . For each family , first we ordered in ascending order its members by the amount of the parameter varied ( here K2 ) , and then we calculated the variation ( here decrease ) of the period as being the difference between the periods of the last and first members of the respective family . The average decrease of the period for each family were: −2 . 43 s ( range −4 . 46 to −0 . 52 ) for realistic instances of families of 2 members; −3 . 16 s ( range −3 . 18 to −3 . 14 ) for realistic instances of families of 3 members; −2 . 41 s ( range −4 . 46 to −0 . 52 ) for burster instances of families of 2 members; −3 . 35 s ( range −3 . 52 to −3 . 18 ) for bursters of families of 3 members; and the only family of 4 members of the bursters had a decrease of −4 . 18 s of the period values . The activity of the realistic HCO group was quite sensitive to P also ( 92 , 970 families of one member , and 3 , 048 families of 2 members ) , but less sensitive than for the realistic burster instances . As expected [3] , the activity of this group was less sensitive to Leak than the activities of the burster and realistic burster groups ( 66 , 611; 11 , 873; 2133; 452; 92; and 7 , respectively , families of 1–6 members ) . Finally , the activity of the realistic HCO group showed a lesser sensitivity to changes in K2 than the sensitivity to Leak ( 68 , 702 families of one member , 13 , 163 families of 2 members , 1 , 299 families of 3 members , 34 families of 4 , and 1 family of 5 members ) . As found by Hill et al . [29] , an increase in K2 resulted in a monotonic decrease of the period for most of the families of this group . Only 19 ( out of 13 , 163 ) families of 2 members showed an increase in period . The average decrease of the period for each family were: −3 . 32 s ( range −9 . 75 to 1 . 29 ) for families of 2 members; −4 . 87 s ( range −9 . 65 to −0 . 38 ) for families of 3 members; −6 . 59 s ( range −9 . 09 to −2 . 31 ) for families of 4 members; and the only family of 5 members had a decrease of −9 . 46 s of the period values . Then , an increase of P resulted in an increase of the period for most P families of 2 members ( 3 , 014 out of 3 , 048 ) . The average increase in the period values was 4 . 255 s ( range −1 . 507 to 9 . 778 ) for all these 2 member families . Based on the large number of multimember families , it seems that in the heartbeat HCO , inhibition changes the influence of P , K2 , and Leak on network activity by making the activity of HCO instances more robust to changes in these parameters .
Similar to the results presented in [3] , our pairwise plots revealed that several parameters work in pairs to produce more burster ( figure not shown ) and realistic burster instances ( Figure 4 ) : increasing h together with more hyperpolarized ELeak; making CaS larger , together with a larger Leak ( not monotonically , but with a wave shape ) ; increasing K2 together with more depolarized ELeak; increasing Leak ( above 50% ) together with more hyperpolarized ELeak - for more depolarized ELeak and smaller Leak there were less number of instances in each group than when ELeak was more hyperpolarized . For the HCOs ( figure not shown ) and realistic HCOs ( Figure 3 ) , decreasing Leak together with more depolarized ELeak produced the most instances ( maximum at Leak = 25% ) . However , from our pairwise plots we cannot assess the potential relationships between three parameters , as for example those stated in [3] between Leak , ELeak and CaS . Our aim here was to find all potential existing correlations in our models whether between two , three or more parameters . Recent studies , both experimental [16] , [17] , [19] , [23] , [27] and modeling [11] , [21] , [20] , [38] , have shown in several systems that consistent activity is maintained despite a 3–5 fold variation from animal to animal or model instance to model instance among ionic and synaptic conductances and that correlations exist between these parameters . Thus the suggestion has arisen that the functional activity of a given neuron may reside in the set of parameter correlation rules it maintains rather than in the value of any particular parameter . However , the precise combinations of parameters that is adequate to preserve functional neuronal or network activity in not fully elucidated for any system . Most studies have reported positive linear correlations [19] , [23] , while recently [21] , [27] have shown negative linear correlations , and several theoretical studies have reported the existence of nonlinear correlations [11] , [20] . Such correlations have been reported in single cells [39] and in networks of two or more cells correlations [19] , [27] and they have been found between two parameters [21] , [27] , as well as three or four parameters [3] , [27] , [40] , [17] . The potential for general insights into mechanisms for bursting in single neurons and HCOs motivated us to pursue this modeling study despite challenges imposed by large 8-dimensional parameter space that it presented . To find potential linear correlations among our varied parameters , we applied PCA to our four activity groups of interest . This method showed that for the bursters and the realistic bursters groups there is a linear correlation between the six parameters out of which Leak , K2 , and P had each about 3 times or more importance than the other three parameters ( Figure 6 ) . Plots ( Figure 7 A , B ) of each group's instances in the 3D space of Leak , K2 , and P showed a linear correlation for each group around the main diagonal ( Figure 10 ) . Two corollaries emerged from our analysis: 1 ) for the bursters and the realistic bursters the range of permissible ELeak's diminishes and ELeak must be more negative as the values of Leak , K2 , and P increase ( Figure 7 ) ; and 2 ) moderate to large values of Leak , K2 , and P produced more realistic bursters than the small values , and small to moderate values of Leak , K2 , and P produced more not realistic bursters than the large values ( Figure 12 ) . The above observations indicate that these three conductances work together to produce burster and realistic burster instances and begin to pinpoint the mechanisms supporting bursting in isolated heart ( HN ) interneurons . Olypher and Calabrese [11] used sensitivity analyses to predict coordinated changes of parameters that would lead to constant activity in the HN HCO model . They found that P opposes both Leak and K2 , with Leak and K2 having negative relative sensitivity of almost half of P‘s relative sensitivity ( positive ) . Our analyses ( Figures 7 and 10 ) yield similar yet contrasting results; Leak opposes P , but in an almost equal relationship ( see their weights in Figure 6 B ) . A small amount of Leak requires a small amount of P to produce bursting , and a large amount of Leak requires a large amount of P ( within some range ) . In addition , if P is small then K2 is small ( within a range ) , and if P is large then K2 must be large . Interestingly , all three parameters have negative weights in the equation given by the PCA method , with P and Leak having almost equal weights and being slightly bigger than the weight of K2 . Thus it appears that none of these three parameters is sufficient by itself to produce burster and realistic burster instances , but they must work together ( in linear combination ) in almost equal amounts towards producing the respective instances . PCA guarantees that all existing linear combinations ( in any number of parameters ) that characterize a set of data will be found , should such linear combinations exists . Thus our bursters and realistic bursters groups are characterized by the single dominant linear combination between parameters returned by the PCA method ( Figure 6 A ) . From a mechanistic standpoint , the observed correlation of Leak , K2 , and P and their corollaries fit our current understanding of bursting in HN neurons . None of these three currents show inactivation and their activation is relatively fast compared to the burst period ( instantaneous in the case of ILeak ) . IK2 is active only during the burst phase , owing to its depolarized range of activation , and provides outward current that limits depolarization . IP is active throughout the burst cycle owing to it broad and shallow activation curve , and it provides the inward current that drives baseline spiking activity . ILeak is also active throughout the burst cycle and provides the outward current necessary for repolarization after the burst . Essentially IK2 must oppose IP during the burst and ILeak must oppose it during the interburst interval . When Leak , K2 , and P are all small then IP is very weak during the interburst interval and a small ILeak ( i . e . , small Leak ) even with a relatively depolarized ELeak can effectively oppose it . But when P is moderate or large ( i . e . , a large IP ) then a large Leak with a relatively negative ELeak ( i . e . , a large ILeak ) is necessary to oppose IP during the interburst interval ( Figure 7 pie charts ) . Based on work with canonical HN models and analyses in the living neurons [29] , [3] , [13] we hypothesize that ICaS is critical for bursting in isolated HN neurons and that the burst duration is controlled by its inactivation dynamics , which in our model are fixed . This hypothesis is further supported by the data of Figure 4 C . Because ILeak is the main determinant of the interburst interval , a small Leak will then lead to short interburst intervals and thus to more not realistic burster instances ( Figure 12 ) . For the groups of realistic and HCO instances , PCA did not find any linear relationship between the parameters ( Figure 5 ) . Plots ( Figure 8 A , B ) of the instances of each HCO group in the 3D space defined by Leak , K2 , and P , which revealed correlations for the burster groups , showed that these groups of instances form complex shape , like a wedge . Goldman et al . [7] in their study on the robustness of activity type in single model neurons found similar relationships between parameters supporting bursting . This non-linearity suggests that these three parameters are not enough to characterize the HCO and realistic HCO groups . We hypothesize that Leak , K2 , and P play the same role in HCOs as outlined above for bursters: the added factor being synaptic inhibition , which provides outward current during the interburst interval . Because synaptic inhibition provides outward current during the interburst interval , the system no longer depends on a large Leak with a relatively negative ELeak to oppose a large IP , and in fact small Leak's are favored ( Figure 8 ) . Note that the pie chart analysis of Figure 8 B shows directly a lack of restriction on ELeak throughout the region of realistic HCO bursting . Figure 3 further corroborate this hypothesis by showing that the number of realistic HCO instances increases dramatically as SynS increases . Our plots in Figure 9 show that one can obtain linear correlations between parameters for HCOs when working in a reduced parameter space . In our case , we reduced the parameter space to only three maximal conductances , Leak , K2 , and P , and kept the rest of the parameters at their canonical values . Then we plotted the instances of our four groups of interests into this new 3D space . The plot of the reduced set of HCO instances still showed a complex relationship between the three parameters , while the plot of the reduced sets of realistic HCO instances ( and also of bursters and realistic bursters instances ) showed linearity . The reduced set of realistic HCOs produced a linear cluster in this space , and the reduced sets of bursters and realistic bursters produced lines similar to those seen to the unreduced sets ( Figure 10 ) . Our analysis here indicates that a strong correlation between Leak , K2 , and P , is critical in determining activity in bursters and realistic bursters and thus bursting activity should be very sensitive to their individual variation . Sensitivity analysis is the most common computational method used to assess the influence of a parameter on the activity type in neuronal models [29] , [11] , [7] , [36] . Previous work showed that the bursting activity of isolated HN model neurons is very sensitive to Leak , so that HN neurons are not robust busters under experimental conditions that alter leak properties [3] . We have confirmed and extended that finding here using our database of model instances by showing that Leak families have only one member . We have similarly shown a strong sensitivity to P and to a slightly lesser extent to K2 . Robustness of activity state requires correlated changes in these three parameters . When configured as an HCO , realistic bursting activity becomes substantially more robust to individual changes in these parameters , which can be seen as both the expanded occupancy of the parameter space in Figure 8 and by the large number of Leak , K2 , and P families with multiple members . Thus mutual synaptic inhibition adds robustness to bursting activity in HN neurons . Several recent studies suggest that correlated parameters could be key factors in maintaining functional activity states in neurons . Goldman et al . [7] found that a model neuron's robustness ( ability to maintain functional activity , e . g . regular bursting ) is determined by its sensitivity to sets of parameter changes . Hudson and Prinz [21] found that conductance correlations contribute to the robustness of critical features of electrical activity . Lamb and Calabrese [41] found partial conductance correlations that contribute to the activity phase of the leech heart motor neurons . The mechanisms that maintain functional bursting in the pyloric CPG of the stomatogastric nervous system of crabs employ several key parameter correlations ( linearly or not ) [17] , [18] , [20] , [21] , [23] which appear necessary for the maintenance of activity state . Likely evolution promoted such mechanisms to maintain robust activity and robustness seems to be achieved in the oscillator heart interneurons of the leech heartbeat CPG by three linearly correlated maximal conductances of P , Leak and K2 . Changes in any of these three parameters ( P , Leak or K2 ) must be accompanied by changes in the other two parameters in a linear correlation to maintain the neurons in a realistic bursting activity mode . Our results imply that these three parameters compensate for each other's variations to keep bursting functional . Moreover they show that linking these neurons by mutually inhibitory synapses into a HCO increases robustness . We leave unanswered for future work the question of how period is modulated while robustness is maintained .
We used Hill et al . 's model [29] of a half-center oscillator ( HCO ) which produces electrical activity ( rhythmic alternating bursting of mutually inhibitory neurons ) similar to that observed in the living system ( in the heartbeat central pattern generator or CPG of the leech ) . The model is publicly available on ModelDB repository ( https://senselab . med . yale . edu/ModelDB/ ) , accession number 19698 . The HCO model consists of a two reciprocally inhibitory model interneurons , represented as single isopotential electrical compartments with Hodgkin and Huxley [30] type intrinsic and synaptic membrane conductances . Each compartment contains 8 voltage-dependent currents , five inward currents INa - a fast Na+ current , IP - a persistent Na+ current , ICaF - a rapidly inactivating low-threshold Ca current , ICaS - a slowly inactivating low-threshold Ca current , Ih - a hyperpolarization-activated cation current ) and three outward currents ( IK1 - a delayed rectifier-like K current , IK2 - a persistent K current , IKA - a fast transient K current ) . The model has two types of inhibitory synaptic transmission between the two interneurons: graded transmission ( SynG ) and spike-mediated transmission ( SynS ) . The graded transmission SynG was modeled as a postsynaptic conductance controlled by presynaptic Ca2+ concentration and the spike-mediated transmission SynS was modeled as a postsynaptic conductance triggered by presynaptic spikes . The values for the maximal conductances and the leak reversal potential ( free parameters in the model ) that we used for our canonical model are CaS = 3 . 2 nS , h = 4 nS , P = 7 nS , K2 = 80 nS , Leak = 8 nS , SynS = 60 nS , SynG = 30 nS , Na = 200 nS , CaF = 5 nS , K1 = 100 nS , KA = 80 nS , and ELeak = −60 mV [31] . The kinetics , voltage-dependencies , reversal potentials of the intrinsic currents , and the synaptic connections of the HCO model interneurons have all been verified and previously adjusted to fit the biological data of leech interneurons [2] , [14] , [29] , [36] , [42] . The differential equations of the model are given in the following . The equation of the membrane potential ( V ) of each neuron is given by:where C is the total membrane capacitance ( ) , Iion is an intrinsic voltage-gated current , ILeak is the leak current , ISynS is the graded synaptic current , ISynS is the spike-mediated synaptic current , and Iinject is the injected current . Voltage-gated currents are given by given by given by where is the maximal conductance , Eion is the reversal potential , and the activation and inactivation variables for the parameters are given in Table 1 . where is the presynaptic membrane potential . where is the maximal synaptic conductance , is the time of a spike event , and M is the modulation variable of the synapse determined from , . The synaptic function is given by where is a normalization constant with , and the decay ( ) and rise ( ) times of the synaptic conductance . In our previous work [31] , we performed extensive simulations of this HCO model by systematically varying eight key parameters ( a brute-force approach ) . Figures 2 A and B show the intrinsic currents and the synaptic conductances of two randomly chosen simulated instances , one a realistic HCO and one a realistic burster . The figures show how the currents interact ( including their ratio and the timing ) to produce the realistic leech bursting activity . One can see in the examples that IA and ICaF are very small , and thus have a small influence on the cell's activity . INa and IK1 are mainly involved is spiking and must be co-varied to maintain constant spiking during the burst . While spike frequency is an important determinant of HCO activity due to inhibition of the opposite neuron , it is better controlled by slow currents and spike amplitude and undershoot do not appear critical for model behavior . We simply came upon a canonical combination for these two currents that produced realistically sized spikes . Moreover , previous analysis involving varying one parameter at a time had identified maximal conductances SynS , SynG , P , K2 , h , CaS , and Leak , and ELeak as critically contributing to bursting behavior [3] . Thus in this analysis we did not vary K1 , KA , CaS , or Na but concentrated on varying the critical parameters . As some selection was necessary due to computational and financial limitations on database size , this selection seemed reasonable . In future we may be able to add the fast currents to the database . All model simulations were started from the same initial conditions , which were different for each of the two neurons and were obtained by running the canonical HCO model [29] for 200 s , such that one of the two neurons was in its bursting state and the other one was being inhibited . The same parameter values were used in each of the paired model neurons . The eight parameters varied were: the maximal conductances of spike-mediated ( SynS ) , graded transmission ( SynG ) , Leak , P , CaS , h , and K2 , across of 0% , 25% , 50% , 100% , 125% , 150% , and 175% of their canonical values and ELeak across −70 , −65 , −60 , −55 , and −50 mV values . After changing a parameter , we ran each model instance for 100 s to allow the system to establish stable activity , and then we ran it for another 100 s , from which we recorded the voltage traces of the electrical activity corresponding to its paired neurons and the corresponding spike times . The firing characteristics were analyzed and recorded into a database named HCO-db . In voltage traces we recognized a spike only if the potential waveform crossed a threshold of −20 mV . We defined a burst as having at least three spikes and a minimum inter-burst interval of 1 second . We defined the cycle period as being the interval between the middle spikes of two consecutive bursts . Phase was calculated on a per cycle basis , as being the delay from the middle spike of a burst of neuron B to the middle spike of the preceding burst of neuron A divided by the interval from this middle spike of the next burst of neuron A to the middle spike of the preceding burst of neuron A . The duty cycle was defined as the percentage of the period occupied by a burst . We defined a half-center oscillator instance ( HCO ) as having: two model interneurons each showing bursting activity with at least two bursts in a 40 s time interval , with each burst having normal spikes ( coefficient of variation of the amplitudes of the spikes within any burst is less than 0 . 07 ) ; a small variation of period ( coefficient of variation of period less than 0 . 05 ) ; relative phase in the range of ( 0 . 45–0 . 55 ) ; and at least one synaptic component present ( either SynS≠0 , or SynG≠0 , or both SynS≠0 and SynG≠0 ) . We considered a realistic HCO instance as being a HCO that showed realistic bursting corresponding to that observed in leech oscillator heart interneurons . Precisely , it was a HCO with period between 5–15 s , average spike frequency between 8–25 Hz , and duty cycle between 50–70% . We defined an isolated neuron instance ( isolated neuron ) as having two identical interneurons ( though started with different initial conditions , but otherwise identical ) , and no synaptic interaction ( i . e . , SynS = 0 and SynG = 0 ) . We defined a burster instance as being an isolated neuron instance for which both neurons had at least two bursts , each with normal spikes , and regular periods ( as defined above for the HCOs ) . Note that burster instances can be thought of as being HCOs with no synaptic connections . We defined a realistic burster as being a burster that showed realistic bursting corresponding to isolated leech oscillator heart interneurons . Precisely , it was a burster with period between 5–15 s , and average spike frequency between 8–25 Hz . Note that realistic bursters can be thought of as being realistic HCOs with no synaptic connections . We define a family as being a sub-set ( of a group ) of instances that have all the same parameter values except one ( e . g . , all realistic bursters that vary only by P constitute a family ) . Note that one can partition a group into families according to the number of members in the family . In our previous work [31] , we created a database of 10 , 485 , 760 HCO simulated model instances ( HCO-db , [32] ) by systematically varying eight key parameters ( a brute-force approach ) . The resulting parameter space includes 10 , 321 , 920 HCO instances which have at least one synaptic component present , and 163 , 840 isolated neuron instances which contain twin neurons without any synaptic interaction . By using our definitions above as criteria , we identified those simulated instances belonging to four groups [31]: functional HCOs encompassing 1 , 202 , 139 HCO instances and their subset of realistic HCOs having 99 , 066 instances , and of bursters encompassing 424 instances , of and their subset of realistic bursters encompassing 307 instances out of the entire database . By querying the HCO-db , we efficiently explored the instances from these four groups to determine which and how intrinsic membrane and synaptic parameters affect their electrical activity . In particular , we were interested in defining the parameter values that can lead to functional output from this circuit that conforms to that observed in the living system . For this , we applied the following methods to our groups of HCO model instances . Principal component analysis ( PCA ) is a powerful tool used in many fields for identifying potentially hidden patterns within a large multidimensional data set . PCA , proposed by Pearson in 1901 [43] , is a mathematical method based on an orthogonal linear transformation that allows for dimensionality reduction of a multidimensional data set without too much loss of information . This transformation converts the original data set into a set of linearly uncorrelated variables called principal components . Each principal component is a linear combination of the original variables . The number of principal components is less than or equal to the number of original variables . PCA allows for reducing the dimensionality of a data set by using only the first few principal components ( considered the most important ) . The first principal component has the largest possible variance and each succeeding component in turn has the next highest variance possible and it is orthogonal to ( i . e . , uncorrelated with ) the preceding components . Since all the principal components are orthogonal to each other , there is no redundant information . We applied the principal component analysis ( PCA ) [43] to our four groups of interest . For this , we used the princomp function provided in the MATLAB Statistics Toolbox [44] to obtain the principal components ( PC ) for each group . This function returns the coordinates of the original data in the new coordinate system defined by the principal components . We can visualize each group of instances within the 3D space defined by the first three principal components obtained for the respective group ( plots not shown , see S3: Figures 1–4 ) . The princomp function also calculates the coefficients of the linear combinations of the original variables ( our parameters ) that generate the principal components . Each coefficient represents the importance of the respective parameter within the principal component ( PCi = , where n is the number of the original parameters pj , and wi , j are the coefficients of these parameters for the principal component PCi ) . For each group of interest , we plotted the percent of variability explained by each principal component ( see Results ) . We used 3D Orthogonal Distance Regression ( ODR ) to assess the relationship between parameters identified in our PCA . Here we focus on the linear regression ( [45] ) and more precisely on the orthogonal linear regression method . The orthogonal linear regression method uses the Principal Components Analysis ( PCA ) method described above to fit a linear regression that minimizes the perpendicular distances from the data to the fitted model ( least square fit is minimum ) . The method is also called Total Least Squares method or Principal Component Regression ( [46] ) . Basically , given a set of points in a 3D space , the ODR method uses the coefficients of the first PCA corresponding to the 3D points to find a line ( called 3D ODR line ) in this 3D space that is least distance from the points . Once found , this line shows the tendency or direction of the points within the 3D space with respect to the three axes of the space . To test for the equality of our group distributions , we applied a two-sample Kolmogorov-Smirnov [47] nonparametric test ( Figures 12A and 12B ) . To perform a two-sample Kolmogorov-Smirnov test on our distributions ( see Results ) we used the kstest2 function in Matlab ( [44] ) . The function was applied with the default Matlab values of ‘Alpha’ = 0 . 05 significance level and ‘Tail’ of ‘unequal’ . We rejected the null hypothesis at the 5% significance level . The null hypothesis stated that the two samples are drawn from the same distribution .
|
Central pattern-generating networks ( CPGs ) must be remarkably robust , maintaining functional rhythmic activity despite fluctuations in internal and external conditions . Recent experimental evidence suggests that this robustness is achieved by the coordinated regulation of many membrane and synaptic current parameters . Experimental and computational studies showed that linearly correlated sets of such parameters allow CPG neurons to produce and maintain their rhythmic activity . However , the mechanisms that allow multiple parameters to interact , thereby producing and maintaining rhythmic single cell and network activity , are not clear . Here , we use a half-center oscillator ( HCO ) model that replicates the electrical activity ( rhythmic alternating bursting of mutually inhibitory interneurons ) of the leech heartbeat CPG to investigate potential relationships between parameters that maintain functional bursting activity in the HCOs and the isolated component neurons ( bursters ) . We found a linearly correlated set of three maximal conductances that maintains functional bursting activity similar to the animal in burster model instances , therefore increasing robustness of bursting activity . In addition , we found that bursting activity was very sensitive to individual variation of these parameters; only correlated changes could maintain the activity type .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"circuit",
"models",
"animal",
"models",
"cellular",
"neuroscience",
"model",
"organisms",
"neural",
"networks",
"computational",
"neuroscience",
"biology",
"and",
"life",
"sciences",
"computational",
"biology",
"neuroscience",
"research",
"and",
"analysis",
"methods"
] |
2014
|
Identifying Crucial Parameter Correlations Maintaining Bursting Activity
|
Thermophilic enzymes are often less active than their mesophilic homologues at low temperatures . One hypothesis to explain this observation is that the extra stabilizing interactions increase the rigidity of thermophilic enzymes and hence reduce their activity . Here we employed a thermophilic acylphosphatase from Pyrococcus horikoshii and its homologous mesophilic acylphosphatase from human as a model to study how local rigidity of an active-site residue affects the enzymatic activity . Acylphosphatases have a unique structural feature that its conserved active-site arginine residue forms a salt-bridge with the C-terminal carboxyl group only in thermophilic acylphosphatases , but not in mesophilic acylphosphatases . We perturbed the local rigidity of this active-site residue by removing the salt-bridge in the thermophilic acylphosphatase and by introducing the salt-bridge in the mesophilic homologue . The mutagenesis design was confirmed by x-ray crystallography . Removing the salt-bridge in the thermophilic enzyme lowered the activation energy that decreased the activation enthalpy and entropy . Conversely , the introduction of the salt-bridge to the mesophilic homologue increased the activation energy and resulted in increases in both activation enthalpy and entropy . Revealed by molecular dynamics simulations , the unrestrained arginine residue can populate more rotamer conformations , and the loss of this conformational freedom upon the formation of transition state justified the observed reduction in activation entropy . Our results support the conclusion that restricting the active-site flexibility entropically favors the enzymatic activity at high temperatures . However , the accompanying enthalpy-entropy compensation leads to a stronger temperature-dependency of the enzymatic activity , which explains the less active nature of the thermophilic enzymes at low temperatures .
To cope with the extremely hot habitats , thermophilic enzymes isolated from organisms thriving in these environments have usually evolved towards high thermal stability . Although thermophilic and mesophilic enzymes are comparably active at their respective temperatures where these enzymes function , thermophilic enzymes are often less active at lower temperatures [1] , [2] . To date , this reduced activity of the thermophilic enzymes at low temperatures remains only partially understood . Comparative study of activity , stability , and flexibility relationships of homologous enzymes from thermophilic , mesophilic , and psychrophilic enzymes showed that enzyme flexibility is often correlated with its activity but is inversely related to its stability [1]–[9] . Thermophilic enzymes generally acquire larger values of activation entropy ( ΔS# ) and enthalpy ( ΔH# ) than those of their mesophilic and psychrophilic homologues [3] , [10] . One popular interpretation of this observation is that the extra stabilizing interactions found in thermophilic enzymes will result in a more rigid enzyme . Increased rigidity has been proposed to explain why thermophilic enzymes are less active at low temperatures , while optimizing local flexibility is a structural adaptation of psychrophilic enzymes to remain active at low temperatures [1] , [2] , [11] . Here we used a pair of thermophilic/mesophilic acylphosphatases homologues ( acylphosphatase from hyperthermophilic archaeon Pyrococcus horikoshii ( PhAcP ) and human common-type acylphosphatase ( HuAcP ) ) as a model to study how local flexibility of the active site affects the enzymatic activity . Acylphosphatases catalyze the hydrolysis of acylphosphates to phosphates and carboxylates , by providing an invariant arginine residue ( Arg20 in PhAcP , Arg23 in HuAcP ) that stabilizes the negative charges in the transition state ( Figure 1 ) [12]–[14] . Despite remarkable conservation of sequence and structure at the active site , the catalytic activity of the thermophilic PhAcP is significantly poorer as compared with its mesophilic homologue HuAcP at low temperatures [14] . To date , the crystal structures of three thermophilic acylphosphatases and four mesophilic acylphosphatases are available ( Figure S1 ) . Structural comparison of the crystal structures reveals that this active-site residue Arg-20 forms a salt-bridge with the C-terminal carboxyl group of the glycine residue Gly-91 in PhAcP ( Figure 1B ) [14] . This salt-bridge is also present in thermophilic acylphosphatases from Sulfolobus solfataricus ( PDB: 2bjd ) [15] and Thermus thermophilus ( PDB: 1ulr ) , but not in mesophilic acylphosphatases from human [16] , cattle [13] , Drosophila [17] , and Bacillus subtilis ( PDB: 2vh7 , 2acy , 1urr , 3br8 ) ( Figure S1 ) . In PhAcP , the formation of the salt-bridge is facilitated by the C-terminal glycine , which can adopt an unusual φ angle of ∼180° . In the cases of thermophilic acylphosphatases from T . thermophlius and S . solfataricus , the C-terminal carboxylate groups are brought in the position to form the salt-bridge by having one less residue . The presence of the salt-bridge probably increases the rigidity of the active-site residue in thermophilic PhAcP by locking the guanido group of Arg-20 to prevent conformational fluctuation during catalysis . In this study , we have shown that this salt-bridge is largely responsible for the observed differences in activation energy , and hence the temperature-dependency of enzymatic activity between thermophilic and mesophilic acylphosphatases . By disrupting the salt-bridge , the activation energy for the thermophilic enzyme was converted to mesophilic-like . Parallel findings were obtained by introducing the salt-bridge into the mesophilic HuAcP . Analysis of thermodynamics parameters showed that the removal of the salt-bridge decreases both the values of ΔH# and ΔS# . Molecular dynamics ( MD ) simulations suggested that the active-site arginine residue could adopt more rotamer conformations in the absence of the salt-bridge . The loss of this conformational freedom upon the formation of the transition state justified the observed reduction in the activation entropy . Finally , the implications on why thermophilic enzymes are less active at low temperatures are discussed .
The salt-bridge between the active-site residue Arg-20 and the C-terminal carboxyl group is found only in thermophilic PhAcP but not in mesophilic HuAcP ( Figure S1 ) . To investigate how this salt-bridge affects the enzymatic activity , we need to engineer a variant of PhAcP in which the salt-bridge was disrupted without perturbing the active site . Molecular modeling suggested that the formation of the salt-bridge required the C-terminal residue Gly-91 to adopt an unusual phi angle of ∼180° . Thus , any non-glycine substitutions could disrupt the interaction . We then replaced the Gly-91 by an alanine residue to yield a variant of PhAcP ( PhG91A ) by site-directed mutagenesis . The substitution did not significantly affect the stability of the enzymes , for the apparent melting temperatures of PhWT ( ∼107°C ) and PhG91A ( ∼106°C ) were similar and there was no significant difference between the values of free energy of unfolding ( ΔGu ) for PhWT ( 58±7 kJ mol−1 ) and PhG91 ( 51±6 kJ mol−1 ) ( Figure S2 ) . The structure of PhG91A was determined by x-ray crystallography at 2 . 4 Å resolution ( Table S1 ) , and it is superimposable with the structure of wild-type PhAcP ( PhWT ) ( Cα root-mean-square deviation ( r . m . s . d . ) of 0 . 21 Å between chain A of 1W2I [PhWT] and 2W4D [PhG91A] ) . Most importantly , the C-terminal alanine residue was confirmed to face away from the active-site arginine residue leading to the disruption of the salt-bridge ( Figure 2A ) . Next , the enzymatic activity of PhG91A and PhWT at different temperatures was determined and the results represented in the Arrhenius plot ( ln kcat versus 1/T ) . As shown in Figure 2C , PhWT and PhG91A both gave straight lines that intersected at ∼298 K . PhWT had a steeper slope denoting a stronger temperature dependency of the enzymatic activity than PhG91A . At ∼298 K , both enzymes shared similar kcat values ( 228±15 and 211±25 s−1 for PhWT and PhG91A , respectively ) ( Table 1 ) . In terms of the activation energy ( Ea ) of the reactions calculated from the slope of the Arrhenius plot , PhWT displayed a greater Ea value ( 49 . 1±1 . 4 kJmol−1 ) than PhG91A ( 32 . 1±1 . 7 kJmol−1 ) ( Table 1 ) . Our data suggest that the removal of the salt-bridge between Arg20 and Gly91 results in a weaker temperature dependency of enzymatic activity . The consequences are that the enzymatic activity of PhWT will be higher at elevated temperatures but becomes more sluggish at lower temperatures . For example , at lower temperature , e . g . 283 K , PhG91A lacking the salt-bridge retains a significantly greater kcat value than PhWT ( Figure 2C ) . As shown earlier , the presence of the salt-bridge increases the activation energy of catalysis . We then questioned whether the introduction of the salt-bridge into the mesophilic homologue would result in the same observation . Two variants of HuAcP were constructed . Firstly , the C-terminal residue of HuAcP was substituted with a glycine residue ( HuG99 ) to engage the salt-bridge with the active-site arginine residue . Secondly , a pseudo-wild-type HuAcP ( HuA99 ) was created by replacing the C-terminal residue with an alanine residue to resemble PhG91A . Crystal structures of HuA99 and HuG99 , resolved at 1 . 5 Å and 1 . 7 Å , respectively ( Table S1 ) , demonstrated that they were superimposable with a Cα r . m . s . d . value of 0 . 17 Å . In a good agreement with our original design , the salt-bridge between the C-terminal carboxyl group of Gly-99 and Arg-23 at the active site was only detected in HuG99 , but not in the pseudo-wild-type HuA99 ( Figure 2B ) . Regarding the Arrhenius plot ( Figure 2C ) , both HuG99 and HuA99 exhibited straight lines that intersected also roughly at 298 K . The salt-bridge bearing variant HuG99 illustrated a steeper slope of the Arrhenius plot , i . e . a stronger temperature dependency of the enzymatic activity , as compared to that of HuA99 lacking the salt-bridge . Importantly , the activation energy ( 52 . 5±2 . 5 kJmol−1 ) for HuG99 was analogous to that for PhWT ( 49 . 1±1 . 4 kJmol−1 ) , while the activation energy for HuA99 ( 29 . 5±1 . 9 kJmol−1 ) and HuWT ( 37 . 6±3 . 1 kJmol−1 ) was comparable to that for PhG91A ( 32 . 1±1 . 7 kJmol−1 ) ( Table 1 ) . Taken together , our data suggest that the presence of the salt-bridge between the C-terminal carboxyl group and the active-site arginine residue increases the activation energy that results in a stronger temperature dependency of the enzymatic activity . To provide further insights into how the active-site salt-bridge affected the enzymatic activity of acylphosphatases , we calculated and summarized the free energy ( ΔG# ) , enthalpy ( ΔH# ) , and entropy ( ΔS# ) of activation , as well as the changes of these parameters ( ΔΔG# , ΔΔH# , ΔΔS# ) upon the removal of the active-site salt-bridge ( Table 1 and Figure 3 ) . The values of ΔΔH# and TΔΔS# for HuWT were somewhat lower than those for HuA99 . Nevertheless , the patterns of changes in the activation thermodynamics parameters were similar in both mesophilic and thermophilic acylphosphatases . Noteworthy , the removal of the active-site salt-bridges , i . e . PhWT versus PhG91A and HuG99 versus HuA99 , led to large negative values of ΔΔS# ( Table 1 ) . As shown in Table 1 and Figure 3 , the large negative values of TΔΔS# due to the removal of the salt-bridge were accompanied by large negative values of ΔΔH# . At ∼298 K , these two effects canceled out each other , and therefore , the changes in ΔG# and the resulting reaction rates were minimal ( Table 1 ) . At <298 K , the enthalpic term ( ΔΔH# ) was smaller than the entropic term ( TΔΔS# ) . As a result , acylphosphatases without the salt-bridge ( PhG91A , HuA99 ) were more active than those with the salt-bridge ( PhWT , HuG99 ) ( Figures 2 and 3 ) . Conversely , at >298 K , the acylphosphatases with the salt-bridge became more active ( Figures 2 and 3 ) . To investigate if the active-site salt-bridge affects substrate binding , we have performed isothermal titration calorimetry to measure the thermodynamics parameters of substrate binding for variants of acylphosphatases using a substrate analogue , S-benzyloxycarbonyl-thiosulfonate ( Table 2 and Figure S3 ) . Our results showed that the removal of the active-site salt-bridge had minimal effect on the substrate binding affinity ( i . e . similar values of Ka and ΔGb ) but resulted in significant decreases in both enthalpy and entropy of binding ( Table 2 ) . The entropic contribution ( TΔΔSb ) of removing the active-site salt-bridge to substrate binding was −6 . 0 and −3 . 8 kJ mol−1 for thermophilic and mesophilic AcP , respectively . Noteworthy , these values were much less negative than the corresponding changes in activation entropy ( TΔΔS# , Table 1 ) , which were −17 . 2 and −23 . 3 kJ mol−1 , suggesting that upon removal of the active-site salt-bridge , the system loses more entropy in the formation of the transition state than the formation of the enzyme-substrate complex . We have shown that the removal of the active-site salt-bridge decreased the activation entropy , suggesting an increase in the local flexibility at the active site . Next , we performed the MD simulations to further characterize the local flexibility of the active site affected by the salt-bridge . Three 10 ns trajectories were obtained for each of the acylphosphatases studied , namely PhWT , PhG91A , HuG99 , and HuA99 . The MD simulations were stable , with values of Cα root-mean-square deviation below ∼1 . 5 Å ( Figure S4 ) . The distance between the C-terminal carboxyl group and the guanido group of the active-site arginine residue was less than 4 Å throughout the entire simulation of PhWT and HuG99 , denoting the presence of the salt-bridges in these proteins . On the other hand , the distance was ∼6 Å throughout the simulation of PhG91A and HuA99 , suggesting the salt-bridge was broken . Moreover , for PhG91A and HuA99 , the removal of the salt-bridge did not affect backbone flexibility significantly , as indicated by the comparable values of Cα root-mean-square fluctuations derived from the MD simulations ( Figure S5 ) . The notable change in the flexibility upon the removal of the active-site salt-bridge was localized in the side-chain conformation of the active-site arginine residue ( Arg-20 in PhAcP and Arg-23 in HuAcP ) . In the crystal structure of acylphosphatases , the active-site arginine residue adopted the mtm180° rotamer conformation ( named after the convention of Lovell et . al . , 2000 [18] ) ( Figure 4 ) . In the simulations of the salt-bridge bearing PhWT and HuG99 , the side-chains of the arginine residues populated mainly the native mtm180° rotamer conformation ( Figure 4A and C ) . In contrast , despite the mtm180° rotamer conformation , transitions to ptt180° , ttp180° , and mtt180° rotamers were prominent in the MD trajectories of PhG91A and HuA99 ( Figure 4B and D ) . Our results suggest that the removal of the active-site salt-bridge allows the arginine residue to populate several more rotamer conformations other than the orientated mtm180° rotamer conformation in the salt-bridge structure .
In this study , we used a “mirror-image” mutation approach [19] to investigate the role of the salt-bridge that restricts the flexibility of the active-site arginine residue on the enzymatic activity of acylphosphatases . Our data clearly demonstrated that the removal of the active-site salt-bridge in the thermophilic PhAcP decreased both the ΔH# and ΔS# , while a parallel trend was observed when the salt-bridge was introduced in the mesophilic HuAcP . Our results strongly indicate that the salt-bridge increases the activation entropy by rigidifying the active-site arginine residue . From the MD simulation analysis , in the absence of the salt-bridge , the active-site arginine residue populates a broader distribution of conformations in the ground state ( Figure 4 ) . That TΔΔS# took more negative values than TΔΔSb ( Tables 1 and 2 ) suggests that the majority of these degrees of freedom of the arginine residue are lost upon the formation of the transition state , in which a highly restrained positioning of the active-site residue is required for the catalysis to optimally occur . Indeed , based on our previously proposed model of an enzyme-substrate complex of the acylphosphatase [14] , the active-site arginine residue has to adopt the mtm180° rotamer conformation in order to stabilize the negative charges developed on the carbonyl group of the leaving group during the formation of the transition state , while other rotamer conformations are not productive ( Figure 5 ) . This loss of the conformational freedom results in a more negative value of ΔS# as in the cases of PhG91A and HuA99 . In our case , the active-site salt-bridge contributes to a decrease of ∼20 kJ mol−1 in the entropic penalty at 298 K ( Table 1 ) , which can translate into a >3 , 000-fold increase in kcat if the ΔH# remains constant . From the entropic point of view , rigidifying the active-site residue should increase the enzymatic activity rather than decrease it . The reduced activity of thermophilic acylphosphatase at low temperatures is caused by the accompanying increases in activation enthalpy that counteracts the entropic term ( Table 1 ) . The cause of such enthalpy-entropy compensation is intriguing . It has been argued that flexibility can contribute to the lowering of activation barriers because it facilitates sampling of conformational sub-states that have lower barriers for catalysis to occur [20] , [21] . In another view by Warshel and co-workers , enzyme catalysis is determined mainly by electrostatic reorganization energy [22] , and the reduction of ΔH# in mesophilic and psychrophilic enzymes is probably a result of the reduction of the reorganization energy but not changes in flexibility [23] . Another possibility is that the presence of a nearby negatively charged C-terminal carboxyl group may destabilize the transition state and thereby increase the activation enthalpy . Nevertheless , the enthalpy-entropy compensation appears to be a general property of weak intermolecular interactions . According to a simple model by Dunitz , the enthalpic and entropic contributions to the free energy for weak intermolecular interactions should compensate each other at ∼300 K [24] . Why is the active-site salt-bridge present in thermophilic acylphosphatases but absent in mesophilic homologs ? Our results suggest that the active-site salt-bridge increases the enzymatic activity at higher temperatures where the entropic term dominates ( Figure 2 ) . For example , the kcat value for PhWT ( 760 s−1 ) at 318 K was significantly higher than that for PhG91A ( 520 s−1 ) . From this point of view , rigidifying the active-site residue is a structural adaption of thermophilic acylphosphatases that favors enzymatic activity at high temperatures . Due to enthalpy-entropy compensation , the salt-bridge also leads to a stronger temperature dependency in enzymatic activity so that the thermophilic acylphosphatase becomes less active at low temperatures . In fact , the low-temperature activities ( at <298 K ) of the thermophilic acylphosphatase ( like in the case of PhG91A ) can be improved by the removal of the salt-bridge that lowers the activation enthalpy and increases the local flexibility of the active-site arginine residues . This observation is consistent with the suggestion that psychrophilic enzymes adapt to remain active at low temperatures by lowering the activation enthalpy [8] , [9] , [25] , [26] . Our results showed that the contribution of the active-site salt-bridge to the thermal stability of the acylphosphatase , if any , is small ( Figure S2 ) . Although many sequence-structure comparisons suggest that thermophilic proteins tend to have more salt-bridges than their mesophilic homologues [27]–[29] , whether salt-bridges stabilizes proteins is context-dependent because the favorable electrostatic interaction of opposite charges may be offset by dehydration penalty and the entropic cost of fixing the salt-bridging groups [30]–[33] . Moreover , surface-charged residues may stabilize proteins through long-range electrostatic interactions [34]–[39] . Using a genetic algorithm that optimizes the surface electrostatic interactions , Makhatadze and co-workers introduced five substitutions to human acylphosphatase and improved its melting temperature by ∼10°C without affecting its enzymatic activity [40] . Their results suggest that one can improve the thermal stability of an enzyme without compromising its activity . Noteworthy , our data also suggest that the active-site salt-bridge is not the sole factor contributing to the reduced activity of the thermophilic acylphosphatase at low temperatures . For instance , the thermophilic PhG91A is consistently less active than the mesophilic HuA99 although both enzymes lack the active-site salt-bridge ( Figure 2C ) . While the two enzymes have similar values of activation enthalpy ( i . e . similar slope in the Arrhenius plot ) , the differences in enzymatic activity are a result of PhG91A having a more negative value of activation entropy ( Table 1 ) . As all active-site residues in acylphosphatases are highly conserved , our data support the conclusion that substitutions at non-active-site residues play a critical role in decreasing the activation entropy and hence the enzymatic activity of the thermophilic acylphosphatase . That non-active-site residues do affect enzymatic activity has also been demonstrated in other enzymes [41] , [42] . Although how non-active-site substitutions affect activation entropy is not known , the reduction in activation entropy is unlikely to be caused by rigidifying the active site of the thermophilic enzyme , which is supposed to increase the activation entropy rather than decrease it . Apparently , substitutions that increase the activation entropy without affecting activation enthalpy may be a good strategy to improve the enzymatic activity of thermophilic enzymes at low temperatures .
The fragments of mutants were amplified by polymerase chain reaction method and subcloned into pET507a , an in-house modified vector with a multiple cloning site inserted between the NcoI and BamHI sites of pET3d ( Novagen ) . DNA sequencing was performed to check the sequence of all mutants created . The primers used for the mutations were as follows: PhG91A forward ( 5′ TAACTACCATGGCCATAGTTAGGGCTCAC 3′ ) and reverse ( 5′ TAACTAGGATCCTCACGCAACGATCCTGAA 3′ ) ; pseudo-WT HuA99 forward ( 5′ TAACTACCATGGCAGAAGGAAACACCCTG 3′ ) ; reverse ( 5′ TAGCGCGGATCCTTACGCTACAATTTGGAAG 3′ ) ; and HuG99 reverse ( 5′ TAGCGCGGATCCTTAGCCTACAATTTGGAAG 3′ ) . Protein samples of all acylphosphatases and their mutants were expressed and purified as described previously [14] , [16] . The continuous optical enzymatic assay for acylphosphatase was performed as previously described [14] using benzoyl phosphate as substrate . The assay was performed in triplicates by incubating substrate from 0 . 05 to 2 . 0 mM with acylphosphatases from 0 . 8 to 1 . 5 nM in 0 . 1 M sodium acetate buffers at pH 5 . 3 . The rate of hydrolysis was monitored by the decrease of the absorbance at 283 nm . For the mesophilic acylphosphatases , the assay was performed at 283 , 288 , 293 , 298 , and 303 K . For the thermophilic acylphosphatases , the temperature range was extended to include 308 , 313 , and 318 K . Enzyme kinetics parameters were obtained up to 318 K because the substrate benzoyl phosphate became too labile at higher temperatures . Calorimetric measurements were carried out using a Nano ITC isothermal calorimeter ( TA Instruments ) at 298 K . Substrate analogue S-benzyloxycarbonyl-thiosulfonate ( 30 mM ) ( Sigma-aldrich ) was titrated in 25 injections of 4 µl each to the protein sample ( 1 . 5 mM ) in 0 . 1 M sodium acetate buffer at pH 5 . 3 in a 1 ml sample cell . The data were analyzed by the program NanoAnalyze provided by the manufacturer . Crystals of PhG91A , HuA99 , and HuG99 were grown using the sitting-drop-vapor diffusion method at 289 K . The crystallization conditions are summarized in Table S1 . The crystals were cryoprotected by soaking in 25% ( w/v ) glycerol for PhG91A or polyethylene glycol-400 ( PEG400 ) for HuA99 and HuG99 in their corresponding mother liquors . The crystals were then loop-mounted and flash-cooled in liquid nitrogen . X-ray diffraction datasets were collected at 100 K using an in-house R-AXIS IV ++ imaging-plate system and a rotating copper-anode x-ray source ( Rigaku MicroMax-007 with VariMax optics ) . The diffraction data were processed with MOSFLM , SCALA , and TRUNCATE in the CCP4 suite [43] . The structures were resolved by the molecular replacement using the crystal structures of wild-type PhAcP [14] and HuAcP [16] as the search templates . Models were built by XTALVIEW [44] and refined by the programs CNS [45] and REFMAC5 [43] . The refined structures were validated by PROCHECK [46] and WHATIF [47] . The Ramachandran analysis was performed using the program MOLPROBITY [48] . The details of the MD simulations are described in Text S1 . In brief , all simulations were performed using GROMACS version 3 . 3 with the all-atom OPLSAA force field [49] , using a 0 . 002 ps time step for 10 ns . Three MD trajectories were obtained for each of the PhWT , PhG91A , HuG99 , and HuA99 , and the structures were analyzed at every 1 ps interval .
|
Although enzymes from thermophiles thriving in hot habitats are more stable than their mesophilic homologs , they are often less active at low temperatures . One theory suggests that extra stabilizing interactions found in thermophilic enzymes may increase their rigidity and decrease enzymatic activity at lower temperatures . We used acylphosphatase as a model to study how flexibility affects enzymatic activity . This enzyme has a unique structural feature in that an invariant arginine residue , which takes part in catalysis , is restrained by a salt-bridge in the thermophilic homologs but not in its mesophilic homologs . Here , we demonstrate the trade-offs between flexibility and enzymatic activity by disrupting the salt-bridge in a thermophilic acylphosphatase and introducing it in the mesophilic human homolog . Our results suggest that the salt-bridge is a structural adaptation for thermophilic acylphosphatases as it entropically favors enzymatic activity at high temperatures by restricting the flexibility of the active-site residue . However , at low temperatures the salt-bridge reduces the enzymatic activity because of a steeper temperature-dependency of activity .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"biochemistry",
"enzymes",
"biology",
"biophysics",
"simulations",
"biophysics"
] |
2011
|
A Rigidifying Salt-Bridge Favors the Activity of Thermophilic Enzyme at High Temperatures at the Expense of Low-Temperature Activity
|
Enteropathogenic Escherichia coli ( EPEC ) strains are defined as extracellular pathogens which nucleate actin rich pedestal-like membrane extensions on intestinal enterocytes to which they intimately adhere . EPEC infection is mediated by type III secretion system effectors , which modulate host cell signaling . Recently we have shown that the WxxxE effector EspT activates Rac1 and Cdc42 leading to formation of membrane ruffles and lamellipodia . Here we report that EspT-induced membrane ruffles facilitate EPEC invasion into non-phagocytic cells in a process involving Rac1 and Wave2 . Internalized EPEC resides within a vacuole and Tir is localized to the vacuolar membrane , resulting in actin polymerization and formation of intracellular pedestals . To the best of our knowledge this is the first time a pathogen has been shown to induce formation of actin comets across a vacuole membrane . Moreover , our data breaks the dogma of EPEC as an extracellular pathogen and defines a new category of invasive EPEC .
The human pathogens enteropathogenic Escherichia coli ( EPEC ) and enterohemorrhagic E . coli ( EHEC ) [1] and the mouse pathogen Citrobacter rodentium [2] are closely related extra-cellular diarrhoeal agents characterized by their ability to colonize the gut epithelium via attaching and effacing ( A/E ) lesion formation ( reviewed in [3] ) [4] . Similarly to other Gram-negative bacteria EPEC , EHEC and C . rodentium encode a type III secretion system ( T3SS ) , which is central to their infection strategy ( reviewed in [5] ) [6] . This complex machinery translocate dozens of effector proteins [7] , [8] directly from the bacteria to the eukaryotic cell cytoplasm ( reviewed in [9] ) . The translocated effectors are targeted to various sub-cellular compartments where they subvert a plethora of cell signaling pathways via interactions with a range of host cell proteins . The host cell cytoskeleton is a common target of T3SS effectors [10] . EPEC , EHEC and C . rodentium translocate the effector Tir into the plasma membrane where it functions as a receptor for the bacterial outer membrane protein intimin [11] . Intimin:Tir interaction leads to activation of N-WASP and formation of actin rich pedestals on which the extracellular bacteria rest [12] . In addition to Tir , A/E pathogens translocate a variety of other effectors which also modulate the host cell cytoskeleton including EspG/EspG2 , which induce depolymerization of the microtubule network [13] , Map , which induces formation of transient filopodia early in infection [14] and EspM which directs formation of actin stress fibers [15] . Map and EspM are members of the WxxxE family [15] , [16] , [17] , which was first grouped together based on conserved peptide motif consisting of an invariant tryptophan and glutamic acid residues separated by three variable amino acids and their shared ability to subvert host cell small GTPase signaling . Small GTPases cycle between an inactive GDP bound and an active GTP bound form , allowing them to function as molecular switches in response to a variety of stimuli . The switch from inactive to active forms results in a conformational change , which allows the GTPase to bind downstream mammalian effectors . Small GTPases are regulated by guanine exchange factors ( GEFs ) , GTPase activating proteins ( GAPs ) and guanine dissociation inhibitor ( GDI ) proteins ( reviewed in [18] , [19] ) . The three best characterized Rho GTPases are RhoA , Rac1 and Cdc42 which are implicated in formation of stress fibers , lamellipodia and filopodia respectively ( reviewed in [20] ) . The WxxxE effectors were originally proposed to be functional mimics of mammalian small GTPases [16] . However , we have recently shown that EspM activates RhoA [15] whereas Map induces filopodia via activation of Cdc42 and RhoA [17] . In addition to Map and EspM we have recently discovered the novel WxxxE effector EspT , which is encoded by C . rodentium and a subset of EPEC strains [21] , including EPEC E110019 which caused a sever outbreak in Finland in 1987 that affected children and adults alike [22] . We have shown that EspT induces formation of lamellipodia and membrane ruffles in epithelial cells via activation of Rac1 and Cdc42 [23] . Membrane ruffles are sheet like structures which are induced by mammalian cells in order to facilitate crawling movement , macro-pinocitosis and receptor recycling ( reviewed in [24] ) . These protrusion are regulated through activity of Rho family GTPases and their downstream effectors ( reviewed in [25] ) . Importantly , a subset of invasive bacterial pathogens hijack and subvert mammalian signal transduction pathways which facilitate formation membrane ruffles in order to promote bacterial entry into mammalian cells . Perhaps the best studied of these pathogens are Salmonella and Shigella which induce extensive membrane ruffles at the site of bacterial attachment ( reviewed in [26] , [27] ) . Salmonella invasion is dependent upon the activity of several T3SS effector proteins including SopE/E2 which act as GEFs for Rac1 and Cdc42 [28] and SopB which activates the RhoG GEF SGEF [29] . Shigella has also evolved several invasive mechanisms . For example the translocator IpaC has been shown to induce ruffles at the site of Shigella entry via the activation of Cdc42 [30] , recruitment of Src kinase [31] and activation of Abl kinase [32] . The Shigella WxxxE effector IpgB1 has also been shown to induce membrane ruffles via interaction with the ELMO DOCK180 complex which results in activation of Rac1 [33] . EPEC , EHEC and C . rodentium are generally considered extracellular pathogens and their attachment sites on epithelial cells are normally characterized by the assembly of an actin rich pedestal rather than membrane ruffles ( reviewed in [3] ) . However , in both rabbit and human biopsies EPEC have been visualized inside enterocytes and detected in the sub mucosa , mesenteric lymph nodes and spleen [34] ( reviewed in [35] ) . Recently Hernandes et al has shown that the atypical EPEC strain 1551-2 is capable of invading cultured epithelial cells in an intimin omicron dependent manner [36] . As EspT induces membrane ruffles similar to those triggered by IpgB1 [37] the aim of this study was to investigate if expression of EspT leads to EPEC cell invasion and to define the underlying mechanism .
A large screen of clinical EPEC isolates for the presence of espT , a T3SS effector ( Fig . S1 ) , has shown that the gene is present in ca . 1 . 8% of the tested strains [21] . In order to investigate the role of EspT in cell invasion we selected to use the espT positive strains E110019 and C . rodentum; the espT negative EPEC , strain JPN15 [38] , was used as a control . In addition , we generated a JPN15 clone that expresses EspT from the bacterial expression vector pSA10 ( pICC461 ) . We infected serum starved HeLa , Swiss 3T3 and Caco2 cells with E110019 , JPN15 and JPN15 expressing EspT; the cells were then fixed and processed for scanning electron microscopy ( SEM ) . The JPN15-infected HeLa and Swiss 3T3 cells displayed characteristic diffuse bacterial adhesion without any noteworthy surface structures . Caco2 cells infected with JPN15 also show a diffuse pattern of bacterial adherence and a concordant localized effacement of microvili ( Fig . 1 ) . HeLa cells infected with JPN15 expressing EspT or E110019 displayed extensive membrane ruffling over the entire cell surface ( Fig . 1 ) ; in the vicinity of adherent bacteria the ruffles surrounded and wrapped individual bacterial cells forming structures which appear permissive for internalization . Swiss 3T3 cells infected with JPN15 expressing EspT or E110019 exhibited extensive dorsal ruffles and lamellipodia in addition to localized membrane ruffles at the site of bacterial attachment ( Fig . 1 ) . Caco2 cells infected with JPN15 expressing EspT or E110019 displayed prominent membrane ruffles at the site of bacterial adherence in addition to effacement of micovili ( Fig . 1 ) . These results show that EspT can induce actin remodeling and surface structures , similar to those associated with Shigella and Salmonella invasion ( reviewed in [26] ) . We have recently shown that remodeling of the host cell actin cytoskeleton by EspT is dependent on Rac1 and to a lesser extent Cdc42 [23] . Rac1 and Cdc42 utilize a plethora of downstream effectors in order to regulate cytoskeletal dynamics ( reviewed in [19] and [25] ) . Several GTPase effectors including IRSp53 , N-WASP , Pak , Wave2 and Abi1 have been previously been implicated in formation of membrane ruffles [39] , [40] , [41] , [42] . By using immuno-fluorescence microscopy we found that both Wave2 and Abi1 were present and co-localized with actin at membrane ruffles and the leading edge of lamellipodia induced by EspT ( Fig . 2 ) , while N-WASP was not ( data not shown ) . The signaling protein IRSp53 has been proposed to participate in Abi1-Wave2-Rac1 complex formation [39] , [43] . While we did not detect any significant enrichment of IRSp53 in lamellipodia , IRSp53 was localized to membrane ruffles nucleated by EspT ( Fig . S2 ) . Taken together these results show that Abi1 and Wave2 are localized to membrane ruffles and lamellipodia induced by EspT but IRSp53 is only recruited to EspT-induced membrane ruffles . Wave2 is a ubiquitously expressed member of the WASP super family of actin regulators which potently activates the Arp2/3 complex [44] . The Wave family of proteins have a modular structure consisting of a N terminal Wave homology domain ( WHD ) , a central proline rich region ( PRR ) and a C terminal Arp2/3 binding domain ( VCA module ) ( reviewed in [45] ) . The WHD domain has been shown to bind Abi1 [42] and the PRR has been shown to interact with the SH3 domain of IRSp53 [39] . We utilized siRNA in order to determine if Wave2 is essential for formation of the EspT-dependent membrane ruffles . Depletion of endogenous Wave2 from Swiss 3T3 cells , confirmed by Western blotting ( Fig . 3A ) , resulted in a marked decrease in formation of membrane ruffles and lamellipodia induced by JPN15 expressing EspT or E110019 , compared with cells treated with scrambled siRNA ( Fig . 3B ) . In order to determine which of the Wave2 domains is required for formation of lamellipodia and membrane ruffles by EspT , we transfected Swiss 3T3 and HeLa cells with full length Wave2 or dominant negative forms of Wave2 lacking the WHD ( ΔBP ) or the acidic Arp2/3 interacting domain ( ΔA ) . Transfected cell were infected for 1 . 5 h with JPN15 expressing EspT and the presence of lamellipodia or membrane ruffles was assessed . Mock transfected cells or cells transfected with full length Wave2 had lamellipodia and membrane ruffles on 80 to 90% of infected cells ( Fig . S3 ) . In contrast , transfection with either the ΔBP or the ΔA Wave2 dominant negative constructs resulted in significant reduction in lamellipodia and membrane ruffle formation ( Fig . S3 ) . This result demonstrates that binding of Arp2/3 to Wave2 is essential for EspT-mediated formation of lamellipodia and membrane ruffles . Furthermore , the observation that the N terminal truncated Wave2 ΔBP construct has a dominant negative effect suggests that the WHD motif is also required for EspT mediated cytoskeletal rearrangements . The fact that the Wave2 ΔBP construct , which is capable of binding IRSp53 but not Abi1 , is not sufficient to induce EspT dependent actin remodeling further indicates that IRSp53 does not play a prominent role in EspT mediated signaling . Induction of membrane ruffles is a mechanism employed by a range of pathogenic bacteria in order to facilitate cell invasion . This method of bacterial invasion is referred to as the trigger mechanism and relies upon induction of actin polymerization to form an entry foci and a macropinocytic pocket ( reviewed in [26] ) . JPN15 expressing EspT and E110019 induce host cell membrane remodeling which is reminiscent of entry foci and membrane ruffles induced by Shigella and Salmonella ( reviewed in [26] ) ( Fig . 1 ) . We used differential staining to visualize invasion of Swiss 3T3 cells by JPN15 , JPN15 expressing EspT , and E110019; Salmonella enterica serovar Typhimurium strain SL1344 was used as a control . In addition , we conducted gentamycin protection assays to quantify cell invasion of Swiss 3T3 , HeLa and Caco2 cells after 3 h infection . Differential immuno-fluorescence staining and gentamycin protection assays were also performed in HeLa and Swiss 3T3 cells infected for 6 h with wild type C . rodentium , C . rodentium ΔespT and complemented C . rodentium ΔespT . For immuno-fluorescence extracellular bacteria were stained pre cell permeabilzation with primary anti O127 ( JPN15 ) , anti O111 ( E110019 ) , anti O152 ( C . rodentium ) or anti LPS ( S . Typhimurium ) antibodies and a secondary antibody coupled to a Cy3 fluorophore ( red ) . The cells were then permeabilized and total bacteria were stained with the same primary antibodies and a secondary antibody coupled to a Cy2 fluorophore ( green ) , Alexafluor 633 phalloidin and Dapi were used to visualize actin and DNA respectively . Adherent JPN15 bacteria were homogenously stained by both the extracellular and total bacterial probes , indicating that this strain is not significantly invasive ( Fig . 4A ) . In cells infected with JPN15 expressing EspT or E110019 a significant proportion of the bacteria were labeled with the total bacterial stain but not by the extracellular probe ( Fig . 4A ) . S . Typhimurium-infected cells exhibited characteristic membrane ruffling at the entry foci and a high proportion of bacteria were labeled only with the total bacterial probe ( Fig . 4A ) . The quantitative gentamycin protection assay revealed that JPN15 does not efficiently invade HeLa , Swiss 3T3 or Caco2 cells , exhibiting an invasion rate of less than 1 . 5% ( Fig . 4B ) . JPN15 expressing EspT was significantly more invasive with an invasion rate of 15 . 5% in Swiss 3T3 , 14 . 3% in HeLa and 7 . 2% in Caco2 cells ( Fig . 4B ) . E110019 invaded Swiss 3T3 , HeLa and Caco2 cells at a rate of 9 . 2% , 11 . 4% and 5 . 8% respectively . The invasive capacity of EPEC was significantly less than S . Typhimuriumin ( Fig . 4B ) . Infection of HeLa cells with JPN15 expressing EspTW63A for 3 h confirmed that the WxxxE motif plays a major role in membrane ruffling and cell invasion ( Fig . S4 ) . E110019 is multi drug resistant , which limits the ability to genetically modified the isolate . In order to determine if cell invasion is mediated by EspT , we infected Swiss 3T3 cells for 6 h with wild type C . rodentium and C . rodentium ΔespT . Infection with wild type C . rodentium resulted in membrane ruffles and cell invasion , while the espT mutant exhibited neither ( Fig . 4B ) . Complementing the mutant with espT expressed from pACYC184 ( pICC489 ) restored membrane ruffle formation and cell invasion ( Fig . 4B and S5 ) . In order to confirm that EspT can promote EPEC invasion of non-phagocytic cells independently of other T3SS effectors we ectopically expressed EspT in HeLa cells prior to infection with EPEC ΔescN , a T3SS null mutant . Cells ectopically expressing EspT displayed membrane ruffling which facilitated the uptake of ΔescN bacteria ( Fig . S6 ) . No membrane ruffles were observed in cells ectopically expressing EspTW63A ( data not shown ) . These results show that EspT induces EPEC cell invasion by a trigger mechanism , analogous to that of Shigella and Salmonella . As actin remodeling by EspT is dependent on activation of Rac1 , Cdc42 [23] and Wave2 ( Fig . 3 ) , we utilized dominant negative constructs of these signaling proteins and Wave2 siRNA to monitor the effect on invasion of JPN15 expressing EspT and E110019 . Swiss 3T3 cells transfected with dominant negative Rac1 ( Rac1N17 ) , Cdc42 ( Cdc42N17 ) , Wave2ΔA truncated in the acidic Arp2/3 interacting region and Wave2ΔBP lacking the WHD were infected for 3 h . The cells were fixed and stained for bacterial invasion as described above . Cells transfected with Cdc42N17 were still permissive of bacterial invasion while cells transfected with the Rac1N17 , Wave2ΔA or Wave2ΔBP dominant negative constructs were not ( Fig . 5B ) . Depletion of Wave2 using siRNA in Swiss 3T3 cells significantly reduced the invasive capacity of both JPN15 expressing EspT and E110019 compared to cells treated with non-targeting siRNA ( Fig . 3A and 5B ) . Thus , Rac1 , Wave2 and Abi1 are essential mediators of EspT-induced bacterial invasion . After the initial invasion of host cells internalized bacteria are often bound within a vacuole which resembles early endosomes ( reviewed in [46] ) . Intracellular bacteria either remain within the vacuole or rapidly escape to the cytoplasm [26] . In order to determine whether invasive EPEC are bound within a vacuole or free in the cytoplasm HeLa cells were infected with JPN15 , JPN15 over expressing EspT ( pICC461 ) and E110019 for 5 min up to 24 h and stained with various vacuolar markers including Early Endosome Antigen 1 ( EEA1 ) , Vacuolar ATPase ( VATPase ) and Lamp1 . Internalized JPN15 expressing EspT and E110019 were labeled with EEA1 while external bacteria and JPN15 lacking EspT were not ( Fig . 6 shows staining at 45 min post infection ) . EEA1 staining was apparent after 5 min and persists up to 1 h post infection ( data not shown ) . At 3 h and up to 12 h post infection the EPEC containing vacuole ( ECV ) was labeled with VATPase whilst external bacteria were not ( Fig . 7B ) . Similarly to the Salmonella containing vacuole ( SCV ) , a subset of ECVs became enriched with the lysosomal glycoprotein Lamp1 after 16 h ( Fig . S7 and S8 ) and appear to adopt a perinuclear localization ( Fig . 6B ) . In order to determine if EPEC bacteria can multiply intracellularly we infected Swiss 3T3 cells with E110019 for 30 min before extracellular bacteria were killed by gentamycin . Infected cells were fixed for immuno-fluorescence microscopy at 2 , 8 , 16 and 24 h post infection . We observed a time dependent increase in the level of intracellular bacteria suggesting that internalized EPEC can multiply within host cells ( Fig . S8 ) . After escaping from the vacuole many intracellular pathogens such as Shigella , Burkholderia and Listeria utilize specialized outer membrane proteins to recruit actin nucleating factors in order to produce a propulsive force ( reviewed in [26] ) . EPEC is synonymous with actin nucleation which leads to formation of Tir-dependent actin rich pedestals [12] . During the course of this study we observed that invasive EPEC were associated with filamentous actin comets reminiscent of pedestals . Confocal X-stacks confirmed that the intracellular EPEC bacteria were associated with pedestal-like filamentous actin structures ( Fig . 7A ) . In order to determine if Tir was localized at the actin nucleation sites , we infected HeLa cells for 1 h with JPN15 , JPN15 expressing EspT and E110019; following washes the cells were treated with gentamycin for a further 8 h . The cells were then stained with anti-VATPase and anti-Tir antisera in conjunction with phalloidin and Dapi staining . HeLa cells infected with JPN15 exhibited extracellular , pedestal-associated , bacteria which were associated with Tir but not with VATPase ( Fig . 7B ) . In contrast , internalized JPN15 expressing EspT and E110019 bacteria were co-labeled with anti-VATPase , actin and Tir ( Fig . 7B ) . Similarly , invasive C . rodentium also formed intracellular pedestals ( Fig . S5 ) , while C . rodentium Δtir was invasive but failed to trigger actin polymerization ( Fig . S8 ) . In addition , the intracellular EPEC ΔescN , internalized by ectopically expressing EspT , were not associated with actin pedestals ( Fig . S6 ) . These results suggest that the actin filaments associated with EPEC contained within the ECV is nucleated in a Tir-dependent mechanism analogous to pedestal formation by extracellular bacteria . In order to confirm this assertion we infected HeLa cells with E110019 for 2 h and processed the cells for transmission electron microscopy ( TEM ) . The TEM confirmed that E110019 bacteria are internalized via ruffle formation ( Fig . 7C ) . E110019 can also be seen forming multiple pedestals with the membrane on opposing surfaces during ruffle formation and closure ( Fig . 7D ) . Moreover , internalized EPEC bacteria contained within the ECV are associated actin pedestals , which are strikingly similar to those normally associated with extracellular EPEC ( Fig . 7C–E ) . Interestingly , bacteria bound within ECVs can form multiple pedestals around their circumference ( Fig . 7E ) . In order to determine if the formation of intracellular pedestals by A/E pathogens plays a role in bacterial replication and survival within host cells we infected Swiss cells with wild type C . rodentium and C . rodentium Δtir for 1 . 5 h and with E110019 for 30 min . Extracellular bacteria were then killed by a gentamycin wash and the cells incubated for a further 6 , 12 or 24 h . We observed that both wild type C . rodentium and E110019 were capable of intracellular replication whereas the C . rodentium Δtir mutant failed to replicate and instead exhibited a slow decline in bacterial numbers over time ( Fig . S8 ) . These results demonstrate that formation of pedestals by invasive A/E pathogens may play a functional role during intracellular survival .
A/E pathogens have been long considered to be extracellular bacteria which do not invade mammalian cells [47] . However , sporadic reports have shown that atypical EPEC strains can invade non-phagocytic cells [36] , [48] . The invasive ability has been linked to intimin-Tir mediated tight association of EPEC with the host cell membrane which is hypothesized to produce immature phagocytosis cups leading to a passive push effect and inefficient internalization [35] , [36] . In this study we demonstrated for the first time that EPEC can actively invade non-phagocytic cells by inducing formation of membrane ruffles , defining a new category of invasive EPEC . Furthermore we demonstrate that this phenomenon is dependent on the T3SS effector EspT which has previously shown activate Rac1 and Cdc42 [23] . We also show that both actin remodeling and invasion is dependent upon a functional EspT as JPN15 expressing a EspTW63A failed to induce membrane ruffles or to invade . Importantly , in a previous report we indicated that expression of EspT might not confer bacterial invasion of epithelial cells [23] . However , these experiments were conducted using EPEC E2348/69 , which in contrast to JPN15 , C . rodentium and E110019 , forms tight microclonies that mask the invasion phenotype ( data not shown ) . Intracellular pathogens have evolved a variety of mechanisms to promote invasion of mammalian cells , including the trigger ( employed by Salmonella and Shigella ) and zipper ( employed by Yersinia and Listeria ) mechanisms ( reviewed in [26] ) . The trigger invasion mechanism is characterized by formation of actin rich membrane ruffles at the site of bacterial attachment , which are regulated by Rho GTPases , particularly Rac1 and Cdc42 and other cytoskeletal regulators such as PI3K [49] . Shigella and Salmonella utilize T3SS effector and translocator proteins such as IpgB1 and IpaC and SopB and SopE/2 to hijack host cell GTPase and phospho-inositol signaling to modulate membrane ruffling and formation of the macropinocytic pocket [28] , [29] , [30] , [37] . Importantly , although both IpgB1 and EspT belong to the WxxxE family of effectors and play a prominent role in bacterial invasion by inducing membrane ruffles , we have recently shown that they activate Rac1 by distinct mechanisms [23] . Downstream of Rho GTPase signaling , membrane ruffle formation is nucleated by the WASP super family proteins including N-WASP and Wave2 . Salmonella invasion has been demonstrated to be at least in part dependent upon the Arp2/3 binding activity of Wave2 and also the association of Wave2 with Abi1 [50] . Wave2 cannot bind Rac1 directly; two different mechanisms have been proposed to describe how a Wave2-Rac1 complex is formed . Innocenti et al and Steffen et al demonstrate that Wave2 binds to Abi1 and two accessory proteins PIR121 and Nap1 which mediate Rac1 binding [42] , [51] . A report by Miki et al proposed that IRSp53 is the protein which links Rac1 to Wave2 [39] . In this study we demonstrate that EspT activation of Rac1 leads to a downstream recruitment of Wave2 , Abi1 and IRSp53 to membrane ruffles . Depletion of endogenous Wave2 using siRNA resulted in a significant reduction in both the level of membrane ruffles induced by strains expressing EspT and their associated invasive capacity . We also show that the Arp2/3 and Abi1 binding regions of Wave2 , but not N-WASP , are required for EspT-induced membrane ruffles and invasion . Furthermore , a construct of Wave2 which retained the IRSp53 and Arp2/3 binding regions but lacked the Abi1 interacting domain had a dominant negative effect on membrane ruffle formation , suggesting that IRSp53 is not required for , but may play an accessory role in , EspT-mediated actin rearrangements . Once internalized Shigella and Listeria quickly escape the vacuole ( reviewed in [52] ) . In contrast , Salmonella remains vacuole bound and utilizes different virulence factors to modify the vacuolar environment , position and interaction with the host endomembrane system in order to create an intracellular replicative SCV ( reviewed in [53] ) . In this study we demonstrated that after invasion EPEC is bound within a vacuole ( ECV ) and remains vacuolated until at least 16 h post infection . We found that the ECV is EEA1 positive for up to 1 h post infection and progresses to being VATPase positive from 3 h to 12 h post infection . Furthermore , 12 h after infection the ECV appears to adopt a peri-nuclear position , which resembles the properties of the SCV . Similarly to the SCV ( reviewed in [53] ) we found that a subset of ECVs become enriched in the lysosomal glycoprotein Lamp1 ( data not shown ) indicating lysosomal fusion with the ECV . We also demonstrate that internalized EPEC bacteria can survive and replicate within host cells in a time dependent manner . Importantly and uniquely , we found that the ECV is associated with filamentous actin tails , which are reminiscent of the extracellular pedestals normally nucleated by EPEC stains . Formation of intracellular actin pedestal were essential for bacterial survival , as the intracellular population of invasive tir mutant declined over time . Formation of extracellular pedestals is dependent upon the T3SS effector Tir [11] . The interaction of Tir with intimin triggers recruitment of the mammalian adaptor Nck which in turn recruits and activates N-WASP leading to Arp2/3 recruitment and actin polymerization [11] , [54] , [55] , [56] . In this study we found that internalized EPEC can localize Tir to the vacuolar membrane in a T3SS dependent manner and that the localization of Tir can promote actin nucleation . Furthermore , we found that a C . rodentium Δtir mutant is still invasive but does not form intracellular pedestals , demonstrating that pedestal formation by internalized bacteria is a Tir-dependent process analogous to that of extracellular bacteria . Additionally using TEM we found that invasive EPEC bound within a vacuole are associated with intracellular pedestals around the circumference of the bacteria . Interestingly , membrane ruffles seen engulfing invading EPEC were occasionally associated with pedestals , suggesting the pedestals can be formed during or after internalization . Canonically actin is recruited to the surface of intracellular pathogens which are non-vacuolated and this recruitment is mediated by outer-membrane proteins which are free to interact with host cell signaling molecules present in the cytoplasm . For example following escape from the vacuole Shigella and Listeria utilize IcsA/VirG and ActA , respectively , to trigger actin polymerization and motility ( reviewed in [57] ) . The Vaccinia virus uses the viral membrane protein A36R in order to generate actin based motility in a similar manner to the extracellular EPEC pedestals [58] . Importantly , the SCV is also associated with an actin nest which is required to maintain the integrity of the vacuole and support the intracellular replication of Salmonella ( reviewed in [59] ) . Due to the positioning of the actin extensions around the entire circumference of EPEC it is unlikely these intracellular pedestals are involved in classical actin-based motility . However , there are reports suggesting that actin polymerization and depolymerization around the periphery of E-cadherin-coated beads can lead to directional movement in process referred to as flashing [60]; for this reason at this stage we cannot rule out the possibility that intracellular pedestals confer actin based motility . Furthermore , formation of intracellular pedestals by invasive EPEC may play a role in maintaining the vacuole integrity in a similar way to that described for other vacuolated pathogens [59] . To the best of our knowledge the current study demonstrates for the first time that an intracellular bacteria is able to recruit filamentous actin comets to the pathogen cell surface whilst encapsulated in a vacuole . In order to survive within intracellular niche vacuolated bacteria must evade host cell lysosome mediated degradation . Interestingly , during the course of this study we observed that internalized EPEC , which were enclosed in ECVs , displaying strong actin staining around their circumference were rarely Lamp1 positive , whereas ECVs which had little or no actin polymerization associated with them were homogenously labeled with Lamp1 ( data not shown ) . Furthermore , we observed that a C . rodentium Δtir mutant was attenuated for intracellular replication . We propose that formation of actin rich intracellular pedestals around the circumference of the ECV by invasive EPEC may constitute a physical barrier to lysosome fusion protecting the enclosed bacteria from lysosomal degradation; however this hypothesis requires further testing . A similar phenomenon has been described for the trafficking of endosomes and lysosomes to wounded sites of plasma membrane . At sites of plasma membrane disruption lysosomes and endosomes are recruited to seal the breach , this process is inhibited if the cortical actin meshwork is stabilized and enhanced when it is disrupted [61] . Similarly the lysosome dependent internalization of Trypanosoma cruzi requires a depolymerization of the cortical actin network to allow lysosome transit to the plasma membrane [62] . Recently , while screen ca . 1000 clinical EPEC and EHEC isolates we found that none of the EHEC strains and only 1 . 8% of the EPEC strains contain espT [21] . Interestingly , espT was found in EPEC E110019 which was linked to a particularly sever outbreak of gastroenteritis in Finland [22] . E110019 was found to be particularly infectious and unusually for EPEC was associated with person to person spread and adult disease [22] . Although we have no clinical data of the other espT positive isolates it is tempting to speculate that the expression of EspT could be at least in part responsible for the hyper virulence of the E110019 strain . Further studies of the invasive EPEC category are needed to assess the risk they pose to human health .
Bacterial strains used in this study are listed in Table 1 . The C . rodentium ΔespT were constructed using the using the one-step PCR λ-red-mediated mutation protocol [63] The O111:H2 E110019 strain was isolated from an outbreak in Finland [22] . All the strains were maintained on Luria–Bertani ( LB ) broth or agar supplemented with ampicilin ( 100µg/ml ) or Kanamycin ( 50 µg/ml ) . Plasmids used in this study are listed in Table 2; primers are listed in Table 3 . espT was amplified by PCR using E110019 genomic DNA as template and cloned into pSA10 [64] ( primer pair 1 and 2 ) . All constructs were verified by DNA sequencing . Site directed mutagenesis of EspT was carried out using a Quickchange® II kit ( Stratagene ) and primers 3 and 4 according to the manufacturers instructions . Plasmids pSA10::espT was used as template for the mutagenic reactions . The pCX340 vector encoding EspT-TEM fusion was constructed after amplification of espT from C . rodentium using primer pair 5 and 6 . The mammalian expression vector pRK5 containing one of RacN17 or Cdc42N17 dominant negatives used in the transfection assays were a gift from Nathalie Lamarche-Vane . The pRK5 encoding Wave2 , and the pDSRED Wave2ΔA and Wave2ΔBP were a kind gift from Laura Machesky via Ray Carabeo . 48 h prior to infection cells were seeded onto glass coverslips at a density of 5×105 cells per well and maintained in DMEM supplemented with 10% FCS at 37°C in 5% CO2 . Caco2 cells were grown in DMEM supplemented with 20% FCS at 37°C in 5% CO2 . The cells were washed in PBS and the media changed every 24 h for 12 days until the cells polarized . 3 h before infection , the cells were washed 3 times with PBS , the media replaced with fresh DMEM without FCS supplemented with 1% mannose and 500 µl of primed bacteria were added to each well , the plates were then centrifuged at 1000 rpm for 5 min at room temperature and infections were carried out at 37°C in 5% CO2 . HeLa cells were seeded on to glass coverslips for infection as previously described above . Wild type EPEC E2348/69 and ΔescN T3SS null mutant were transformed with the pCX340 vector encoding EspT-TEM-1 fusion; an NleD-TEM fusion was used as a positive control . Translocation assay was performed as described previously [65] . Coverslips were washed 3 times in PBS and fixed with 3% Paraformaldehyde ( PFA ) for 15 min before washing 3 more times in PBS . For immuno-staining , the cells were permeabilized for 5 min in PBS 0 . 5% Triton X100 , washed 3 times in PBS and quenched for 30 min with 50 mM NH4Cl . Pre prermeabilized samples were not treated with triton X100 . The coverslips were then blocked for 1 h with PBS 0 . 5% BSA before incubation with primary and secondary antibodies . The primary antibody mouse anti EEA1 ( BD biosciences ) and mouse anti VATPase ( Gifted by Prof . D . Holden ) were used at a dilution of 1∶100 , while rabbit anti O127 , anti O111 , anti O152 , anti Tir and goat anti CSA-1 ( salmonella LPS , gifted by Prof D . Holden ) were used at a dilution of 1∶500 . Rabbit anti Wave2 ( SantaCruz Biotechnology ) and Mouse anti Abi1 ( Abcam ) were used at 1∶200 dilutions . Coverslips were incubated with the primary antibody for 1 h , washed three times in PBS and incubated with the secondary antibodies . Donkey anti-rabbit IgG conjugated to a Cy2 or Cy3 fluorophore , donkey anti-mouse IgG conjugated to a Cy5 or Cy5 fluorophore , donkey anti goat IgG conjugated to a Cy2 or Cy3 fluorophore ( Jackson laboratories ) were used at a 1∶200 . Actin was stained using AlexaFluor 633 phalloidin , Oregon Green phalloidin or Rhodamine phalliodin ( Invitrogen ) at a 1∶100 dilution . All dilutions were in PBS/0 . 5% BSA . Coverslips were mounted on slides using ProLong® Gold antifade reagent ( Invitrogen ) and visualized by Zeiss Axioimager immunofluorescence microscope using the following excitation wavelengths: Cy3 – 550nm , Cy5 – 650nm and Oregon Green – 488nm . All images were analyzed using the Axiovision Rel 4 . 5 software . Confocal X stacks were taken using a Leica Sp2 microscope . Cell boundaries were determined using actin staining and Abobe photoshop software . Swiss 3T3 cells or HeLa cells were transfected with pRK5 encoding EspT , RhoAN19 , RacN17 , Cdc42N17 dominant negatives fused to a Myc tag , pDSRED encoding Wave2 , Wave2ΔA or Wave2ΔBP by lipofectamine 2000 ( Invitrogen ) according to the manufacturer's recommendations . The cells were incubated at 37°C in a humidified incubator with 5% CO2 for 16 h , washed twice in PBS before having their media replaced with DMEM as described previously . Transfected cells were infected with the appropriate strain as described above . HeLa cells were seeded at a density of approximatetly 5×106 cells per well 24 h prior to transfection of either Wave2 siRNA pool or a non-targeting pool supplied by Dharmacon using HiPerFect ( Qiagen ) according to the manufacturers instructions . The media was changed 16 h after transfection and the cells were allowed to recover for 12 h before being trypsinated and seeded at a density of 5×106 cells . The siRNA procedure was repeated for a total of 3 rounds before the cells were used . Levels of Wave2 and tubulin were then detected by western blotting using anti wave2 ( Santa Cruz ) and anti tubulin ( Sigma ) antibodies . Cells were then infected with the appropriate strain and processed for immuno-fluorescence microscopy as previously described . Cells seeded into the wells of a 24 well plate were infected as described above for 6 h at 37°C in 5% CO2 . The pre-gentamycin plates were washed 5 times in PBS and then permeabilzed for 15 minutes with 1% saponin in sterile water before plating in triplicate on LB plates in dilutions ranging from 100 to 10−7 . The post gentamycin samples were washed 5 times with PBS after the final wash the PBS was replaced with serum free DMEM containing 200µg/ml of gentmaycin and the cells incubated for 1 h at 37°C in 5% CO2 . The plates were then washed a further 5 times in PBS before permeabilization and plating as described above . The pre and post gentamycin plates were then incubated for 15 h in a static 37°C incubator and the colony forming units ( cfu ) were counted . The percentage of invasion was calculated based on the ratio of cfu on the pre and post gentamycin plates . Glass coverslips were seeded and infected for 2 h with the appropriate strains as described above . The cells were washed 3 times in phosphate buffer pH7 . 2 and then fixed with 2 . 5% Gluteraldehyde ( Agar ) in phosphate buffer pH7 . 2 for 15 min . The coverslips were then washed with phosphate buffer pH7 . 2 a further 3 times before being post fixed in 1% Osmium Tetroxide for 1 h . The cells were then washed 3 times in phosphate buffer before being washed for 15 min in graded ethanol solutions from 50% to 100% to dehydrate the samples . The cells were then transferred to an Emitech K850 Critical Point drier and processed according to the manufacturer's instructions . The coverslips were coated in gold/palladium mix using a Emitech Sc7620 minisputter to a thickness of approximately 370Å . Samples for scanning electron microscopy ( SEM ) were then examined blindly at an accelerating voltage of 25 kV using a Jeol JSM-6390 . 6 well plates were seeded and infected for 2 h with the appropriate strains . The cells were washed 3 times in phosphate buffer pH7 . 2 and then fixed with 2 . 5% Gluteraldehyde in phosphate buffer pH7 . 2 for 15 min . The plates were then washed with phosphate buffer pH7 . 2 a further 3 times before being removed from the plate using a Teflon scraper and subsequently harvested into in eppendorf tube . The eppendorfs were then centrifuged at 10 , 000 RPM to pellet the cells . The cell pellets were post fixed in 1% Osmium Tetroxide for 1 h , followed by 1% buffered tannic acid for 30 min and then a 1% aqueous sodium sulfate rinse for 10 min . The sample was dehydrated using an ethanol-propylene oxide series ( with 2% uranyl acetate added at the 30% step ) and embedded in Epon-araldite for 24 h at 60°C . Ultrathin sections ( 60 nm ) were cut with a Leica EMUC6 ultramicrotome , contrasted with uranyl acetate and lead citrate , and viewed with an FEI 120-kV Spirit Biotwin TEM . Images were obtained with a Tietz F415 digital TemCam .
|
Enteropathogenic E . coli ( EPEC ) is an important diarrheal pathogen responsible for significant infant mortality in the developing world and is increasingly associated with sporadic outbreaks in the developed world . The virulence strategy of EPEC revolves around a conserved Type 3 secretion system ( T3SS ) which translocates bacterial effector proteins directly into host cells . EPEC is considered to be a non-invasive pathogen which intimately adheres to host cells and polymerizes actin rich pedestals on which extracellular bacteria rest . Recently we have identified the T3SS effector EspT which activates the mammalian Rho GTPases Rac1 and Cdc42 , resulting in the formation of membrane ruffles and lamellipodia . In this study we dissect the signaling pathway utilized by EspT to nucleate membrane ruffles and demonstrate that these ruffles can promote EPEC invasion of host cells . Furthermore , we show that internalized EPEC are bound within a vacuole . We also report for the first time the ability of a bacterial pathogen to form actin comet tails across a vacuole membrane . In addition to providing novel insights into the subversion of cellular signaling by invasive pathogens , our study also breaks the long held dogma of EPEC as an extracellular pathogen and will have implications on how future EPEC infections are diagnosed and treated .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"microbiology/cellular",
"microbiology",
"and",
"pathogenesis"
] |
2009
|
The T3SS Effector EspT Defines a New Category of Invasive Enteropathogenic E. coli (EPEC) Which Form Intracellular Actin Pedestals
|
Cilia are evolutionarily conserved hair-like structures with a wide spectrum of key biological roles , and their dysfunction has been linked to a growing class of genetic disorders , known collectively as ciliopathies . Many strides have been made towards deciphering the molecular causes for these diseases , which have in turn expanded the understanding of cilia and their functional roles . One recently-identified ciliary gene is ARL2BP , encoding the ADP-Ribosylation Factor Like 2 Binding Protein . In this study , we have identified multiple ciliopathy phenotypes associated with mutations in ARL2BP in human patients and in a mouse knockout model . Our research demonstrates that spermiogenesis is impaired , resulting in abnormally shaped heads , shortened and mis-assembled sperm tails , as well as in loss of axonemal doublets . Additional phenotypes in the mouse included enlarged ventricles of the brain and situs inversus . Mouse embryonic fibroblasts derived from knockout animals revealed delayed depolymerization of primary cilia . Our results suggest that ARL2BP is required for the structural maintenance of cilia as well as of the sperm flagellum , and that its deficiency leads to syndromic ciliopathy .
Cilia are short , protruding organelles often referred to as “signaling hubs” . These microtubule-based structures are involved in diverse functional roles , and impairment of their structure or function often leads to a class of genetic diseases known as “ciliopathies” [1] . Cilia contain a highly organized structure , consisting of a 9+0 ( motile and immotile ) or a 9+2 ( motile ) microtubule arrangement , starting with triplet tubules at their base ( basal body and transition zone ) , doublet tubules throughout the axoneme , and singlets at their tip [2 , 3] . Despite the retention of this core structure throughout the body , cilia in each tissue are modified to impart unique functionality , a feature that reflects the broad range of ciliopathy phenotypes . For instance , the cilium of photoreceptors , the light-sensing neurons of the retina , has the function of both connecting the different segments of the cell ( the inner and the outer segment ) and of allowing the transport of proteins and metabolites across these compartments [4] . Meanwhile , cilia present in the embryonic node during gastrulation can bend and sense fluid-flow , the event responsible for the left/right-patterning of the visceral organs [5] . On the other hand , the sperm flagellum is the longest in the body ( ~100μm in mice ) and possesses additional accessory structures not present in other cilia ( illustrated in S2 Fig ) , including the transient microtubular-based structure , the manchette , which assists in axoneme growth [6] . Ciliary defects have been associated with retinitis pigmentosa ( RP ) , a progressive form of retinal degeneration leading to loss of photoreceptor cells and vision . A recent study has associated nonsense mutations in the polyglutamylase gene TTLL5 with retinal degeneration and male infertility [7] . In support of these findings , Ttll5 knockout mice , in addition to visual impairment , show flagella that are detached from the sperm head , disrupted axoneme patterns with loss of tubulin doublets , and a severe loss of motility in sperm cells [8 , 9] . Similarly , defects in the Intraflagellar Transport Protein 27 ( IFT27 ) are linked with the RP-associated Bardet-Biedl syndrome and cause sperm malformations leading to infertility in mice [10 , 11] . One protein previously linked to RP is the ADP-Ribosylation Factor Like 2 Binding Protein ( ARL2BP ) [12–14] . In agreement with the cilia-related phenotypes , ARL2BP localizes to the connecting cilia of photoreceptor cells [15] , and our previous findings showed that loss of ARL2BP results in abnormal doublet microtubule structure of the axoneme and shortened cilia in photoreceptor cells [16] . In this study , we report the identification of two homozygous mutations in the gene ARL2BP in three Portuguese patients from two consanguineous families displaying RP and male infertility . The murine knockout model for the same gene showed similar phenotypes , including retinal degeneration , immotile sperm cells and impaired spermatogenesis , as well as situs inversus and increased brain ventricular volume . Our data highlight a novel ciliopathic entity linking two structurally similar , yet functionally different , ciliary organelles—the photoreceptor connecting cilium and the spermatozoon flagellum , associating vision and the reproductive system .
The index subject P1 ( ID: LL1 ) was initially evaluated at age 40 and diagnosed with retinitis pigmentosa . Born from a consanguineous union ( Fig 1A ) , and originating from Portugal , he was first diagnosed with myopia at age 8 and developed night blindness and photopsia by the age of 26 . Over the following 10 years , he developed a progressive loss of vision and bilateral constriction of the peripheral visual fields . The patient noticed a more pronounced reduction in visual acuity at age 37 , associated with onset of photophobia ( both eyes ) , and later , metamorphopsia in the right eye . At age 42 , his best corrected visual acuity ( BCVA ) was 0 . 2 in the right eye ( OD ) and 0 . 25 in the left eye ( OS ) . Slit-lamp examination revealed moderate opacities of the crystalline lenses and generalized chorioretinal atrophy , including punched-out lesions , affecting the periphery and the posterior pole , upon dilated fundus examination ( Fig 2A ) . Scarce pigment deposits were visible in the peripheral retina as well as bilateral optic disc pallor and attenuated retinal vessels ( defining the characteristic triad of retinitis pigmentosa ) . Fluorescence angiography showed generalized window defects and atrophy of the photoreceptor/pigment epithelium complex . Fundus autofluorescence showed spotty areas of hypo-autofluorescence outside of the vascular arcades and a central ring of hyper-autofluorescence in the posterior pole ( Fig 2C ) . At last examination , at age 43 , marked progression of peripheral visual field constriction was noted in both eyes , which was more severe in the OD ( limited to the central 10–15 degrees ) . Spectral domain optical coherence tomography ( SD-OCT ) revealed pronounced atrophy of the retinal layers , with enhanced visualization of the choroidal vessels ( Fig 2E ) . Electroretinograms ( ERG ) showed no response to light stimuli , either from cone or from rod photoreceptor cells ( S1 Fig ) . The patient’s past medical history revealed cardiac arrhythmia under bisoprolol treatment and infertility . Due to unsuccessful conception , a full spermiogram analysis was performed , the diagnosis resulting in severe asthenozoospermia ( complete absence of motility ) . Table 1 summarizes this analysis , as compared to the reference normal values described by the World Health Organization [17] . The causes of infertility were investigated by exploring possible occurrence of deletions in the Y chromosome , as described for the Sertoli cell-only syndrome ( OMIM: 305700 ) . The presence of 21 genetic markers ( STS ) on chromosome Yq and two markers on Yp was verified ( S1 Table ) , therefore excluding this possibility . A fresh sperm sample , originally collected for immunofluorescence analysis for the present study , showed that most sperm heads were detached from flagellum , and approximately 80% of intact cells had a shorter flagellum ( Fig 3A ) . Nevertheless , the percentage of normal sperm formed was within standard WHO reference values . Patient P2 ( ID: LL89 ) , a woman , was diagnosed with a classical form of retinitis pigmentosa at the age of 36 years . The patient was born from a consanguineous union between first degree cousins of Portuguese origin ( Fig 1A ) . Initial visual complaints started with photophobia at 11 years of age and progressed to night blindness and photopsia by the age of 25 . Constriction of the peripheral visual field was noted at the age of 28 . Disease progression led to a sharp reduction in visual acuity at age 38 and to loss of color vision by the age of 40 . The patient underwent cataract surgery at age 40 and 48 . At 50 she developed metamorphopsia in the OS . At the last ophthalmologic examination , patient P2 showed a BCVA of 0 . 3 in the OD , 0 . 16 in the OS , and severe constriction of the peripheral visual field , with bilateral tunnel-like vision , restricted to the 5 degrees central . Dilated fundus examination revealed attenuated retinal vessels , pale optic discs ( Fig 2B ) and a marked generalized chorioretinal atrophy in the posterior pole and in the periphery , with areas of complete atrophy in the inferior peripheral region of the right eye . Mild to moderate mottled pigment deposition in the retinal periphery , along with some irregular white patches in the posterior pole near the superior vessels , was also observed . Fundus autofluorescence was notable for generalized hypo-autofluorescence with a wide hyper-autofluorescent ring in the posterior pole ( Fig 2D ) . SD-OCT revealed pronounced atrophy of the outer retinal layers and the presence of a foveal cyst ( Fig 2F ) . Patient P2’s paternal aunt ( P3 ) and her uncle ( P4 ) were also diagnosed with RP , showing similar progression and manifestations . Specifically , P3 began to be visually impaired in her 20s with the onset of night blindness . Her visual field progressively shrank for both eyes and considerably worsened after the birth of her two daughters in her 30s , evolving to a state of complete blindness by the age of 50 . There is no clinical history of other ciliopathic disorders in this family , with the exception of P4 , who suffered from chronic bronchitis . This latter patient died of multiple myeloma prior to this study , thus preventing further clinical analysis . Interestingly , P4 never conceived any offspring in spite of multiple attempts ( as reported by his sister , patient P3 ) , whereas both female patients P2 and P3 had two healthy children . None of the patients in this study had self-reported respiratory conditions or subjective complaints ( except for P4 ) , metabolic disturbances , situs inversus or skeletal abnormalities . The DNA of P1 was initially screened for mutations in the TTLL5 gene , due to the similarities in phenotype , but the results were negative . Following whole-exome sequencing ( WES ) , the data were analyzed using an internal in silico pipeline assessing variant frequency in the general population , quality , and predicted impact at the protein level . This filtering resulted in 7 homozygous changes , of which 6 resided in autozygous regions ( Fig 1B ) . Among these 6 variants , a single-nucleotide substitution disrupting the canonical consensus donor splice site downstream of exon 3 in ARL2BP ( NM_012106 . 3:c . 207+1G>A , hg19 ) was identified ( Fig 1C ) . P2 was studied using the same methodology , and 20 of the 23 rare homozygous variants resulting from our pipeline were located in autozygous regions . These included a 4-bp deletion in the coding sequence of ARL2BP exon 1 , resulting in a frameshift and creation of an early stop codon at the beginning of exon 2 ( c . 33_36delGTCT:p . Phe13ProfsTer15 ) ( Fig 1C and 1D ) . P3 was also confirmed to carry the same homozygous mutation ( Fig 1D ) . RT-PCR analysis was performed on cDNA derived from P1’s sperm RNA and compared with RNA from a healthy control donor . Gel electrophoresis revealed the absence of a band corresponding to the expected PCR product in P1 , and the appearance of a smaller fragment ( Fig 4A ) . Sanger sequencing of this latter fragment showed that in P1’s sperm ARL2BP transcripts were abnormally spliced , and that c . 207+1G>A caused the skipping of exon 3 , the fusion of exon 2 with exon 4 , and the shift of the canonical reading frame , leading to premature termination at the 9th codon of exon 4 ( Fig 4B and 4C ) . To determine the role for ARL2BP in spermiogenesis , the final stages of spermatogenesis , we investigated the localization of ARL2BP within the sperm cell . In the human control sample , ARL2BP localized at the base of the flagellum , as well as at the equatorial zone of the sperm head ( Fig 3A ) . In cells from P1 , we observed most of the sperm heads separated from their tails , though some intact sperms had a shortened tail . In the majority of these cells , staining for ARL2BP was non-specific with faint background ( Fig 3A ) . In approximately 2–5% of the spermatozoa from P1 , staining was similar to that of the control sample , suggesting residual expression of the wildtype , correctly spliced ARL2BP isoform in this patient , which often occurs for mutations affecting splicing sites [18–21] . In agreement with findings from human samples , ARL2BP was found in the sperm head at the head-tail connecting apparatus ( HTCA ) and principal piece in murine sperm ( Fig 3B ) . In contrast , no signal was observed in KO murine sperm , demonstrating the specificity of the antibody used . Concurrent with the identification of the male patient , we discovered that male Arl2bp KO mice were infertile , as they were not yielding any litters during the generation of the murine Arl2bp KO model to study blindness [16] . Therefore , we examined sperm motility with live imaging and found that the KO sperm were immotile , in agreement with the human patient phenotype ( S1 and S2 Videos ) . Of note , testis size and weight were comparable between WT and KO ( Fig 5A ) , and Heterozygous males were comparable to WT in every murine examination performed throughout this study . Morphological analysis of the testes by H & E staining revealed normal spermatogenesis in KO animals . However , sperm release into the lumen appeared impaired , with a smaller lumen area , an absence of sperm tails , and an increase in residual bodies ( RB ) ( Fig 5B ) . Inspection of the murine KO sperm revealed a drastically decreased epididymal sperm cell count ( Fig 5C ) , and additional light microscopy images of cauda epididymis sperm revealed that all Arl2bp KO sperm had gross morphological defects , including numerous detached heads and tails , kinked necks , bent tails , stubby tails , abnormal heads , and cytoplasmic bulges attached to the tails ( Fig 5D ) . These results are consistent with a high incidence of sperms with abnormal morphology ( 92% ) reported in patient P1 ( Table 1 , Fig 4A ) . The severity of sperm immotility in the KO mice spurred the investigation into the morphology and formation of the sperm tail and accessory structures . We first examined the sperm tail core ( axoneme ) using microtubule-associated markers ( antibodies listed in Table 2 ) . Despite severe loss in protein levels in sperm lysates from Arl2bp KO animals ( Fig 6A ) , Acetylated Tubulin ( AcTu ) and Glutamylated Tubulin ( GluTu , GT335 ) were found in the sperm tail in WT and KO animals ( Figs 3B and 6D ) . Interestingly , testes cross-sections from KO mice stained with GluTu revealed a further irregularity in tail shapes , as the tails were spiraled ( Fig 6C ) . Retinitis Pigmentosa GTPase Regulator ( RPGR ) and Sperm Flagellar Protein 2 ( SPEF2 ) , both axoneme-associated proteins , displayed highly diminished and spotty staining in KO murine sperm ( Fig 6E ) . These findings show that loss of ARL2BP does not interrupt the initiation of microtubular axoneme growth . However , shortened axonemes and a decrease in axoneme-associated protein localization to the sperm tail indicates that maturation of the axoneme is impaired . To determine if the assembly of the accessory structures is normal in the absence of ARL2BP , we assessed sperm tails using markers such as A-Kinase Anchoring Protein 4 ( AKAP4 , fibrous sheath , FS ) and Outer Dense Fiber Protein 1 ( ODF1 , outer dense fiber ) . Both markers were absent in murine KO sperm , though present in murine KO testes lysates ( Fig 6A and 6E ) . Additionally , both forms of AKAP4 were at the expected distribution in WT animals [22] , with most of the soluble , non-assembled precursor to AKAP4 ( pro-AKAP4 , 82kDa ) in the testes samples , while in the sperm lysates the phosphorylated form of AKAP4 assembled into the FS ( Ph-AKAP4 , 109kDa ) was the most represented ( Fig 6A ) . In contrast , both forms of the protein were present in the KO testes , with neither present in the sperm lysates ( Fig 6A ) . Supporting this finding , staining in KO testes sections revealed retention of AKAP4 in the residual bodies shed during spermiogenesis , and an absence of AKAP4 from sperm tails . In comparison , WT testes displayed AKAP4 in both residual bodies and luminal sperm tails ( Fig 6B ) . Conversely , we observed staining for Heat-Shock Protein 60kDa ( HSP60 , mitochondrial sheath ) in murine KO sperm ( although in reduced amounts ) ( Fig 6E ) . This finding was independently confirmed by the presence of pyruvate dehydrogenase protein ( Pyr Deh . , mitochondrial sheath ) in sperm lysates ( Fig 6A ) . Furthermore , staining of WT and KO testes cross-sections using Mitotracker showed mitochondria in elongating spermatids ( Fig 6C ) , indicating the formation of a mitochondrial sheath . All together , these results demonstrate a failure to complete spermiogenesis by the inability to form the outer dense fiber layer or assemble the fibrous sheath in Arl2bp KO animals . ARL2BP staining in the sperm head and the presence of abnormally shaped sperm heads in the male patient and KO mice prompted our investigation of acrosome development throughout spermatogenesis . The acrosome spreads like a cap around the nucleus in stages V-VII ( cap phase ) , as observed in the WT testes ( Fig 7A ) . However , KO animals displayed acrosomal irregularities exemplified by clusters of acrosomal granules instead of the cap structure ( Fig 7B ) . Furthermore , at later stages in spermatogenesis , there were instances of irregularly shaped acrosomes surrounding nuclei . Fig 7C shows acrosome caps in KO testes at stages VIII-XI that are irregularly shaped instead of round , with a delicate appearance . While at stage XII , there were few misshapen acrosomes displaying a “molar tooth” shape ( Fig 7D ) . Of note , most acrosomes on elongating spermatids in KO testes appear normal , suggesting that ARL2BP is not crucial for acrosome formation . Extensive analysis of ultrastructural images from WT and KO murine testes and epididymis tissue corroborated our earlier observations of decreased sperm count and lack of sperm tails in the KO . WT sperm tails demonstrated the organized structure illustrated in S2 Fig ( Fig 8A and 8B ) . Additionally , spermatogenesis appeared to be relatively normal through early tail formation in spermiogenesis , including formation of the manchette with centrally located centrioles in both WT and Arl2bp KO ( Fig 8C and 8D ) . However , significant abnormalities in head , neck , and tail ultrastructure were noted in later stages of spermiogenesis in Arl2bp KO animals . Longitudinal and cross-sections of Arl2bp KO sperm tails revealed that microtubules were present in parallel arrays in the proximal tail region , but they did not form the canonical 9 + 2 axoneme arrangement ( Fig 8C ) . Microtubules were singlets and were not paired , and some seemed to be incomplete tubules ( Fig 8C2 and 8E2 ) . There was also evidence of uneven thickness in the tail staining with tubulin ( GT335 and AcTu , Fig 4B ) , which coincides with the uneven microtubules presented in the ultrastructural images ( Fig 6E and 6F ) . Nevertheless , most tubules were associated with electron-dense material that appears to be outer dense fibers and/or fibrous sheath ( Fig 8C and 8E ) . The outer dense fibers and fibrous sheath were not properly organized and were scattered in various portions of the tail , though present , which is consistent with the reduced levels of ODF1 and AKAP4 markers ( Fig 8E and 8F ) . The mitochondrial sheath contained centrally located mitochondria , but they were not properly organized ( Fig 8F ) . Additionally , Arl2bp KO sperm at stage IX displayed manchette ( Fig 8D2 ) , and in sperm from testis sections , the basal plate and capitulum were present in the neck , but segmented columns were either absent or severely disrupted ( Fig 8F2 ) . We were unable to detect sperm at later stages in development after disassembly of the manchette and were therefore unable to fully assess manchette structure throughout spermatogenesis . Altogether , these data confirm the necessity for ARL2BP in sperm flagellum structure and formation . Human patients , as well as KO animals , were further examined for additional phenotypes typically observed in ciliopathies . The human patients in this study did not report any symptoms associated with impaired cilia in the trachea or kidney . This finding is corroborated by our observations in Arl2bp KO mice ( S3 , S4 and S5 Videos ) . However , we observed a high number of KO mice displaying situs inversus or heterotaxy , an asymmetric Left/Right ( L/R ) positioning of the internal organs caused by defects in the nodal cilia of developing embryos ( Fig 9A ) . The association of ARL2BP mutations in humans with situs inversus is variable , and the patients reported in this study did not possess situs inversus . However , a previous study identified one patient with a mutation in ARL2BP with situs inversus [15] . In mice , this phenotype is highly penetrant , as most Arl2bp KO animals exhibit situs inversus ( 55% ) , with 28% possessing heterotaxy of either the heart or stomach ( Fig 9B ) . Furthermore , tracking and statistical analysis revealed that the number of KO’s produced from Heterozygous x Knockout or from Heterozygous x Heterozygous parents did not follow Mendelian ratios when examined from full-term litters or from embryonic day 13 . 5 ( e13 . 5 ) ( chi square value of 39 . 19 , p<0 . 001 , and chi square value of 7 , 0 . 01<p<0 . 005 , respectively ) . However , litters examined from e7 . 5 did follow Mendelian ratios ( chi square value of 0 . 96 ) suggesting that there may be partial embryonic lethality mediated by the absence of ARL2BP at or near the time of gastrulation ( after day e7 . 5 ) ( Fig 8C ) . Unfortunately , further studies delineating the timing of embryo loss was not in the scope of this study . Nonetheless , these results establish ARL2BP’s essential role in node-determined laterality during murine development . An additional ciliated tissue investigated in the Arl2bp KO mice was the cerebral ventricles . These are fluid-filled cavities in the brain that contain ciliated tufts on the apical surface of ependymal cells . There are four interconnected cavities , including the two lateral ventricles and third ventricle in the forebrain that are connected to the fourth ventricle in the brain stem by the cerebral aqueduct present in the midbrain . The motile cilia in these cavities are responsible for the circulation of cerebrospinal fluid ( CSF ) from the lateral ventricles , where it is produced , to the surrounding space of the brain and spinal cord . Disruption of the CSF flow results in enlarged ventricles and hydrocephaly due to build-up of CSF in the ventricles . The gross morphology of the rest of the brain appears similar to WT in size and shape . White matter axon tracts and gray matter volume appeared consistent between genotypes , indicating that deletion of Arl2bp does not result in gross structural brain malformations . Remarkably , a 4-fold increase in lateral ventricular volume was observed in the brains of KO mice compared to WT littermates using micro-CT scans ( Fig 9D and 9E ) . Of note , the third ventricle also appeared enlarged in Arl2bp KO mice , however we were not able to accurately quantify this due to problems with segmentation from the surrounding space . Additionally , there was no noticeable difference in the cerebral aqueduct or fourth ventricle between WT and Arl2bp KO mice , nor was there any sign of obstruction in the cerebral aqueduct or fourth ventricle , a common cause of hydrocephaly . Interestingly , the majority of KO mice did not present external signs of hydrocephaly . Though the larger ventricular volume points to a role for ARL2BP in CSF flow regulation , a detailed behavioral analysis would need to be completed to understand if these structural changes cause any behavioral deficits . To assess this phenotype in our patients , we looked for signs pointing to chronic hydrocephalus . None were shown nor reported by patients P1 and P3 . On the other hand , patient P2 suffered from frequent migraines and recently underwent a cerebral CT imaging examination , which revealed normal ventricular volumes and absence of other morphological anomalies . To further investigate the role for ARL2BP , mouse embryonic fibroblasts ( MEFs ) were generated from Arl2bp KO mice and WT littermates . Primary cilia were induced in MEFs after removal of serum from the growth media . After 48 hours of serum starvation , the percentage of ciliated cells was comparable between WT and KO ( Fig 10C and 10D ) . MEFs lacking ARL2BP possessed significantly shorter cilia ( average of 2μm ) than WT MEFs ( average of 2 . 7μm ) ( Fig 10A and 10B ) . Importantly , WT MEFs did express ARL2BP , whereas Arl2bp KO MEFs did not ( Fig 10E ) . To determine if the cilia present in Arl2bp KO MEFs had difficulty in primary cilia depolymerization or re-entry into the cell cycle , cilia resorption was assessed after addition of serum to cilia-induced cells . Interestingly , after 2 hours of serum addition , a significantly higher percentage of KO MEFs retained their cilia than what was observed in WT ( Fig 10C and 10D ) . Further observation of KO and WT MEFs revealed that cilia resorption was not statistically different between them from 6 to 24 hours ( Fig 10D ) . Importantly , cell cycle distribution in KO MEFs was similar to that of WT , as evaluated through cell sorting and observed throughout primary MEF cell maintenance ( S2 Table ) , thus the observed effects on primary cilia statistics are cell cycle independent . These results indicate that loss of ARL2BP affects the initial depolymerization of primary cilia .
Spermatogenesis begins in the basal compartment at the outer edge of the seminiferous tubules and continues inward until fully formed spermatids are released into the lumen and on to the epididymal tissues . The last stage , spermiogenesis , is characterized by sperm tail formation , beginning with axoneme growth and assisted by a transient microtubular structure , the manchette [6] . Without ARL2BP , sperm tails fail to elongate . Morphological analysis with Transmission Electron Microscopy ( TEM ) revealed that the 9+2 axoneme structure was disorganized , with no detectable doublet microtubules ( DMT ) . The outer dense fibers and fibrous sheath structures still accumulate around the tail during spermatogenesis but fail to assemble appropriately . This coincides with the loss of ODF1 ( ODF ) and AKAP4 ( FS ) staining , and the increase of residual bodies present in testes of KO animals , an indicator of failed spermiogenesis . Furthermore , the increased levels of phosphorylated AKAP4 in the testes instead of the sperm of KO mice points to the inability to properly assemble the FS present throughout the principal piece . The presence of p-AKAP4 in the testes lysates is likely related to the fibrous sheath fragments seen in the cytoplasmic bulges present in electron micrograph images of the testes , as well as the AKAP4-containing residual bodies . The misassembled FS is also linked to the irregular axoneme structure , as the ODFs attach directly to their corresponding DMTs in the principal piece [27 , 28] . FS assembly occurs in a distal to proximal direction , following formation of the axoneme . Without the stable interaction and growth of the DMTs and ODFs there is not creation of a true “distal end” . Likely , the smaller , disorganized clumps that show FS periodicity and rib-like appearance in the ultrastructural images is due to the inability of the FS precursor to assemble correctly along the distal axoneme . Furthermore , AKAP4 processing and FS assembly is dependent on the formation and proper localization of the annulus . This is a septin ring-structure formed during spermiogenesis that travels to its position between the mid- and principal pieces after midpiece formation is complete [29] ( S2 Fig ) . If the midpiece forms improperly , the annulus cannot localize appropriately , resulting in abnormal processing of AKAP4 and a failure to complete spermiogenesis . These defects in FS , and thereby principal piece formation , are accompanied by issues in midpiece formation . In the midpiece , the ODF does not directly bind to the microtubule doublets , but it is held in place by , and in early spermatogenesis attached to , the surrounding mitochondrial sheath ( MS ) [30 , 31] . The lack of a properly assembled MS could be attributed to the poorly assembled ODFs . Therefore , it is possible that without ARL2BP , the malformation in the microtubule structure causes impaired assembly of all periaxonemal structures and a failure to complete spermiogenesis . This is evident by the lack of principal pieces ( which assemble last ) in sperm tails from the KO model and in the majority of sperms from the human sample . The manchette is a transient microtubule- and F-actin-based structure formed during spermiogenesis . The formation and assembly of the MS , ODF , FS , and acrosome is mediated by intra manchette transport ( IMT ) , a process resembling intraflagellar transport ( IFT ) [29 , 32 , 33] . Given that the MS , ODF , FS and acrosome assembly depends on ARL2BP , and that the manchette forms in Arl2bp KO sperm , we hypothesize that IMT is impaired or that there is a delay in the manchette disassembly . This hypothesis is supported by animal models that lack SPEF2 , a protein required for the axonemal central pair in sperm flagellum , and MEIG1 , a protein required for spermiogenesis . Loss of either of these proteins results in abnormal head and tail formation , which in both cases are linked to an elongated or delayed manchette [34–39] . Though it is unlikely that ARL2BP directly binds with SPEF2 or MEIG1 , the abnormally shaped heads and impaired assembly of accessory structures with loss of these proteins is similar , indicating that ARL2BP may be involved in IMT or manchette disassembly . Furthermore , we showed ARL2BP localizing in the equatorial zone of the acrosome , as well as abnormal acrosome formation in the Arl2bp KO sperm . In mammals , proteins localizing at the equatorial segment of the acrosome are involved in the initiation of sperm-oocyte fusion [40 , 41] . However , it remains unclear if ARL2BP has any direct role in fertilization . Taken together with the data observed from the WT and Arl2bp KO MEFs , the delay in cilia depolymerization closely relates to the notion that ARL2BP is involved in manchette disassembly . Therefore , we hypothesize that ARL2BP is essential for depolymerization of microtubules in the manchette and MEF primary cilium . We cannot , however , exclude the possibility that lack of ARL2BP in Sertoli cells underlies part , or all , of the phenotypes observed in sperm from patients and Arl2bp KO mice . Considering that Sertoli cells are essential to the regulation of spermatogenesis , and are located directly next to sperm cells throughout their developmental lifetime , it is possible that loss of ARL2BP compromises the ability of these cells to nourish the sperm cells or take up the residual bodies . L/R-asymmetry is determined by the mono-ciliated cells in the embryonic node during gastrulation . There is a leftward-fluid flow generated in this region by motile monocilia , which bend the immotile cilium of the crown cells and signals for asymmetrical protein expression related to eventual organ placement [42 , 43] . Therefore , if this fluid flow is disrupted , the laterality of the organs will be affected . It was shown in an elegant study by Nonaka , et . al . 2002 that complete reversal of the flow ( rightward ) resulted in complete situs inversus [44] . However , what determines situs inversus vs other forms of laterality placements are not known and is thought to be stochastic . To date , one patient with defective ARL2BP was reported to have situs inversus [12] . This laterality defect was also observed in KO mice , but not in patients from the two families of this study , who have normal organ position ( situs solitus ) . However , there was an abnormally high incidence ( 55% ) of situs inversus , compared with normal laterality or heterotaxia in the KO mice . Since the axonemes in photoreceptors and sperm tails are shorter with loss of ARL2BP , it is likely that nodal cilia are also shorter in the KO animal . Interestingly , throughout the breeding process , we noticed a significant decrease in the expected number of KO animals after laterality had been determined ( e7 . 5 ) . Therefore , we hypothesize that some KO embryos possessing heterotaxia died embryonically . Since different forms of heterotaxia are related to embryonic lethality ( heart malformations , congenital heart disease ) , this could be possible [45 , 46] . This could also explain why only one patient identified with ARL2BP mutations possesses situs inversus , while the incidence in the mouse model is much higher . Though these questions could not be answered within the scope of this study , these results provide an interesting avenue for further research . Multi-ciliated cells are present in the ventricles of the brain , throughout the trachea , and in the fallopian tubes . Patients with mutations in ARL2BP were not reported to have any symptoms related to defects in these cilia ( sinusitis , otitis media , hydrocephaly , or female infertility ) . Among patients of this study , only P4 was reported with chronic bronchitis . The late age of onset of this respiratory issue and the fact that the patient is presently deceased , makes it difficult to discern possible association of this to mutated ARL2BP-driven ciliary defects . To note , the mouse model also did not display any related symptoms . Interestingly however , live imaging of dissected trachea of KO mice possessed tufts of normal cilia , immotile cilia , and cilia with uncoordinated beating ( S3 , S4 and S5 Videos ) [47] . Though overall , the tracheal cilia from the KO mice were able to move fluid in one direction , similar to WT . Likewise , the ventricular volumes in KO mouse brains were significantly increased but neither of these defects were enough to cause a discernible phenotype . Therefore , we consider that ARL2BP is not required for the function of multi-ciliated cells , though it does appear to be necessary for the robust performance of these cilia . With our findings , we validate the past association of ARL2BP mutations with the blinding disease RP , and bring the first evidences of ARL2BP involvement in spermatogenesis . We thereby prove that additional cilia-related phenotypes originate from ARL2BP deficiency , with manifestations that are similar in human and mouse . Furthermore , we provide a first insight into the disease mechanisms associated with ARL2BP mutations in relationship to defective ciliogenesis , pointing to an essential role for this protein in the maintenance of normal structure and homeostasis of cilia and flagella .
Protocol # 09/14 approved by the Cantonal Committee ( Vaud Canton ) for Research Activities on Human Subjects , Title: "Molecular Genetics of Ocular Diseases"—written consent was given . On 3/15/2017 Comissão de Ética para a Saúde do Instituto de Oftalmologia Dr . Gama Pinto approved research on human patients . Written consent was given . On 4/28/2017 Comissão de Ética para a Saúde do Instituto de Oftalmologia Dr . Gama Pinto approved a few amendments to the human patient protocol . Written consent was given . This study followed the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health . All animals used in this study were handled and housed according to approved Institutional Animal Care and Use Committee ( IACUC ) protocol # 1803013440 of West Virginia University . The approved euthanasia procedure used was carbon dioxide inhalation followed by cervical dislocation . Patients were recruited from the Instituto de Oftalmologia Dr . Gama Pinto in Lisbon , Portugal . DNA was extracted from peripheral blood leukocytes ( subjects P1 and P2 ) , and from saliva ( subject P3 ) , after obtaining written informed consent . RNA was extracted from a sperm sample from patient P1 . A control sperm and DNA sample were provided by a healthy donor . Ophthalmologic examination included assessment of BCVA , slit-lamp examination , dilated fundus examination , fundus photography , visual fields , and optical coherence tomography ( OCT ) . Full‐field ERGs were also recorded , following the International Society for Clinical Electrophysiology of Vision ( ISCEV ) protocol . Semen analysis ( subject P1 ) was carried out by standard procedures by an andrology laboratory , and according to WHO guidelines [17] . Additional clinical features were assessed based on patients’ clinical history . WES was performed using 2 μg DNA derived from peripheral blood mononuclear cells . Protein-coding regions were captured using the HiSeq Rapid PE Cluster Kit v2 , and an Illumina HiSeq 2500 instrument was used for paired-end sequencing . Single nucleotide variants and small insertions and deletions were detected using the Genome Analysis Tool Kit ( GATK v4 . 0 ) software package , using the Best Practice Guidelines identified by the developers [48] . The pathogenicity of the detected genetic variants was assessed after functional annotation through ANNOVAR [49] and with the mean of in-house scripts [50] . Genomic regions with high homozygosity were determined using the AutoMap software . Primer pairs for ARL2BP exons and flanking intron boundaries were designed using the CLCbio Genomics Workbench ( Qiagen ) . PCR amplification was performed in a 10 μl total volume containing 2 ng genomic DNA or cDNA , 1x GoTaq buffer , 0 . 1 mM dNTPs , 10 μM of each primer , and 5 U/μl of GoTaq polymerase ( Promega ) . PCR products were purified using ExoSAP-IT , USB , or extracted from agarose gel using Nucleospin Gel and PCR Clean-up ( Macherey-Nagel ) . Sanger sequencing was performed by a service provider ( Fasteris , http://www . fasteris . com/ ) . Reverse transcription was performed on 1 μg RNA using a mix of random primers and the GoScript Reverse Transcription Protocol ( Promega ) . Mice were euthanized by CO2 inhalation and testes were dissected and separated from epididymis and fat . Testis samples were weighed and immediately frozen in liquid nitrogen . Prior to analyses , the samples were homogenized in phosphate buffered saline ( PBS , with protease inhibitor cocktail ( Thermo Fisher #A32955 ) ) and a NanoDrop spectrophotometer was used to measure protein concentrations . Samples were analyzed by SDS-PAGE gel , followed by transfer onto polyvinylidene difluoride membranes . After blocking the membranes with western blot blocking buffer ( LiCor 004864 Classic ) for 30 min at room temperature , they were incubated with the primary antibodies overnight at 4°C . The membranes were then washed in PBST ( PBS with 0 . 1% Tween-20 ) 3 times for 5 minutes ( 3 X 5 min ) at room temperature before incubation in secondary antibody ( goat anti-rabbit Alexa 680 , goat anti-rat Alexa 680 , or goat anti-mouse Alexa 800 ) for 30 min at room temperature . After 3 x 5 minutes of washes with PBST , membranes were scanned using Odyssey Infrared Imaging System . For testis , the anesthetized animal was perfusion fixed with 4% PFA and then the testis was dissected out . This was followed by incubation in 30% sucrose/PBS overnight at 4°C . Afterward , testes were incubated in a 1:1 mixture of 30% sucrose in PBS and OCT ( Cryo Optimal Cutting Temperature Compound , Sakura ) for 1 hr , and flash frozen in OCT . Staining was performed following the same protocol as detailed below . For sperm , the animal was euthanized by CO2 inhalation , and the epididymis was collected . The tissue was then minced , and sperm were allowed to swim out for 5–10 minutes ( in the case of knockout , entire volume of liquid was collected ) and collected into an Eppendorf tube . 100μl of sperm suspension was added to a Superfrost Plus slide and allowed to completely dry by placing the slides on a hot plate ( setting 3 , Corning Hot Plate ) . Cells were fixed in 4% PFA for 15 minutes followed by three 5 minute washes in PBS . Ice-cold methanol was then added for 2 minutes , followed by three 5 minute washes in PBS . The cells were then blocked overnight at 4°C ( PBS with 5% goat sera , 0 . 5% TritonX-100 , 0 . 05% sodium azide ) . The next day , the cells were incubated with primary antibody for 2 hours at RT , followed by three 5 minute washes in PBS . The cells were then incubated with secondary antibody for 1 hour , followed by three 5 minute washes in PBS . They were then mounted with ProLong Gold anti-fade reagent and viewed on a Nikon confocal microscope . All of the immunofluorescence performed on testes tissue and epididymal sperm was performed on at least 3 KO animals with corresponding littermate controls . Additionally , at least 2 fields of view were taken for each stain with each animal and representative images were selected . Immunofluorescence for human P1 and control sperm samples was performed as described in previously-published protocols [7 , 51] . The same anti-ARL2BP primary antibody described above was used in the human sperm staining experiments , in addition to a secondary goat anti-rabbit antibody conjugated with Alexa Fluor 488 ( Invitrogen ) . Pictures were taken on a Zeiss LSM 780 confocal microscope . Mice were euthanized by CO2 inhalation and epididymal sperm were dissected out and placed in 1ml PBS . Sperm were collected , spun at 1500 RPM for 3 minutes and re-suspended in 50μl of PBS ( knockout ) or 4% paraformaldehyde ( wild type; and further diluted 1:10 in PBS after 5 min ) . Sperm cells were counted using a hemocytometer . Embryos were harvested at E13 . 5 and separated from each other and the placenta in PBS . The tissues were minced with sterile razor blades . The tissues were then trypsinized with 6ml total of 0 . 25% Trypsin/EDTA . 7ml of MEF media ( catalog # 10-013-CV from Corning ) DMEM containing glucose , pyruvate , and L-glut + 15%FBS , 1% Pen/Strep ) was then added to each tube and spun down at 1200 RPM for 8 minutes . Cells were then added to a 100mm dish containing 25ml of MEF media and allowed to grow overnight at 37°C with 5% CO2 . Cells were maintained for up to 4 passages . To induce cilia formation , cells were grown to 90% confluency and serum starved for 48 hours . Depolymerization of cilia was done by the addition of sera ( after 48 hours of starvation ) , and cells were collected for staining after 2 hours , 6 hours , 12 hours , or 24 hours of serum addition . For immunocytochemistry , cells were fixed with 4% PFA for 15 min ( or -20°C Methanol for 2 minutes for centrosomal staining , acetylated tubulin and pericentrin ) , washed with PBS 3 x 5 minutes , and blocked for 30–60 minutes ( PBS with 5% goat sera , 0 . 5% TritonX-100 , 0 . 05% sodium azide ) . The remaining steps were performed in the same way as outlined in the immunofluorescence section . Mice were euthanized by CO2 asphyxiation and testes were dissected and separated from epididymis and fat . Testes were decapsulated and 1-3mm pieces were placed into 4% paraformaldehyde and 2% glutaraldehyde in 0 . 2M cacodylate buffer overnight at 4°C . Sperm were also collected from epididymal tissues and placed in the same fixative in the same conditions . Testis and sperm samples were then post-fixed in 1% osmium tetroxide , dehydrated in a series of increasing ethanol concentration , and embedded in Embed812 resin ( Electron Microscopy Sciences , Hatfield , PA ) . Thin sections were cut and stained with UranyLess ( EMS , Hatfield , PA ) and photographed on a Hitachi HT7700 TEM ( NSF grant #1229184 ) . TEM analysis was performed on two WT mice and two KO mice . Transcardial perfusions were performed on anesthetized 60 day old mice with 4% paraformaldehyde . The brain was carefully dissected out of the skull and fixed in 4% paraformaldehyde for 2 days at 4°C . The brains were then transferred to stability buffer ( 4% w/v paraformaldehyde ( pH 7 . 2 ) , 4% w/v acrylamide , 0 . 05% w/v bis-acrylamide , 0 . 25% w/v VA044 initiator , 0 . 05% w/v Saponin , in 1xPBS ) for 3 days at 4°C , and then underwent nitrogen desiccation , followed by a 3 hour incubation at 37°C . After a two-day staining in 0 . 1N iodine , the brains were embedded into 3% agarose for imaging . The brains were imaged on a Bruker SkyScan 1272 MicroCT scanner ( Cu 0 . 11μm filter , 1500ms exposure , and 8μm resolution ) , and 3D reconstruction of ventricular volume was performed with Seg3D software . All data are presented as mean ± standard error margin . Immunoblots and ciliated cell counts were analyzed by unpaired , two-tailed t test ( n = 3 ) . For cilia measurements , staining was performed in triplicate with 100 cilia measured for each and data were visualized with the ggplot2 package in R version 3 . 3 . 2 . Image and densitometry analysis were performed using ImageJ 1 . 50i . For ciliated cell counting , staining was performed in triplicate with 100 cells ( and their cilia ) counted for each . Mendelian ratios were analyzed using a chi square test with 1 degree of freedom . Chi square values of 3 . 84 or higher are statistically significant ( p>0 . 05 ) . Our research has been conducted in accordance with the tenets of the Declaration of Helsinki and was approved by the Institutional Review Boards of our respective Organizations . Protocol # 09/14 was approved by the Cantonal Committee ( Vaud Canton ) for Research Activities on Human Subjects , Title: "Molecular Genetics of Ocular Diseases" . On 3/15/2017 Comissão de Ética para a Saúde do Instituto de Oftalmologia Dr . Gama Pinto approved research on human patients . On 4/28/2017 Comissão de Ética para a Saúde do Instituto de Oftalmologia Dr . Gama Pinto approved amendments to the human patient protocol . Written informed consent was obtained from all patients prior to the sample collection . This study followed the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health . All animals used in this study were handled and housed according to the approved Institutional Animal Care and Use Committee ( IACUC ) protocol #1803013440 of West Virginia University . The approved euthanasia procedure used was carbon dioxide inhalation followed by cervical dislocation .
|
The flagellated tails of sperm cells require a stringent developmental process that is essential for motility and fertility . The components that comprise the sperm tail assemble in regulated steps with protein processing , transport , and structural assembly dependent on each other for sperm tail maturity . In this work , we have identified ARL2BP , a previously retinal-associated protein , to be essential for sperm tail development and assembly . We show that without functional ARL2BP in humans or mice , sperm tails fail to develop , starting with the assembly of the core microtubular structure within the tail . Loss of ARL2BP also effects other ciliated cells , indicating a unique role for ARL2BP in ciliary microtubule formation . This research on ARL2BP provides further understanding on the links between vision and fertility . This work also demonstrates how genomic studies for human patients and murine models can coincide to provide greater insight into disease .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"meiosis",
"medicine",
"and",
"health",
"sciences",
"reproductive",
"system",
"microtubules",
"sperm",
"head",
"cell",
"cycle",
"and",
"cell",
"division",
"cell",
"processes",
"reproductive",
"physiology",
"germ",
"cells",
"animal",
"models",
"model",
"organisms",
"experimental",
"organism",
"systems",
"cellular",
"structures",
"and",
"organelles",
"cytoskeleton",
"sperm",
"research",
"and",
"analysis",
"methods",
"acrosomes",
"animal",
"cells",
"chromosome",
"biology",
"animal",
"studies",
"mouse",
"models",
"testes",
"spermatogenesis",
"cell",
"biology",
"anatomy",
"cilia",
"physiology",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"genital",
"anatomy"
] |
2019
|
Mutations in ARL2BP, a protein required for ciliary microtubule structure, cause syndromic male infertility in humans and mice
|
The bacterial PorB porin , an ATP-binding β-barrel protein of pathogenic Neisseria gonorrhoeae , triggers host cell apoptosis by an unknown mechanism . PorB is targeted to and imported by host cell mitochondria , causing the breakdown of the mitochondrial membrane potential ( ΔΨm ) . Here , we show that PorB induces the condensation of the mitochondrial matrix and the loss of cristae structures , sensitizing cells to the induction of apoptosis via signaling pathways activated by BH3-only proteins . PorB is imported into mitochondria through the general translocase TOM but , unexpectedly , is not recognized by the SAM sorting machinery , usually required for the assembly of β-barrel proteins in the mitochondrial outer membrane . PorB integrates into the mitochondrial inner membrane , leading to the breakdown of ΔΨm . The PorB channel is regulated by nucleotides and an isogenic PorB mutant defective in ATP-binding failed to induce ΔΨm loss and apoptosis , demonstrating that dissipation of ΔΨm is a requirement for cell death caused by neisserial infection .
The genus Neisseria is comprised of the human pathogenic species N . gonorrhoeae ( Ngo ) and N . meningitidis , which cause gonorrhea and meningitis , respectively . The attachment of bacteria to epithelial cells results in transfer of the outer membrane porin PorB to the host cell cytoplasmic membrane [1] , [2] and mitochondria [3] , [4] . Infection by Ngo causes loss of membrane potential ( ΔΨm ) across the inner mitochondrial membrane ( IMM ) and release of cytochrome c , which is required for activation of caspases and induction of apoptosis [5] . When expressed in host cells , PorB translocates to mitochondria and efficiently causes the breakdown of ΔΨm , but fails to induce the release of cytochrome c and subsequent apoptosis under these conditions [6] . This suggested that PorB is required , but is not sufficient , to induce apoptosis , and that a second signal is needed to induce cytochrome c release during infection . Pro-apoptotic signals like growth factor withdrawal , DNA damage or cytoskeletal rearrangement lead to the activation of so-called BH3-only proteins , pro-apoptotic members of the Bcl-2 family [7] . Active BH3-only proteins cause the oligomerization and pore formation of Bax and Bak in the outer mitochondrial membrane ( OMM ) [8] , [9] . We recently demonstrated that signaling cascades originating from the initial interaction of gonococci with host cells specifically induce the release of the cytoskeletal associated proteins Bim and Bmf , which are both required for the full induction of apoptosis by gonococcal infection [10] . Bim and Bmf activate proapoptotic Bak and Bax proteins , inducing OMM perforation followed by the release of caspase-activating factors into the cytosol and activation of apoptosis [11] . Thus , Bim- and Bmf-initiated events may act in cooperation with mitochondrial PorB in apoptosis induction . Whereas targeting of PorB to mitochondria and its crucial role in Ngo-induced apoptosis are well established [3] , [6] , the molecular mechanism by which PorB causes loss of ΔΨm remains unknown . Studies with yeast mitochondria have indicated that import of PorB , and also other bacterial porins , into mitochondria might follow the same pathway as the endogenous mitochondrial porin , voltage-dependent anion-selective channel ( VDAC ) [6] , [12] . In addition , as bacterial PorB and mitochondrial VDAC are both classical β-barrel proteins , structural similarities may facilitate recognition of bacterial PorB by the mitochondrial protein import machinery . In general , uptake of newly synthesized proteins from the cytosol is mediated by the TOM complex , the translocase of the mitochondrial outer membrane [13] . Within this complex , the general import pore is formed by Tom40 [14] . At the OMM of yeast mitochondria , both endogenous VDAC [15] and a PorB derivative [6] target Tom40 to enter mitochondria . VDAC and all other mitochondrial β-barrel proteins tested so far are subsequently transferred to the SAM/TOB complex ( sorting and assembly machinery ) in the mitochondrial outer membrane [16] , [17] . Interestingly , the core component of this complex Sam50/Tob55 , shows remarkable similarities to the bacterial outer membrane protein Omp85 [18]; Omp85 has been shown to mediate membrane insertion of PorB and other β-barrel proteins in Neisseria meningitidis [19] , [20] . Considering the obvious homologies between the Omp85 family members [18] , bacterial PorB should be recognized and inserted into the OMM by the SAM/TOB complex . This would also be in agreement with the general rule that β-barrel proteins are found neither in bacterial nor in mitochondrial inner membranes . However , if PorB accumulates in the OMM , it is difficult to explain how it dissipates ΔΨm , since this would require massive ion flux across the IMM . Here , we investigated the role of mitochondrial targeting of PorB during the course of infection-induced apoptosis . Our data demonstrate that the cooperation of PorB and signaling pathways activated by BH3-only proteins induces the release of cytochrome c and the activation of caspases . We show that PorB avoids Sam50/Tob55 and Sam37/Mas37 , two core components of the SAM/TOB complex , to integrate into the IMM . As a result , mitochondria lose their ΔΨm and the structural integrity of the cristae is dramatically altered . We propose that these modifications at the IMM are essential early events in Ngo-induced apoptosis .
Our previous observations that PorB of pathogenic Neisseria efficiently targeted mitochondria and induced ΔΨm loss without triggering the release of cytochrome c ( [3] and Fig . 1A ) , confirmed that these were independent processes , at least in our model . Therefore , we reasoned that PorB-expressing HeLa cells lack signals upstream of mitochondria to efficiently release cytochrome c . Such potential upstream signals are mediated by BH3-only proteins , which we recently identified as necessary factors for gonococci-induced apoptosis [10] . To demonstrate the interplay of PorB-triggered ΔΨm loss and BH3-only protein-induced pathways for the release of proapoptotic factors , we activated Bak in HeLa cells using the cell permeable BH3-mimetic compound BH3I-2 . BH3I-2 activates Bak ( Fig . 1B ) by interfering with its binding to Bcl-XL [21] . We then transfected HeLa cells with the porB gene of strain VPI ( PorBNgo ) and monitored PorB expression , ΔΨm and cytochrome c distribution , using immunofluoresence microscopy . As a control , PorBNmu of the commensal strain N . mucosa , which does not target mitochondria [6] , was expressed . The mitochondria of cells expressing PorBNmu were unchanged in comparison to control cells; in contrast , those expressing PorBNgo lost ΔΨm but stained positive for cytochrome c ( Fig . 1A ) , as previously described [6] . Addition of BH3I-2 had no effect on either the membrane potential or the cytochrome c content of mitochondria of PorBNmu-expressing cells ( Fig . 1A ) . Only PorBNgo-expressing cells stained negative for cytochrome c in the presence of BH3I-2 ( Fig . 1A ) , suggesting that mitochondrial targeting of PorBNgo sensitizes cells for the complete release of cytochrome c upon Bak activation . To test whether BH3I-2 treatment induces caspase cleavage and activation in PorBNgo-expressing cells , western blot analysis was performed to detect active caspase 3 . Caspase 3 remained inactivated in cells treated only with BH3I-2 ( Fig . 1C , lane 3 ) . PorBNgo expression alone resulted in minimal caspase activity; however , upon addition of BH3I-2 a sharp increase in activity was elicited ( Fig . 1C , lanes 7 and 8 ) . Interestingly , dissipation of ΔΨm by treatment of cells with the uncoupling reagent CCCP ( carbonyl cyanide m-chlorophenyl hydrazone ) or Antimycin A , an inhibitor of the electron transport chain , also caused activation of caspases when combined with BH3-I2 treatment ( Fig . 1C , lanes 4 and 6 ) , although to a lesser extent . These results suggested that PorBNgo-induced ΔΨm loss facilitates the release of cytochrome c , which in turn activates caspase-3 . The release of cytochrome c during apoptosis is a multi-step process that requires a complete remodeling of the IMM [22] . Loss of cristae structure and subsequent condensation of the mitochondrial matrix is often detected in mitochondria devoid of ΔΨm [23] , [24] . Accordingly , we tested whether infection and/or PorB expression induce remodeling of mitochondria . Electron microscopy ( EM ) revealed that most of the mitochondria in infected apoptotic cells were highly condensed and appeared dark ( Fig . 2A ) , a phenomenon explained by an increased electron diffraction of the condensed matrix . When HeLa cells were treated with the caspase inhibitor zVAD prior to infection , at least 60% of the mitochondria from these cells displayed a dark matrix and loss of cristae ( Fig . 2A , B ) . Immunogold-labeling of PorB-transfected cells revealed that most PorB containing mitochondria underwent extensive condensation of the matrix and loss of cristae structure ( Fig . 2C , D ) . We therefore conclude that the presence of PorB triggers a major reorganization of the IMM . Since induction of ΔΨm loss is most likely the crucial step in the sensitization of mitochondria by PorB and gonococcal infection , we investigated the underlying mechanisms of this process . Since PorB is structurally similar to other β-barrel proteins , it should be recognized by the SAM/TOB complex [25] , [26] and insert into the OMM . To test this assumption , we isolated mitochondria from HeLa cells , incubated them with radiolabelled PorB and subsequently subjected these mitochondria to carbonate extraction at different stringencies ( pH 10 . 8 and 11 . 5 ) to remove loosely attached PorB . The membrane protein VDAC was present in the pellet at both pH , but the soluble protein Hsp60 was found completely in the supernatant only at more stringent conditions , at pH 11 . 5 . Although the amount of soluble PorB increased with the increase of stringency of carbonate extraction , a fraction of radiolabelled PorB remained associated with membranes even after carbonate extraction at pH 11 . 5 ( Fig . S1A ) . Upon import into mitochondria , the amount of carbonate-resistant PorB increased with time , but only traces of PorB were detected in the pellet after carbonate-extraction of mock samples containing no mitochondria ( Fig . S1B ) . Thus , carbonate-resistant PorB is not formed independently of mitochondria , but instead is a fully imported and membrane-integrated fraction . To determine if OMM proteins are involved in the uptake of PorB , we pretreated the isolated mitochondria with trypsin to remove cytosolic domains of OMM proteins and compared the rate of import in pretreated and untreated mitochondria . The amount of carbonate-resistant PorB after import was significantly reduced in trypsin-pretrated mitochondria , reaffirming the dependence of PorB import on the TOM complex ( Fig . 3A , Fig . S1E and [6] ) . Likewise , PorB import was reduced ( by 20–30% ) in mitochondria of tom40kd-2 cells after 5 days of knockdown induction with Dox ( Fig . 3B ) . Of note , even after knockdown , traces of Tom40 that can function in import are still present in the mitochondria ( Fig . 3B and not shown ) . Importantly , PorB import into mitochondria isolated from pLVTHM cells containing an empty vector was unchanged upon treatment with Dox , excluding a Dox-dependent side effect ( not shown ) . In addition , we have previously shown that import of VDAC into mitochondria isolated from tom40kd-2 cells was reduced , along with other TOM components , including Tom20 and Tom22 [27] . Hence , our findings support the notion that the TOM complex is involved in the import of both VDAC and PorB . We then addressed the possible role of the SAM complex in the import of PorB using HeLa-derived cell lines with inducible knockdown of Sam50 ( sam50kd-2 ) and Metaxin 2 ( mtx2kd-2 ) [27] , a putative mammalian homologue of yeast Sam35 . VDAC import into the mitochondria isolated from these cell lines was clearly restricted after the induction of knockdown , confirming our previous data ( Fig . S2A , B ) , [27] ) and demonstrating the functional downregulation of Sam50 and Metaxin 2 in these mitochondria . Contrary to our expectations , in Sam50- or Metaxin 2-depleted mitochondria , PorB import was unimpeded ( Fig . 3C , D ) . As observed before [27] , in mitochondria with the knockdown of Sam50 , levels of Tom40 were likewise reduced in the range of 40–50% ( Fig . 3C ) . However , we did not see any decrease of PorB import into these mitochondria in spite of the reduction of Tom40 amounts . This can be explained by the fact that even a strong Tom40 knockdown of more than 90% affected PorB import only moderately ( Fig . 3B ) ; a Tom40 reduction of 40–50% might simply not be sufficient to cause any effects . Previously published data had suggested that PorB follows the same import route as VDAC [6] . However , our data indicated that the SAM/TOB complex marks a crucial branching point of the pathways . To investigate further , we used a similar approach to most previous studies [25] , [26] and used yeast to test whether PorB is a substrate of the SAM complex . PorB transport was monitored in isolated yeast mitochondria using radiolabelled PorB . The protease-protected fraction of PorB was only obtained in the presence of mitochondria ( Fig . S1C ) , ruling out a non-specific aggregation of PorB and confirming import into the mitochondria . As reported previously , the homologous PorB of the non-pathogenic strain N . mucosa was not imported into mitochondria ( Fig . S1D; [6] ) . PorB and other β-barrel protein import efficiencies into mitochondria were then monitored . Similar to our findings with human mitochondria ( Fig . 3B , C , D ) , PorB import efficiencies were comparable to WT levels in the bacterial mutant sam50-1 strain ( Fig . 3E ) , whereas in the yeast mutant tom40-4 import was reduced to levels reported for the mitochondrial porin VDAC [15] ( Fig . S1E ) . Import of VDAC into sam50-1 mitochondria was impaired ( Fig . S2C ) ; in contrast , import of the IMM dicarboxylate carrier ( DIC ) remained unchanged ( Fig . S2D ) , supporting a specific role for the SAM complex in the transport of OMM β-barrel proteins [17] , [28] . Similar results were obtained with mitochondria isolated from a Δsam37 strain lacking the Metaxin 1 homologue Sam37/Mas37 [16] , [29] ( Fig . 3G , S2E & S2F ) , an essential factor for the mitochondrial import of β-barrel proteins [16] . Collectively , these findings suggest that PorB , although a β-barrel protein , avoids the SAM complex during import into mitochondria . To determine the precise intramitochondrial location of PorB , we prepared highly purified outer membrane vesicles from yeast mitochondria [30] . Unexpectedly , PorB was not present in the vesicles but only in the pellet fraction ( Fig . S3 , lane OMV vs . lane MPF ) . In fact , using these conditions PorB mainly formed aggregates in the cytosol , compromising further analysis . Therefore , we first imported radiolabelled PorB into mitochondria isolated from the WT strain , and then separated mitochondrial membrane vesicles by sucrose density centrifugation . Imported PorB accumulated partially in a carbonate-resistant form , as was the case in human mitochondria , indicating its integration into the membrane ( Fig . S1A and data not shown ) . Whereas PorB accumulated in higher density fractions of the gradient , the other β-barrel protein , OMM porin VDAC , was found in the upper part of the gradient . The majority of PorB accumulated in the same higher density fraction as the γ-subunit of the IMM ATP synthase ( F1γ ) ( Fig . 4A ) . This fractionation pattern clearly demonstrated that PorB does not accumulate in the OMM and that PorB is at least partially associated with the IMM; however , a significant fraction of the newly imported PorB seems to form aggregates , as suggested by the presence of PorB in the high density fractions 8 , 9 , and 10 ( Fig . 4A ) . Cytosolic PorB aggregates have never been observed in mammalian cells , suggesting that they express PorB in a fully import-competent state in vivo . The immediate dissipation of ΔΨm in HeLa cells shows that PorB enters the mitochondria in an active form [6] . We expressed PorB in HeLa cells and determined its sub-mitochondrial localization by immunogold electron microscopy . The vast majority of the PorB gold particles localized to the IMM and matrix , similar to the endogenous IMM protein Tim23 ( Fig . 4B , C ) . For comparison with an endogenous OMM marker , we included labeling of Tom22 , which showed a clearly different distribution ( Fig . 4B , C ) . Taken together , these findings show that PorB avoids interactions with the SAM machinery and is instead directed to the intermembrane space ( IMS ) . The distribution of PorB in HeLa cells indicates that , at least in human mitochondria , PorB preferentially associates with the IMM . It is known from previous multichannel studies that PorB is able to form pores of high conductance [2] . Whereas single channel data have been reported on N . meningitidis PorB [31] , the investigations on gonococcal PorBs were restricted to multichannel recordings . However , only single channel analysis can clarify whether gonococcal PorBs can form a high-conductance channel in the IMM . Upon addition of purified PorB to either side of a planar lipid bilayer with lipid composition corresponding to the IMM [32] , single-channel currents were readily detected ( Fig . 5A ) . Although PorB exhibited a dynamic gating behavior with a multitude of conductance states ( Fig . 5A & Fig . S4 , insert ) , a main conductance state of Λ = 420±15 pS was also identified from a linear current-voltage relationship under symmetrical buffer conditions ( Fig . S4 ) . Previous observations with multi-channel recordings [2] indicated a voltage-dependent gating behavior of PorB , raising the question whether PorB channels remain open or closed under the physiological conditions present in the IMM ( i . e . at ΔΨm of around 150 mV [33] ) . Gating transitions were rarely observed at lower voltages ( not shown ) ; PorB channels exhibited a typical three step channel closing at voltages above 60 mV ( Fig . 5A ) . Considering the ΔΨm of the IMM and the voltage dependence of PorB , it is obvious that PorB channels incorporated into the IMM would in fact be arrested in a closed state and the capability of PorB to uncouple the mitochondria would be limited . Since it is known that PorB shows an affinity for nucleotides [2] we analyzed the effect of ATP on PorB channels in detail . The concentration of ATP in mitochondria is known to range from 0 . 6 to 6 . 0 mM [34] . After addition of physiologically relevant amounts of ATP to PorB-containing bilayers , two intriguing effects were observed . First , the amplitude of the single-channel conductance was significantly reduced ( Fig . 5A , B , D ) accompanied by a drastic change in PorB gating behavior ( Fig . 5A , D ) : The channel did not display the typical three step gating transitions but remained mainly in an open state exhibiting increased gating frequency , manifested as flickering ( Fig . 5D ) . The second and physiologically even more important effect was the predominant loss of the voltage-dependent closure of PorB upon ATP addition ( Fig . 5C ) . In intact mitochondria , where ATP is abundant , the channels would be forced to stay open even at a ΔΨm of 150 mV , allowing the flux of large currents across the IMM and subsequent rapid dissipation of ΔΨm . The effects were verified by adding ATP to both sides of the membrane ( Fig . 5B , C , left-most , centre-left ) or separately to either the trans or the cis compartment , to test for side specificity ( Fig . 5B , C , centre-right and rightmost ) . Since the usual channel closure is inhibited at the side of ATP addition and we do not know the exact orientation in which PorB incorporates into the membrane , it was important to establish that the effect of ATP on the PorB channel was not side-specific ( Fig . 5B , C ) . These observations on single PorB channels allowed us to draw several important conclusions: ( i ) PorB inserts easily into a lipid bilayer , regardless of membrane potential . ( ii ) High concentrations of mitochondrial ATP stabilize the PorB channels in an open state . ( iii ) Considering a typical size and a simplified volume to surface ratio , a single open PorB channel should dissipate ΔΨm in about 0 . 8 ms ( for details see Protocol S1 ) . Next , we tested if PorB can perforate the IMM in living cells using a calcein quenching assay for infected and porB transfected cells ( for details see Protocol S1 ) . Transfection of PorB ( Fig . 5E ) or infection with Ngo ( Fig . 5F & S5 ) led to the loss of calcein staining after cobalt chloride quenching , confirming IMM permeabilization upon PorB translocation . IMM permeabilization was not a consequence of apoptosis induction since preincubation of the cells with the caspase inhibitor zVAD-fmk failed to prevent IMM permeabilization or the loss of ΔΨm upon Ngo infection ( Fig . 5F & S5 ) . In conclusion , these data confirm that PorB is able to form pores in the IMM , leading to its permeabilization and loss of ΔΨm . We have previously shown that PorB binds ATP via lysine residues potentially located in the PorB channel or the loop 3 region [2] . To identify the specific residues involved , we performed site-directed mutagenesis of several lysine residues to generate PorB derivatives deficient in ATP-binding . Mutant neisserial strains , differing only in the PorB derivative expressed [35] , were tested for ATP binding as described previously [2] . Even though the exchange of lysine 98 for glutamine ( PorBK98Q ) reduced ATP-binding by more than 70% in comparison to WT PorB ( PorBNgo ) ( Fig . 6A ) , PorBK98Q still clearly localized to mitochondria in transfected cells ( Fig . S6 ) . Interestingly , ATP did not affect the channel properties of PorBK98Q . Neither the voltage-dependent gating of the PorBK98Q channel ( Fig . 6B , C ) nor the voltage-dependent open probability ( Fig . 6D , E ) was influenced by ATP , confirming that the effects of ATP on the PorB channel depend on lysine residue 98 . To determine the effect of the K98Q mutation in vivo , we infected HeLa cells with isogenic neisserial strains harboring WT PorBNgo ( N920 ) and mutant PorBK98Q ( N886 ) and then measured the loss of ΔΨm by using tetramethylrhodamine ethyl ester perchlorate ( TMRE ) staining and FACS analysis . Despite both strains having similar infection rates , significantly more cells infected with strain N886 carrying mutant PorBK98Q retained their ΔΨm as compared to cells infected with WT neisserial strain ( Fig . 7A , B ) . Consistently , cells transfected with the PorBK98Q expression construct retained their ΔΨm in a number of cases ( Fig . S6 ) . Maintenance of ΔΨm was never observed with the WT PorB construct . Moreover , the potential of PorBK98Q mutant strain N886 to induce apoptosis in HeLa cells was strongly reduced ( Fig . 7C ) and infected cells failed to release cytochrome c from their mitochondria ( Fig . 7D ) . When cells were transfected with WT PorBNgo prior to infection with N886 , they released cytochrome c ( Fig . 7D ) , confirming that induction of ΔΨm loss and a second unknown signal triggered by N . gonorrhoeae infection are required for the induction of host cell apoptosis .
In this study , we demonstrate that PorB induces the reorganization of mitochondrial cristae and sensitizes mitochondria to release cytochrome c in response to infection and BH3-only protein induced signaling pathways . The initial prerequisite for PorB to sensitize infected cells to apoptosis is its ability to cause ΔΨm loss , probably by bypassing the OMM-SAM complex and accumulating in the intermembrane space . Our previous data suggested that release of pro-apoptogenic factors by mitochondria during apoptosis induced by gonococcal infection requires two independent steps: The sensitization of mitochondria and the perforation of the OMM . The cooperative effect of PorB at the mitochondria and separate signals elicited by infected cells could be demonstrated in PorB-expressing cells during infection with strain N886 , which carries the ATP-binding mutant of PorB , PorBK98Q ( Fig . 7D ) . Our previous data also suggested that the infection-induced perforation of the OMM depends on the BH3-only proteins Bim and Bmf and the pro-apoptotic BH1-3 proteins Bak and Bax , activated by the interaction of the pathogen with host cells [10] , [11] . The data presented here now demonstrate that mitochondria-targeted PorB in combination with a BH3-only mimetic compound is sufficient to induce cytochrome c release and the activation of caspases . This statement is supported by our observation that ( i ) only PorBNgo targets the mitochondria and not the targeting-deficient derivative PorBNmu , and ( ii ) only PorBNgo recombinant cells and not the non-recombinant neighboring cells responded to BH3I-2 treatment . It is interesting to note that PorB from meningococcal strain H44/76 overexpressed in HeLa cells fails to induce ΔΨm dissipation but instead protects against apoptosis [36] , whereas PorB from strain Z2491 has a similar uncoupling activity as gonococcal PorB [6] . Recent data suggest that invasive and carrier strains of N . meningitides from the same clonal complex induce or inhibit apoptosis , respectively . PorB from the invasive isolate promotes apoptosis induction [37] , supporting a role for PorB in the life and death decision of Neisseria-infected cells . As shown before [6] , the amounts of PorB are relatively similar in mitochondria isolated from Neisseria-infected cells and cells transfected with PorB . However , the efficiency of transfection is approximately 30–40% , whereas the efficiency of infection is nearly 100% . Therefore , during transfection we estimate that 2- to 3-fold more PorB is present in a cell than during infection . Nevertheless , we observe the similar phenotype , a major reconstruction of cristae structures , in both infected and PorB recombinant cells . In this aspect , the transfection and infection model seem to be comparable . The reconstitution of cristae structures is required for sensitization of mitochondria for the release of cytochrome c . Previous work has shown that the majority of cytochrome c is sequestered in the closed cristal compartments ( 85% ) ; as a result structural changes in mitochondria are required to achieve complete and rapid release of cytochrome c [22] . Moreover , ΔΨm loss and matrix condensation contribute to cytochrome c release [24] . In the context of the biogenesis of endogenous mitochondrial proteins , it is not surprising that we find that the uptake of PorB into mitochondria is mediated by the TOM complex . This complex is involved in the import of practically all mitochondrial preproteins from the cytosol [38]–[40] . However , our observation that the import of PorB is independent from the SAM/TOB complex was unexpected . All previously tested endogenous mitochondrial β-barrel proteins , including VDAC , Tom40 , Sam50/Tob55 , and Mdm10 , were transferred to this complex and subsequently inserted into the OMM [25] , [41] . Only recently , subunit Sam35 of the SAM/TOB complex was shown to recognize a specific sorting signal in the C-terminal part of mitochondrial β-barrel proteins and to initiate the insertion of these proteins into the OMM [42] . PorB is obviously lacking this sorting signal and is therefore not recognized as a substrate by the SAM complex . Interestingly , C-terminal sequences were also reported to be relevant in the sorting of β-barrel proteins in bacteria [43] . Our first attempt at directing PorB to the OMM by introducing a C-terminal segment of VDAC failed ( data not shown ) , probably due to requirements in the tertiary structure of the protein , as shown previously for PorB [6] . Purified PorB spontaneously integrated into liposomes consisting of a broad variety of different lipids , from both the outer and inner mitochondrial membranes . Insertion of PorB into the lipid phase is accompanied by specific conformational changes ( see Protocol S1 and Fig . S7 ) and requires neither additional assembly factors nor energy sources . A similar finding has been reported for the integration of soluble human VDAC into lipid bilayers [44] . Previously , we found significant amounts of PorB were associated with the OMM upon overexpression of the protein in yeast [6]; however , here we found that highly purified OMM vesicles are devoid of PorB . In HeLa cells , PorB shows the similar distribution pattern as the IMM protein Tim23 . Interestingly , the secretin PulD , a β-barrel protein in the Gram-negative bacterium Klebsiella oxytoca , accumulates in the bacterial plasma membrane if assembly in the outer membrane is blocked [45] . We propose a similar mechanism for PorB: PorB trapped in the mitochondrial intermembrane space has the possibility to integrate into the IMM . Since most proapoptotic factors of the mitochondria are stored in the IMS , recent investigations have concentrated on the question of how their release is triggered by the opening of the OMM . However , we found that during Ngo-induced apoptosis , events at the OMM are preceded by essential modifications at the IMM . Considering a typical size and a simplified volume to surface ratio , a single PorB channel would dissipate ΔΨm in about 0 . 8 ms ( for details see Protocol S1 ) . Thus , a single open PorB pore in the IMM is sufficient to short-circuit a whole mitochondrion . At potentials Vm≥±60 mV , the PorB channel is enclosed in membranes with a lipid composition comparable to that of the IMM , but ATP binding arrests the channel in an open state even at voltages physiological for the IMM of about 150 mV . In previous experiments with time-averaged multichannel recordings , we observed an apparent ATP-dependent decrease in the voltage sensitivity of PorB channels [2]; however , the time-averaged analysis of these multichannel recordings did not allow the resolution of single-gating events . Here , our single-channel analyses unequivocally showed that ATP abolishes the voltage-dependent closure of PorB channels . In summary , following insertion into the IMM , PorB is arrested in an open state by ATP at the prevailing ΔΨm , i . e . by the normal physiological activity of the mitochondria . The attenuated phenotype of the neisserial strain N886 expressing PorBK98Q provided clear in vivo evidence for a crucial role of ATP binding in dissipating ΔΨm and inducing apoptosis during infection . Our recent data suggest that cell death induced by Ngo requires the close interplay of two interdependent signaling cascades , one leading to the activation of Bak [11] and the other to the sensitization of the mitochondria by translocated PorB ( Fig . 8 ) . For the latter , PorB utilizes an existing mitochondrial transport pathway , the TOM complex , but bypasses the SAM complex , and activated by ATP dissipates ΔΨm . This intriguing and so far unique example is a consequence of the evolution of the SAM machinery . SAM has diverted so much from the parental Omp85 machinery that it is now unable to recognize and sort PorB into the OMM . In this way the bacterial effector PorB functions as an integral element of the host's cell death machinery .
N . gonorrhoeae N242 ( strain VPI; Opa+ , PorBIA ) [46] and N920 ( strain MS11; Opa+ , PorBIA ) expressing PorB of N242 [35] have been described . N886 ( strain MS11; Opa+ , PorBK98Q ) was constructed using the same strategy as for N920 but a mutant porB gene of N242 was transformed instead of the wildtype derivative [35] . Gonococci were routinely grown on GC agar base plates ( Becton Dickinson , Difco and Remel ) supplemented with Proteose Pepton Nr . 3 ( Difco ) and 1% vitamin mix for 14–20 h at 37°C in 5% CO2 in a humidified atmosphere . Opa phenotypes were monitored by colony morphology under a stereo microscope or by immunoblotting . HeLa cells ( human cervix carcinoma ) were grown in RPMI 1640 ( Gibco ) supplemented with 10% heat inactivated fetal calf serum ( FCS ) in the presence of 5% CO2 . Cells were seeded 24 h before infection and washed several times with RPMI without supplements . Infections were routinely performed at a multiplicity of infection ( MOI ) of 1 without centrifugation . For inhibition of caspases , cells were incubated with 50 µM zVAD-fmk ( Bachem ) for 15 min prior to infection and throughout the respective infection period . Cells were seeded on coverslips and transfected with pCMV-Tag-1 , containing either the porBNgo ( P . IA ) or porBNmu gene with an N-terminal FLAG-Tag [3] , using Lipofectamine 2000 ( Invitrogen ) according to the manufacturer's protocol . Twenty four hours post-transfection , cells were stained with 150 nM MitoTracker ( Molecular Probes ) , dissolved in cell culture media for 30 min at 37°C , washed with phosphate buffered saline ( PBS ) and fixed in 3 . 7% paraformaldehyde ( PFA ) . Fixed cells were permeabilized using 0 . 2% Triton X-100 , and nonspecific binding was blocked by using 1% goat serum . Samples were stained using anti-FLAG ( Sigma ) antibody , followed by detection with fluorochrome-coupled secondary antibodies ( Jackson Immuno Research ) . Samples were analyzed under a Leica confocal microscope using TCS software . 5×105 cells per sample were harvested in 100 µl loading buffer and 20 µl of the protein lysates were separated by SDS-PAGE and transferred to nitrocellulose or polyvinylidenfluorid ( PVDF ) membranes . The following antibodies and sera were used in this study: anti-β-Actin ( Sigma ) ; anti-Bak NT ( Upstate ) ; anti-Bak ( Ab-1 ) ( Millipore ) ; anti-Bax NT ( Upstate ) ; anti-Bax ( 6A7 ) ( BD Pharmingen ) ; anti-cleaved Caspase-3 ( Cell Signalling ) ; anti-FLAG ( Sigma ) ; anti-Hsp60 ( Stressgen Bioreagents ) ; anti-VDAC ( Abcam ) ; anti-Tim23 ( BD Biosciences ) ; antibodies against human Tom40 and Sam50 were a gift from N . J . Hoogenraad , and against mouse Metaxin 2 ( cross-reactive with human Metaxin 2 ) a gift from P . Bornstein . Antibodies against yeast mitochondrial proteins Tom40 and Tim23 were a gift from N . Pfanner and C . Meisinger . Equal loading was routinely confirmed by appropriate loading controls . Quantitative analysis of immunoblots was performed by using the open source software ImageJ ( http://rsbweb . nih . gov/ij/index . html ) . Trypsin treatment of mitochondria was performed as previously described [47] . To induce the knockdown by RNA interference , cells were grown for 5 to 7 days in the presence of 1 µg/ml doxycycline as previously described [27] . Efficiency of the knockdown was assessed by western blot . For trancription/translation purposes , PorB was cloned into pGEM-4Z vector ( Promega ) with two additional methionines at its C-terminus and in vitro transcribed/translated in the presence of 35S-methionine/cysteine ( GE Healthcare ) using the TnT Quick Coupled System ( Promega ) . Mitochondrial isolation and import of proteins were performed essentially as described previously [27] . A detailed protocol is available as Protocol S1 . Mitochondria were isolated from yeast cells as described previously and used for import of 35S-labelled mitochondrial porin ( VDAC ) , dicarboxylate carrier ( DIC ) , or the ATP-synthase γ-subunit ( F1γ ) following standard procedures [15] . A detailed protocol is available as Protocol S1 . Purified azolectin ( 60 mg/ml ) in n-decan or a lipid mixture corresponding to the lipid composition of inner mitochondrial membranes ( 60 m/ml ) [32] in n-decan was used to produce stable planar lipid bilayers by using the painting technique [48] , [49] . Purified PorB was applied directly below the bilayer in the cis chamber . Buffer conditions were symmetrical with 250 mM KCl , 10 mM Mops-Tris ( pH 7 . 0 ) in the cis/trans compartment . Two Ag/AgCl electrodes covered by 2 M KCl-agar bridges were inserted into each chamber with the trans chamber electrode connected to the headstage ( CV-5-1GU ) of a Geneclamp 500 current amplifier ( Axon Instruments ) and this was used as a reference for reported membrane potentials . Current recordings were carried out using a Digidata 1200 A/D converter . Data analysis was performed by self-written Windows-based SCIP ( single-channel investigation program ) in combination with Origin 7 . 0 ( Microcal Software ) . Current recordings were performed at a sampling interval of 0 . 1 ms , filtered with a low-pass-filter at 2 kHz . Voltage ramps were carried out by continuously increasing the voltage at a rate of 5 mV/s . After incorporation of single PorB channels into the bilayer by spontaneous insertion , control currents were recorded . Subsequently nucleotides were added either to both sides or separately to the cis or trans side of the membrane . Interactions between the added nucleotides and PorB channels were examined after stirring the aqueous solutions on both sides of the planar lipid bilayer . Binding of ATP by PorB was assayed by chemically crosslinking radiolabelled ATP to neisserial strains expressing different PorB derivatives as previously described [2] . ATP binding was quantified using AIDA Image Analyzer software . For immuno-EM analysis , the cells were fixed with 3% PFA in stabilizing buffer ( 1 mM EGTA , 4% PEG 6000 or PEG 8000 , 100 mM PIPES pH 6 . 9 ) and embedded in 10% Gelatine/PBS . Small blocks of the samples were infiltrated overnight in 2 . 3 M sucrose/0 . 1 M Na-phosphate buffer . Ultra-thin sections were cut at −120°C with a diamond knife . The sections were transferred onto carbon-coated pioloform-film on TEM-grids . The sections were then blocked and reacted with the primary antibody against FLAG-tag ( Sigma ) , Tom22 ( GeneTex ) , Tom20 ( BD Biosciences ) , Tim23 ( BD Transduction Laboratories ) and secondary antibodies coupled with 6 or 12 nm gold particles . For the analysis of mitochondrial membrane potential cells were harvested by trypsinization and washed with phosphate buffered saline ( PBS ) before staining with 100 nM tetramethylrhodamine ethyl ester perchlorate ( TMRE ) ( Molecular Probes ) in growth media at 37°C , 5% CO2 for 30 min . After staining , cells were washed twice with PBS and immediately analyzed by FACS analysis .
|
PorB is a bacterial porin that plays an important role in the pathogenicity of Neisseria gonorrhoeae . Upon infection with these bacteria , PorB is transported into mitochondria of infected cells , causing the loss of mitochondrial membrane potential and eventually leading to apoptotic cell death . Here , we show that PorB enters mitochondria through the TOM complex , similar to other mitochondria-targeted proteins , but then bypasses the SAM complex machinery that assembles all other porin-like proteins into the outer mitochondrial membrane . This leads to the accumulation of PorB in the intermembrane space and the integration of a fraction of PorB into the inner mitochondrial membrane ( IMM ) . In the IMM , ATP-regulated pores are formed , leading to dissipation of membrane potential and the loss of cristae structure in affected mitochondria , the necessary first steps in induction of apoptosis . Our work offers , for the first time , a detailed analysis of the mechanism by which PorB targets and damages host cell mitochondria .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"infectious",
"diseases/sexually",
"transmitted",
"diseases",
"infectious",
"diseases",
"cell",
"biology/cellular",
"death",
"and",
"stress",
"responses",
"microbiology/innate",
"immunity",
"cell",
"biology/membranes",
"and",
"sorting",
"infectious",
"diseases/bacterial",
"infections",
"microbiology/cellular",
"microbiology",
"and",
"pathogenesis",
"microbiology/medical",
"microbiology"
] |
2009
|
Bacterial Porin Disrupts Mitochondrial Membrane Potential and Sensitizes Host Cells to Apoptosis
|
Mammalian meiocytes feature four meiosis-specific cohesin proteins in addition to ubiquitous ones , but the roles of the individual cohesin complexes are incompletely understood . To decipher the functions of the two meiosis-specific kleisins , REC8 or RAD21L , together with the only meiosis-specific SMC protein SMC1β , we generated Smc1β-/-Rec8-/- and Smc1β-/-Rad21L-/- mouse mutants . Analysis of spermatocyte chromosomes revealed that besides SMC1β complexes , SMC1α/RAD21 and to a small extent SMC1α/REC8 contribute to chromosome axis length . Removal of SMC1β and RAD21L almost completely abolishes all chromosome axes . The sex chromosomes do not pair in single or double mutants , and autosomal synapsis is impaired in all mutants . Super resolution microscopy revealed synapsis-associated SYCP1 aberrantly deposited between sister chromatids and on single chromatids in Smc1β-/-Rad21L-/- cells . All mutants show telomere length reduction and structural disruptions , while wild-type telomeres feature a circular TRF2 structure reminiscent of t-loops . There is no loss of centromeric cohesion in both double mutants at leptonema/early zygonema , indicating that , at least in the mutant backgrounds , an SMC1α/RAD21 complex provides centromeric cohesion at this early stage . Thus , in early prophase I the most prominent roles of the meiosis-specific cohesins are in axis-related features such as axis length , synapsis and telomere integrity rather than centromeric cohesion .
After completing premeiotic DNA replication mammalian germ cells enter meiosis and undergo two meiotic cell divisions without any further DNA replication . Haploid gametes are produced . Meiosis features highly specific chromosome structures and behaviour to ensure proper chromosome segregation , exchange of genetic information , and maintenance of genome integrity ( reviewed in [1] ) . In leptonema the four sister chromatids become increasingly compacted and each pair of sister chromatids forms an axial element ( AE ) , most often characterized by the axial element proteins SYCP2 and SYCP3 . The compacted AEs start to pair in zygonema , i . e . the two homologous pairs ( homologs ) of sister chromatids synapse and form the synaptonemal complex ( SC ) , which is complete in pachynema . The SC thus contains four sister chromatids . Each pair of sister chromatids is held together by cohesins , the two pairs are embedded in SC proteins . Once synapsed , the AEs are called lateral elements ( LEs ) of the SC . The protein SYCP1 is centrally located in the SC between the LEs and serves as a marker for synapsis . Homologous recombination between the two homologs requires the introduction of programmed double strand breaks ( DSBs ) by the topoisomerase-type enzyme SPO11 . These breaks , which can be visualized by staining for double-strand break repair proteins such as the meiosis-specific DMC1 , are introduced in leptonema and are processed into recombination intermediates until pachynema . In diplonema the SC between homologs disassembles , the homologs desynapse , but remain linked through a few chiasmata , the sites of meiotic recombination , until the homologs are separated in anaphase of meiosis I and the recombination process is completed . Mammalian meiocytes express four meiosis-specific subunits of the core cohesin complex in addition to the ubiquitously expressed five cohesin proteins SMC1α , SMC3 , RAD21 , SA1/STAG1 or SA2/STAG2 . The meiosis-specific cohesins include one SMC protein , SMC1β , the two kleisins RAD21L and REC8 , and a stromal antigen protein , SA3/STAG3 ( for recent reviews see [2–5] . Theoretically , 18 distinct protein complex can be formed from combinations of these proteins , and so far , at least 6 different cohesin complexes were reported [6–16] . The spatiotemporal appearance of these complexes and their individual roles throughout meiosis are incompletely understood . Immunofluorescence ( IF ) data derived mainly from staining mouse testis sections or spermatocyte or oocyte chromosome spreads from different stages of prophase I showed distinct patterns of individual cohesin proteins indicating different roles for the various cohesin complexes . The scheme in Fig 1A roughly illustrates the kinetics of presence of individual cohesin proteins in mouse spermatocytes . In mice of both sexes , SMC1α , SMC1β , and SMC3 are associated with unsynapsed ( not yet synapsed ) , synapsed and desynapsed regions in all stages of prophase I . SMC1α is gradually lost from the chromosomes in diplonema and not detected in metaphase I . SMC1β and much of SMC3 remain associated with the centromeric region until metaphase II . STAG3 behaves similarly to SMC3 , and the three kleisins show distinct patterns , which have not yet been entirely clarified as the reports do not agree on all details . Based on imaging studies of spermatocytes it seems as if RAD21 disappears early in prophase I and reappears for a short period in mid to late pachynema and diplonema . REC8 is first seen in preleptotene cells , probably at the onset of premeiotic replication , associates initially all along the spermatocyte chromosomes and remains on centromeres up to metaphase II . RAD21L becomes detectable on chromosomes in leptonema when they start forming AEs , and vanishes at around mid-pachynema . Prior to synapsis , REC8 and RAD21L were observed in a mutually exclusive pattern on the chromosomal axes [8 , 9 , 13 , 15] . Mouse mutants deficient in individual cohesins have revealed very important aspects of their roles . Fig 1B provides an overview of some of the most relevant phenotypes of these cohesin mutants . SMC1β deficient male and female mice are infertile , male meiocytes arrest at a stage when chromosomes have reached an early/mid pachynema structure . With respect to the developmental stage within the seminiferous tubules the cells reached stage IV . It is important to distinguish between the stage of development reached within a section of the seminiferous tubules , and the chromosome features characteristic for a certain stage of meiosis . While in mouse mutants the tubular development may reach a certain more advanced stage , the cells may show chromosome features that are reminiscent of an earlier stage . In other words: the tubules may develop further even though the cells are delayed or blocked in forming the corresponding chromosome structure . Therefore one needs to differentiate between the testis tubule stage and the “chromosomal stage” . Partial loss of cohesion , partial asynapsis , telomere deficiencies , and AEs/SCs that are shortend in length by about half are the prominent phenotypes observed in Smc1β-/- spermatocytes [17 , 18] . In the absence of REC8 both sexes are sterile , the spermatocytes arrest in a late zygonema-like stage based on their chromosomal appearance . Here , synapsis protein SYCP1 is deposited between sister chromatids instead of between homologs [19 , 20] . RAD21L deficient spermatocytes do not properly form AEs and synapsis between homologs is abrogated . Spermatogenesis arrests in a zygonema-like chromosomal stage and the males are sterile , whereas females develop age-related infertility [21] . REC8 and RAD21L double deficient spermatocytes are devoid of AEs and SCs and arrest in a leptonema-like chromosomal stage , defined based on the absence of AEs [22] . A similar dramatic phenotype was recently demonstrated for STAG3 deficient mice . Their spermatocytes feature no or–in case of residual low levels of STAG3 proteins–very short AEs , fail in synapsis , lose some centromeric and telomeric sister chromatid cohesion and are sterile [11 , 12 , 14 , 16] . In the complete STAG3 deficiency , the cells develop maximally to a testis tubular stage IV , but chromosomally they reflect leptotene cells as there are no axes ( Winters et al . , 2014 ) . Very recently , double mutants of Stag3 with either Rec8 or Rad21L were described and show increased centromeric cohesion defects , very short AEs and , in case of the Stag3-/-Rec8-/- strain , are similar in phenotype to the Rec8-/-Rad21L-/- spermatocytes [23] . The different kind and/or severity of the phenotypes of mutants in distinct cohesin proteins indicates that specific cohesin complexes contribute during spermatogenesis to distinct processes , which only partially overlap . The functional complexity of the concert of cohesin complexes in meiocytes , however , is far from being sufficiently understood . To further decipher the function of specific meiotic cohesin complexes in male meiosis , we investigated the roles of meiosis-specific kleisins together with the only meiosis-specific SMC protein , SMC1β . Mouse strains carrying deficiencies in SMC1β and either the REC8 or the RAD21L kleisin were generated . The analysis of these double mutants allowed us to determine whether the kleisins act in an SMC1β-based complex . When there were additive effects of double-deficiencies , this would indicate functions of the kleisins in a separate complex , which must be an SMC1α complex . Indeed , we suggest synergistic action of SMC1α and SMC1β complexes to establish proper AE length , synapsis and to maintain telomere integrity . Both meiotic kleisins act together with the two SMC1 variants in these roles . Very early in meiosis I , i . e . in leptonema , the meiosis-specific cohesins are not strictly required for centromeric cohesion .
An earlier report showed that the third kleisin RAD21 contributes little or not at all to the formation of AEs in spermatocytes , since there are extremely short SYCP3-positive AEs when the other two kleisins were absent , i . e . in Rad21L-/-Rec8-/- spermatocytes [10] . At least in this mutant background and for this central function , RAD21 cannot compensate for the loss of the two meiosis-specific kleisins . Thus , it appeared as if the majority of cohesin complexes in prophase I are based on either RAD21L or REC8 . Here , we used single and double mutants lacking either REC8 or RAD21L only , or in combination with the SMC1® deficiency to dissect the individual contributions of these cohesins to AE formation . Staining for SYCP3 was used to measure axis length at the most advanced spermatocyte stage in each "single-knockout" ( SKO ) and Smc1β-/-Rec8-/- and Smc1β-/-Rad21L-/- DKOs compared to wild type ( wt ) ( Fig 3A and 3B ) . The most advanced stage was assigned based on the appearance and extent of SYCP1 and γH2AX staining , where SYCP1 was present on some chromosomes and the previously diffuse γH2AX signal was reduced to one or two cloud-like structures ( S1A and S1B Fig ) . We took into consideration that some short axes may represent fragments of the same chromosomes and therefore divided the total axes length of a cell by the normal number of chromosomes ( 21 including X and Y separately ) . In case of asynapsis of entire chromosomes the number of axes was increased , and we divided the total axes length by this increased number of axes , since each individual axes–whether synapsed or not–was added to the total axes length of a cell . Asynapsis of entire chromosomes was determined by counting the number of CENP-A signals for centromeres . This number was the same in all mutants at the leptotene stage ( see below ) , and if increased at later stages the CENP-A signals on separate axes indicated asynapsis or loss of sister chromatid cohesion and thus an increased number of axes . The total axes length per cell ( Fig 3B ) was divided by the number of CENP-A positive chromosome axes . In the most advanced spermatocytes of the Smc1β-/-Rec8-/- mutant , the axes are very short with an average axis length of 2 . 351 +/- 0 . 262 ( SD ) μm and thus shorter than the corresponding SKO or wt ( wt: 6 . 56 +/- 0 . 253 μm; Smc1β-/-: 3 . 27 +/- 0 . 451 μm; Rec8-/-: 3 . 67 +/- 0 . 247 μm; Fig 3A and 3B ) . Thus , the removal of REC8 in addition to SMC1β further reduces axis length . Therefore , an SMC1α/REC8 complex should exist , unless one would propose a very distinct role of REC8 independent of any cohesin complex , which is very unlikely . The removal of this SMC1α/REC8 complex supposedly accounts for the additional length reduction in this mutant background . Based on the analysis of the Smc1β-/-Rec8-/- mutant , this SMC1α/REC8 complex contributes a moderate roughly 14% to axis length ( i . e . the further reduction by 0 . 92 μm seen in the Smc1β-/-Rec8-/- mutant compared to the Smc1β-/- mutant ) . Such numbers are obviously approximations only for a normal cell as they reflect the contributions in a mutant background . At any rate , this SMC1α/REC8 complex appears to be a minor complex . This is in agreement with the low efficiency or absence of co-precipitation of SMC1α and REC8 from wt or Smc1β-/- testis extracts reported earlier [9 , 13 , 17] . Despite the moderate reduction is axis length in Rec8-/- and Rad21L-/- cells , the total axes length per cell is similar to wt ( Fig 3B ) . This originates from the high levels of asynapsis in these mutants . From the almost total reduction of axis length in STAG3 deficient spermatocytes [16] it is clear that cohesins determine the entire axis length . In the Smc1β-/-Rec8-/- mutant with only 2 . 35 μm of axis length left , these remaining axis–app . 36% of wt length only–must also be provided by some cohesin complex ( es ) . Thus , the remaining app . 36% of axis length that still exists in Smc1β-/-Rec8-/- spermatocytes has to be supported either by SMC1α/RAD21 or SMC1α/RAD21L complexes , the only remaining complexes . Due to non-homologous associations of AEs and to gaps in SYCP3-positive AEs , the measurement of axis length in Rad21L-/- spermatocytes is very difficult , but an estimate that only takes non-associated , gap-less and clearly identifiable axes into account yields a length roughly comparable to that of the Rec8-/- strain ( Fig 3B ) . In Smc1β-/-Rad21L-/- spermatocyte spreads we observed very short SYCP3-stained axes , which often appeared as dots rather than as filaments; they measured 1 . 17 +/- 0 . 27 μm ( Fig 3A and 3B ) . Thus , in contrast to the Rad21L-/- or Smc1β-/- SKOs , the removal of SMC1β and RAD21L almost entirely abolishes formation of SYCP3-positive axes , with no obvious compensatory effect . This suggests that besides SMC1β complexes , an SMC1α/RAD21L complex contributes to axis formation . This further suggests that an SMC1α/RAD21 complex contributes little if any to axes length . In at least one report anti SMC1α precipitation co-precipitated RAD21L [13] . Since the combined loss of RAD21L and REC8 also causes almost complete loss of axes [10] , this supports the above notion that an SMC1α/RAD21 complex does not significantly contribute to axis length and RAD21 cannot compensate for the loss of the two other kleisins , at least under conditions where other cohesins are absent . Similarly , since the Smc1β-/-Rad21L-/- and the Rec8-/-Rad21L-/- [10] DKOs essentially abolish axis formation , but the Smc1β-/- , the Rec8-/- and the Rad21L-/- SKOs do not , the SMC1α/RAD21L and the SMC1β/REC8 complexes must very prominently if not almost entirely support axis formation , with the above mentioned minor contribution of SMC1α/REC8 . A potential role of RAD21 should therefore be mostly confined to other , specific functions such as supporting pachynema/diplonema events , perhaps formation of chiasmata , consistent with the reappearance of RAD21 seen in some studies at this stage . It should be noted that all numbers provided here as percentage of contribution to axis length are based on comparison with the wt situation . In any mutant , compensatory mechanisms may arise that may affect these numbers such as increased expression or stability of the remaining cohesin complexes . Thus , conclusions are qualitative and only roughly quantitative . However , any compensatory effect , if it exists , may be very minor , since axes are reduced to almost dots in the Smc1β-/-Rad21L-/- spermatocytes , and a very similar observation has been made in a STAG3 deficient mutant–no rescue by other STAG proteins [16] . Therefore , the numbers suggest the relative importance of particular complexes to axis length not only in the specific mutant backgrounds , but very likely also with respect to wt cells . In any case , the analysis reveals the presence and functional capacities of certain cohesin complexes in spermatocytes . Synapsis is impaired in all mutants as co-staining for SYCP1 and SYCP3 showed ( S1B and S1C Fig ) . However , it is not possible to precisely quantify the extent of synapsis for the individual mutants , since in absence of REC8 the SYCP1 deposits between sister chromatids yielding a “false” signal , and in the Smc1β-/-Rad21L-/- mutant very small axes or dots appear . Many ( app . 70% ) of these extremely small structures carry SYCP1 , but the small size precludes quantification . Co-staining for HORMAD1 , an asynapsis marker , and SYCP3 confirmed the extent of synapsis failure in all the mutants , since HORMAD1 is present in all cases ( S1D Fig ) . Detailed analyses of SYCP3- and SYCP1-stained axes of the wt and DKO spermatocytes by super-high resolution OMX microscopy ( Fig 3C ) showed the expected central localization of SYCP1 between the two SYCP3 axes in wt cells . In the Smc1β-/-Rec8-/- mutant , where no synapsis occurs and the SYCP3-stained axes therefore consist of two sister chromatids , the deposition of SYCP1 between the sister chromatids is clearly observed and is in agreement with previous reports that described this phenotype for the Rec8-/- strain [19 , 20] . In the Smc1β-/-Rad21L-/- mutant , SYCP1 is also deposited between sister chromatids , but in addition some individual SYCP1-positive sister chromatids were observed , typically 7 per cell ( +/- 5 . 0; n = 40 ) . This data indicates moderate loss of sister chromatid cohesion in this mutant and show that SYCP1 does neither require two sister chromatids in cohesion nor chromatids in SYCP3-mediated close proximity , nor two AEs , to associate along chromosomes . The deposition of SYCP1 at a single chromatid also suggests , that in mutant backgrounds SYCP1 deposition is not necessarily an indicator for synapsis . Therefore we prefer not to designate SYCP1 deposition on pairs of sister chromatids as "synapsis between sister chromatids" . In wt mouse spermatocytes , the sex chromosomes X and Y only pair at a short , centromere-distal region called the pseudo-autosomal region , PAR [24] . The largely unsynapsed sex chromosomes form a special chromatin domain , the sex body , which features silencer chromatin marks . This X/Y association is seen only in wt cells ( Fig 3A , arrow ) , as is the characteristic sex body chromatin staining by γH2AX as one intense structure with the sex chromosomes embedded ( S1A Fig , arrow ) . Synapsis of homologs depends on programmed double-strand breaks ( DSBs ) , which are repaired with progressing synapsis . Cohesin SMC1β is not required for generation of DSBs , but was shown to support their repair , which is delayed in absence of SMC1β [25] . DSBs can be visualized by staining of DSB repair proteins such as RAD51 or the meiosis-specific DMC1 . In all mutants reported here , DMC1 foci and thus DSBs are produced ( Fig 4 ) . Quantification of DMC1 in the DKOs is very difficult as the axes are short or only dots exist where one cannot distinguish individual foci , particularly in the more advanced stages . Therefore we cannot provide exact numbers . The initial numbers of DMC1 foci , as much as recognizable , appeared to be very similar in the mutants and not unlike wt . The same was observed for RAD51 foci . In the more advanced stage of Smc1β-/-Rec8-/- cells the number of DMC1 foci is reduced to two or more foci per short axis . A similar reduction was observed in Rec8-/- spermatocytes but not in Smc1β-/- spermatocytes , where repair is delayed as previously reported [25] . One may speculate that in the Smc1β-/-Rec8-/- spermatocytes alternative repair pathways such as between sister chromatids supported by SYCP1 localized between sister chromatids are enhanced or that the foci are less stable . We also found several discrete DMC1 foci in Smc1β-/-Rad21L-/- spermatocytes , but due to the extremely short axes we cannot clearly distinguish foci . We observed a reduction of DMC1 signals in Smc1β-/-Rad21L-/- spermatocytes in the advanced stage , indicating that repair of DSBs happens or that the foci are not stable . The analysis of centromeres of meiotic chromosomes reveals both , synapsis at centromeres and for centromeric sister chromatid cohesion . In wt pachynema spermatocytes 21 centromere signals indicate complete synapsis ( except the sex chromosomes with their PAR-distal and thus non-synapsed centromeres ) and complete centromeric sister chromatid cohesion . The appearance of 22 to 40 centromere signals indicates either loss of synapsis , or loss of centromeric cohesion in presence of full synapsis , or partial loss of both cohesion and synapsis . More than 40 centromere signals indicate loss of some or all ( 80 signals ) centromeric cohesion and synapsis . In early zygotene cells , where no synapsis exists and thus this analysis of cohesion is less perturbed by synapsis , more than 40 centromere signals ( from 40 AEs , i . e . 2 x 19 autosomes plus X and Y ) indicate loss of centromeric cohesion . Since the DKOs develop to an early/mid zygonema-like chromosomal stage , we analyzed all cells at the leptotene/early zygotene stage to be able to test for centromeric cohesion independent of synapsis , i . e . we expect 40 centromere signals in wt cells . Previously it has been shown that depletion of REC8 causes synapsis failure . However , REC8 is not required for the establishment of centromeric sister chromatid cohesion in meiocytes as in the most advanced Rec8-/- spermatocytes never more that 40 centromeres were observed [19] . We also observed in average 37 . 76 ( +/- 1 . 787 , n = 42 ) centromeres in Rec8-/- spermatocytes by staining with anti-centromeric antibodies ( ACA ) ( S2 Fig ) , which recognize centromeric and pericentromeric heterochromatin . Staining for CENP-A , an inner centromere component showed 40 . 5 ( +/- 2 . 84; n = 34 ) signals in leptonema Rec8-/- spermatocytes ( Fig 5A and 5B ) . This indicates proper centromeric cohesion despite absence of REC8 . In Smc1β-/-Rec8-/- spermatocytes we observed an average of 36 . 12 ( +/- 5 . 930; n = 33 ) ACA signals , and 41 . 4 ( +/- 4 . 5; n = 21 ) CENP-A signals ( Fig 5A and 5B; S2 Fig ) , again suggesting maintenance of centromeric cohesion , although the very slight increase in CENP-A signals may hint at a minor trend towards weakening of centromeric cohesion . The situation is similar in the other mutants: Rad21L-/- spermatocyte spreads showed 30 . 44 ( +/- 4 . 330; n = 52 ) ACA and 39 . 8 ( +/- 6 . 4; n = 59 ) CENP-A signals on average in accordance with a previous report that used ACA [21] . The low number of distinguishable ACA signals may be caused by telomere fusions ( see below ) , which can bring the relatively broad centromeric heterochromatin in very close proximity . CENP-A , which provides a more specific signal just at the inner kinetochore , would not be as much affected . Smc1β-/-Rad21L-/- spermatocytes showed 39 . 09 ( +/- 5 . 39; n = 46 ) ACA signals and 41 . 0 ( +/- 4 . 6; n = 22 ) CENP-A signals . Statistically significant differences were not found for CENP-A signals , and only some differences between certain pairs of mutants were statistically different for the ACA signals with a p-value >0 . 05 . Overall there appears to be slightly more variation in centromere numbers in the DKOs , but this suggests very limited if any weakening of centromeric cohesion . This data indicates that none of the three meiotic cohesins analyzed here are required for early prophase cohesion . Thus , an SMC1α complex , i . e . SMC1α/RAD21 , must provide most if not all centromeric cohesion at this very early stage of male meiosis . This is consistent with the notion that SMC1β is only expressed after entry into meiosis [7] , and is in agreement with the only partial loss of centromeric cohesion in okadaic acid-induced metaphase I Smc1β-/- spermatocytes when these were derived from zygonema cells . When the cells originated from early/mid pachynema , complete loss of centromeric cohesion was observed [17] . This suggests that SMC1β complexes are loaded onto meiotic chromosomes during prophase I , or at least become cohesive then , and successively take over the duty of maintaining centromeric cohesion from the SMC1α/RAD21 complex , which progressively vanishes after entry into meiosis . How can one reconcile this with our analysis of Smc1β-/-Spo11-/- and Smc1β-/- spermatocytes , which showed some centromeric cohesion deficiency in absence of SMC1β [25] ? In that study , spermatocytes of the most advanced stages were analyzed , i . e . late zygonema for Smc1β-/-Spo11-/- or early/mid pachynema for Smc1β-/- . At this stage , partially synapsed axes carrying three separate centromeres were observed in Smc1β-/- cells , which clearly indicate both , synapsis failure combined with loss of cohesion at the centromeres in at least 8% of the cases . If centromeric synapsis fails , the one strong signal derived from 4 very closely juxtaposed centromeres falls apart into two signals; 3 or 4 signals then indicate loss of cohesion . In Smc1β-/-Spo11-/- chromosomes , which do not synapse , one or two ACA signals were observed and indicated loss of cohesion in about a third of the cells . We suggest that while at leptonema the SMC1α/RAD21 complex still provides most centromeric cohesion , it is partly replaced by SMC1β complexes when synapsis starts to happen , i . e . in zygonema . This interpretation would fit to the above notion of loading of SMC1β complexes onto centromeres after entry into meiotic prophase I . In the absence of SMC1β , SMC1α and SMC3 localize to AEs of early prophase I cells [17] suggesting that the SMC1〈/SMC3 heterodimer forms complexes with RAD21L , REC8 or RAD21 or either of them in vivo . Evidence from immuno precipitation experiments differs somewhat between distinct reports [9 , 13 , 15 , 17] , but neither of the three complexes can be surely excluded at this time . We analyzed chromosomal localization of the remaining cohesin proteins in Smc1β-/-Rec8-/- and Smc1β-/-Rad21L-/- spermatocytes in comparison to wt and the SKOs ( Fig 6; for single channel images see S3 to S9 Figs ) . The one cohesin present in all known cohesin complexes , SMC3 , localizes to the chromosomes in all mutants , although at much reduced amounts in the two DKOs . The pattern also becomes more dotty , much less uniform in the DKOs and also in Rad21L-/- than in wt cells . Whether the occasional signals located off the axes are true chromatin signals cannot be ascertained , although such signals rarely appear outside the nuclei; still at least some can be unspecific signals derived from sticky chromatin . The presence of SMC3 in the DKOs indicates that there are SMC1α complexes , either with RAD21 and/or with either RAD21L or REC8 . This is consistent with the notion above on the role of SMC1α complexes in cohesion and axis length . SMC1β is present in REC8 and RAD21L deficient cells , indicative of SMC1β in association with either one or two of the other kleisins that are present in these mutants . SMC1α is present along the chromosome axes in all SKOs , as well as in the Smc1β-/-Rec8-/- chromosome spreads . A somewhat weaker signal was observed in the Rad21L-/- spreads . This may indicate the disappearance of an SMC1αRAD21L complex . The Smc1β-/-Rad21L-/- preparations show the weakest SMC1α signal , if any . In most cells no specific SMC1α signal was observed . Since the removal of SMC1β in addition to RAD21L should , in principle , not affect SMC1α that is not associated with RAD21L , the nearly complete absence of SMC1α is not immediately intuitive . We think that the absence of a discernible signal may be due to the extremely short , dot-like axes , which would not allow SMC1α to efficiently associate with them . Thus , the SMC3 and SMC1α patterns are largely consistent with each other , although different antibodies generate different intensities , which therefore can hardly be directly compared . Similarly , RAD21L is present in all spreads except those of Rad21L-/- and Smc1β-/-Rad21L-/- mice . Widespread signals for REC8 are seen in the Rad21L-/- spreads , much less is present in Smc1β-/- and Smc1β-/-Rad21L-/- samples . In agreement with the above notion of a minor contribution of SMC1α/REC8 complex to axis formation , this suggests that the majority of REC8 complexes are SMC1β based . RAD21 is more chromatin-spread and axes-associated in wt , and there is still some RAD21 on Smc1β-/- and Rad21L-/- spermatocyte chromosomes , but less on Rec8-/- chromosomes . Very little is seen on either of the DKO spreads , likely because these cells do not enter pachynema , the stage when RAD21 would reappear . Together this is in accordance with the notion that SMC1α/RAD21 complexes exist on spermatocyte prophase I chromosomes . Consistent with the finding of normal numbers of centromere signals in all mutants , we found that those other two kleisins that are present in a given kleisin mutant localize to the centromeres . The only meiosis-specific SA-type cohesin STAG3/SA3 localizes all along the wt and Smc1β-/- chromosomes indicative of SMC1α/STAG3 complexes as reported before [17] . STAG3 also associates along Rec8-/- chromosomes and is present at least in a dotty pattern on Rad21L-/- chromosomes . However , Smc1β-/-Rec8-/- chromosome spreads show STAG3 signals , which suggest that a complex of SMC1α/STAG3 provides the signals seen on Smc1β-/- and on DKO chromosomes . The STAG3 signals seen on Smc1β-/-Rad21L-/- miniature axes are consistent with a SMC1α/STAG3/REC8 complex , which should not be eliminated in this DKO . In immuno precipitation experiments , STAG3 co-precipitated with either of the SMC1 variants , and precipitated with RAD21 , RAD21L and REC8 although not in all experiments reported [9 , 13 , 17 , 21] . While obviously the absence of one cohesin does not preclude others to associate with chromosomes , one potential caveat of the analysis is that chromosome association or cohesin expression of one cohesin could be increased if another one is missing . However , the strong phenotypes observed in each mutant clearly indicate that no full , probably not even relevant , compensation by other cohesins exist . Changes in expression levels of other cohesins in a particular single cohesin mutant were not observed so far [9 , 13 , 17 , 21] . Earlier we reported telomere deficiencies in Smc1β-/- spermatocytes [26] . These deficiencies included shortened telomeres , SCs without telomeres , telomeres that have apparently been broken off SCs , and telomere fusions . To test the contribution of individual kleisins to telomere integrity , we stained chromosome spreads of wt and all mutants by FISH for telomeric sequences ( telo-FISH ) , by anti-RAP1 for this telomere-specific protein , and by anti-SUN1 for association of telomeres with the SUN/KASH complex ( reviewed in [27] ) , which anchors telomeres at the nuclear membrane in early prophase I ( Fig 7A–7D; S11 Fig ) . Because of the very short or dot-like axes of the two DKOs it is not possible to quantify individual telomeres or telomere-like structures , but a qualitative description can be provided . In wt cells , 19 autosomes in full synapsis generate 38 telomere signals , the X and Y chromosomes yield 3 telomere signals , since the synapsed PAR telomere appears as one , together 41 signals . The loss of PAR synapsis in all mutants increases this number to 42 . A fully unsynapsed autosome shows 4 signals , if there is additional loss of cohesion , 8 signals emerge . Telomere aberrations such as telo-less axes or solitary telomere fragments perturb these numbers . In all mutants , telo-FISH shows aberrant telomeres with telomeric DNA that seems to have ruptured off the axes , with axes that lack telomere signals , and with telomeric ends tightly associated , perhaps fused , or clustered ( Fig 7A ) . These phenotypes are most prominently seen in SMC1β and RAD21L deficient mutants . Rec8-/- spermatocytes show the fewest telomeric aberrations , although often 3- to 4 chromosomes display a low intensity FISH signal at one end . FISH signals are directly proportional to telomere length ( see below and , for example [26] ) . RAP1 staining ( Fig 7B ) confirms that many telomeres are deficient in the DKOs , since many axes lack RAP1 signals . These aberrations were also confirmed by staining for TRF2 ( see below; Fig 8 and S12 Fig ) . Similarly , the SKO and DKO show reduced number of SUN1 spots ( Fig 7C ) , indicative of failure to associate with the nuclear periphery . We counted the number of SUN1 telomere signals in all the mutants of the most advanced stage . We observed in average 45 . 32 ( +/- 3 . 787 , n = 26 ) SUN1 signals in wt zygotene/pachytene spermatocytes . In Rec8-/- , Smc1β-/- and Rad21L-/- spermatocytes we observed 64 . 08 ( +/- 5 . 787 , n = 25 ) , 43 . 3 ( +/- 2 . 787 , n = 26 ) and 42 . 3 ( +/- 6 . 787; n = 26 ) SUN1 signals , respectively . In Smc1β-/-Rec8-/- and Smc1β-/-Rad21L-/- spermatocytes we observed in average only 44 . 3 ( +/- 5 . 987 , n = 23 ) and 39 . 3 ( +/- 10 . 587 , n = 34 ) signals ( S11 Fig ) . Differences were statistically significant with a p-value <0 . 05 for the comparisons of wt versus Rec8-/- , Smc1β-/- versus Rec8-/- , Rec8-/- versus Smc1β-/- Rec8-/- , Rec8-/- versus Rad21L-/- , and Rec8-/- versus Smc1β-/-Rad21L-/- . Thus , the only statistically significant difference was observed with the REC8 deficiency . Detailed interpretation of these numbers is difficult as many processes contribute in different ways . Increased SUN1 numbers may result from unsynapsed chromosomes that each form SUN1 foci . Telomere fragments may also form SUN1 foci as seen mostly in Rad21L-/- cells . Decreased SUN1 foci most likely reflect the loss of telomere ends seen in all mutants , not compensated for by unsynapsed chromosomes . Why , for example , there are fewer SUN1 foci in the Smc1β-/-Rec8-/- cells than in the Rec8-/- spermatocytes , which show comparable levels of asynapsis , can only be speculated about: SMC1β may be much more required to preserve telomeric DNA and its structure than REC8 , consistent with the many aberrations seen in the Smc1β-/- spermatocytes [26] . Overall , this data reflects the expected telomere and telomere attachment deficiencies . Measurements of the intensity of the FISH signal ( Fig 7D ) showed that in all mutants there is a shift towards lower intensity , which indicates shorter telomeres . Telomere intensities peak in the wt at 35000 to 40000 units . The Smc1β-/- spermatocytes display a peak around 20000 units and thus feature shorter telomeres as reported before [26] . In Rec8-/- spermatocytes the median intensity is at app . 7500 units and thus telomeres are even shorter than in Smc1β-/- spermatocytes . Unexpectedly , Smc1β-/-Rec8-/- spermatocytes show more intense telomere signals , indicating that other modes of telomere protection become effective if these two cohesins are absent . The median intensity in Smc1β-/-Rec8-/- spermatocytes is similar to that in Smc1β-/- spermatocytes , and the difference is not statistically significant . Both however show higher intensity and thus longer telomeres than in Rec8-/- spermatocytes . This indicates SMC1β is a main contributor to telomere length and it does so without REC8 , i . e . in a different complex . The effect of REC8 deficiency can thus only be brought about by an SMC1α/REC8 complex . Rad21L-/- and Smc1β-/-Rad21L-/- spermatocytes show very short telomeres , with average intensities peaking at around 10 . 000 , and there is no statistically relevant difference between these two strains . This suggests that RAD21L is mainly associated with SMC1β in this function . The variation in length is particularly extensive in the Rad21L-/- and Smc1β-/-Rad21L-/- spermatocytes . Together this suggests that an SMC1β/RAD21L complex and an SMC1α/REC8 complex are mainly responsible for proper telomere length . There is no additive effect of removing RAD21L in addition to SMC1β . This supports the notion that an SMC1α complex featuring RAD21L does not significantly contribute to telomere length . To reveal ultrastructural features of telomeres we performed super-resolution imaging ( SIM ) on anti TRF2-stained telomeres of wild-type and mutant spermatocytes ( Fig 8 , S12 Fig ) . The analysis confirmed the presence of telomere aberrations on mutant chromosomes as described above ( Fig 8 ) . In addition , we observed loop-like structures on many wild-type telomeres , but rarely on mutant telomeres ( Fig 8 , S12 Fig ) . Plotting a 3D-image from signal intensities to analyze contour plots shows a circle of 4 telomere spots in many wt instances ( S13 Fig ) . Quantification of these loop-like structures showed that 64% of the wt , but less than 10% of the mutant chromosomes carry such structures ( S14 Fig ) . Multiple telomere signals were seen in a third of wt samples , but in half or more of the mutants . The mutants often ( 36 to 48% ) also showed only one telomere signal per chromosome , i . e . one end lacked a signal , which happened only in 4% of wt cases . Stretches of telomere signals were observed only in mutants . We assume that almost all wt telomeres feature these loop-like structures , since depending on the specific plane the telomeres were looked at , one may not be able to see all of them as distinct circles , and some may be lost upon chromosome spreading . The non-paired ends of sex chromosomes of wt often also show circles . This suggests that cohesins , particularly SMC1β complexes , support formation of a more closed conformation at the very end of telomeres , which may represent a protective structure . These loops are reminiscent of TRF2-positive t-loops reported from somatic cells [28] and of telomere complexes reported recently for spermatocytes [29] . In conclusion , the different cohesin complexes that exist in mammalian spermatocytes contribute distinctly to different structures and processes in these cells . S1 Table summarizes the most important observations . Some of our conclusions assume that there is no role of kleisins independently of a cohesin complex , i . e . independently of either SMC1α or SMC1β . Formally this can hardly be excluded , but there is no evidence for this . We think the assumption that kleisins work only within cohesin complexes is very reasonable . So far all known functions of kleisins are consistent with their association with cohesins , and thus the interpretations provided above and below are the most straightforward . Several complexes contribute to axes formation and define their length , but to different extent . As determined in mutant backgrounds , SMC1β complexes determine about half of axes length , SMC1α complexes provide the other half with an SMC1α/RAD21L complex supporting axes length most prominently with roughly one-third , the SMC1α/REC8 complex only contributes a minor fraction . A significant contribution by RAD21 complexes is unlikely . The additive effect of distinct cohesin complexes to axes length suggests that the amount of cohesin available to be loaded onto meiotic chromosomes determines axes length , perhaps more so than the particular type of cohesin , i . e . whether it is an SMC1α or SMC1β cohesin complex . Whether individual complexes prefer to associate with certain sequences or DNA structures along chromosomes is not known but not unlikely given the association of cohesin in mitotic cells with binding sites for transcriptional regulators . Synapsis is supported by all complexes , although to a different extent . It also remains unclear by which mechanism ( s ) –directly or indirectly–synapsis is promoted by cohesins beyond formation of an axis-loop-structure . Sex chromosome pairing at the short PAR is particularly vulnerable to loss of any cohesin , since all mutants fail in X/Y pairing . The previously observed dependence of X/Y pairing on cohesin dosage supports this notion [30] . Centromeric cohesion at the leptonema/early zygonema stage does not depend significantly on the meiosis-specific cohesins and thus relies on cohesion established during premeiotic S phase by SMC1α cohesin . Together with earlier publications it becomes clear that with progression of meiosis , cohesion increasingly depends on meiosis-specific cohesins , since SMC1α vanishes and SMC1β becomes prominent . Telomeres suffer from any absence of meiosis-specific cohesins , but the most from absence of RAD21L or REC8 with SMC1β . The mode of telomere protection , however , remains to be elucidated , but the TRF2 patterns revealed here by SIM hint at loop-like protective structures at spermatocyte telomeres . T-loops were initially described in 1999 [31] , and TRF2-dependent t-loops recently demonstrated for somatic cells were indeed suggested to protect telomeres from non-homologous end-joining and ATM-triggered DNA damage signaling [28] .
Smc1β–/–mice have been previously described [32 , 33] . In Smc1β–/–mice , exon 10 was targeted representing 40% of the hinge domain . Generally , mice were bred and maintained in the animal facility of the Medical Faculty , Technische Universität Dresden ( Dresden , Germany ) according to institutional guidelines . All experiments were performed with approval by the State of Saxony . Rad21L-/- and Rec8-/- mice were generated as described previously [19 , 34] . All mice were in the C57BL/6 genetic background . Number of mice used for the experiments: N = 5 , Smc1β-/-; N = 4 , Rec8-/-; N = 4 , Rad21L-/-; N = 4 , Smc1β-/- Rec8-/-; N = 4 , Smc1β-/- Rad21L-/- . Surface-spread chromosomes were prepared by detergent spreading adapted from Wojtasz et al . [35] . Testis was taken from the sacrificed mice and tunica albuginea was removed . Tubules were digested in 1 ml of 1 μg/ml of collagenase type I—PBS buffer for 10’ at 32°C with slight agitation . Tubules were the centrifuged to pellet the cells and excess collagenase was removed . Pellet was then resuspended in 500μl of 0 . 025% trypsin and incubated for 5’ at 32°C . Then 200 μl of media with FCS was added to the Single cell suspension . Cells were then filtered through 40 μm to remove the cell debris and centrifuged . Pellet was then resuspended in 300 μl of PBS . Now single cell suspension was used for the chromosome spreads . 1 . 5 μl of single cell suspension were dropped on 7 μl of 0 . 25% of NP40 . Cells were allowed to lyse for 2 mins and then fixed by adding 24 μl of S fix ( 1% paraformaldehyde , 10 mM sodium borate buffer pH 9 . 2 ) . Samples were incubated for 1 hour at room temperature in a humid chamber . Slides were dried under a hood and washed two times for one minute with 0 . 4% Agepon ( AgfaPhoto ) and another three times for one minute with water . Slides were used immediately or kept at -20°C until IF staining . Testis were removed from sacrificed mice and placed in 2% ( v/v ) of formaldehyde/PBS for 40’ at RT for fixation before incubation in 30% sucrose/PBS overnight . Subsequently , testes were mounted in O . C . T ( Sakura Finetek Europe ) , shock-frozen on dry ice and stored at -80°C . 8μm thick sections were made from the frozen testis , placed on the slides and dried for at least 30 min at RT . Then slides were treated with ice cold methanol for 10’ and 1’ with ice cold acetone . After completely drying , the slides were kept at –80°C or used immediately for the staining . The tubular stages were defined primarily based on cell associations and DAPI staining ( centromeric and pericentric heterochromatin clustering ) as described in [36] . Chromosome spreads and sections were treated in the same way . Slides were blocked with either blocking buffer ( 2% BSA , 0 . 1% Triton X in PBS ) or 10% goat serum for at-least 1hr at RT before the primary antibody treatment . Slides were incubated with primary antibodies for at-least 3 hrs . at 37°C . Then slides were washed with blocking buffer and incubated with secondary antibodies for at-least 1hr . After the secondary antibody treatment slides were washed with blocking buffer and mounted with Vectashield containing 1μg/ml of DAPI . Statistics was performed using the 1-way Anova test , the Dunn’s test , the Whitney-Mann test or the Wilcoxon test as indicated . Telo-FISH of the G-strand was performed using the Telomere PNA FISH/Cy3 kit ( Dako ) . The hybridization were done for 3 h at RT after denaturation at 80°C for 5 min . Cells from WT , SKO and DKO mice were always hybridized at the same time and compared with each other . Telomere intensity were obtained with equal exposure between all the genotypes and the relative length of telomeres was estimated by measuring the fluorescence intensity using ImageJ . Fluorescence was visualized with Zeiss Axiophot fluorescence microscope and analysis of images was performed using ImageJ version 1 . 43u . Image analysis of SIM images was done using the 3D surface plot plugin in of ImageJ . Grid size and smoothing was kept as 256 and 10 . 0 values , respectively , for all images . The following antibodies ( Tables 1 and 2 ) were used in this study:
|
Unlike somatic cells , which feature two different cohesin complexes , in spermatocytes at least six distinct cohesin complexes form , whose concerted functions are little understood . This study focuses on three meiosis-specific cohesins . Meiosis features specific chromosome structures and dynamics , and we revealed individual contributions of meiotic cohesin complexes to chromosome axes length , centromeric cohesion , telomere integrity and synapsis . The only meiosis-specific SMC protein , SMC1β , was removed leaving only complexes based on the universal SMC1α . In addition to SMC1β , either one of the two meiosis-specific kleisins REC8 or RAD21L , proteins that close the cohesin ring-like structure , were eliminated . “Double-knockout” mutants were compared to the single “knockouts” and wild-type . Telomeres and chromosome synapsis are impaired to different degrees in all mutants . In early prophase I prominent roles of meiosis-specific cohesins are in axis length and synapsis rather than centromeric cohesion . Removal of SMC1β and RAD21L almost completely abolishes all chromosome axes . Centromeric cohesion is initially provided by SMC1α complex ( es ) . Later in meiosis , SMC1β ensures centromeric cohesion , suggesting functional replacement of SMC1α . Thus , different cohesin complexes in spermatocytes contribute distinctly to different structures and processes in these cells , but there is also some functional redundancy .
|
[
"Abstract",
"Introduction",
"Results",
"and",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"meiosis",
"chromosome",
"staining",
"spermatocytes",
"immunofluorescence",
"staining",
"chromosome",
"structure",
"and",
"function",
"centromeres",
"cell",
"cycle",
"and",
"cell",
"division",
"cell",
"processes",
"chromatids",
"germ",
"cells",
"telomeres",
"sperm",
"research",
"and",
"analysis",
"methods",
"specimen",
"preparation",
"and",
"treatment",
"staining",
"synapsis",
"animal",
"cells",
"chromosome",
"biology",
"cell",
"biology",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"chromosomes"
] |
2016
|
Distinct Roles of Meiosis-Specific Cohesin Complexes in Mammalian Spermatogenesis
|
Previous studies have shown that leprosy multi-drug therapy ( MDT ) does not stop the progression of nerve function impairment . There are no prospective studies investigating the evolution of nerve anatomic abnormalities after treatment . We examined leprosy patients aiming to investigate the evolution of nerve ultrasonography ( US ) abnormalities and the risk factors for poor outcomes after MDT . We performed bilateral US of the ulnar ( U ) , median ( M ) and common fibular ( CF ) nerves in 9 paucibacillary ( PB ) and 64 multibacillary ( MB ) patients before and after MDT . Forty-two patients had leprosy reactions ( type 1 , type 2 , acute neuritis ) during the study . We analyzed nerve maximum cross-sectional areas ( CSA ) , echogenicity and Doppler signal . Poor outcomes included a post-treatment CSA above normal limits with a reduction of less than 30% ( U , M ) or 40% ( CF ) from the baseline , echogenicity abnormalities or intraneural Doppler in the post-treatment study . We found that PB and patients without reactions showed significant increases in CSA at CF , whereas MB and patients with reactions had CSA reduction in some nerves after treatment ( p<0 . 05 ) . Despite this reduction , we observed a greater frequency of poor CSA outcomes in the MB compared to the PB ( 77 . 8% and 40 . 6%; p>0 . 05 ) and in the patients with reactions compared to those without ( 66 . 7% and 38 . 7%; p<0 . 05 ) . There was significantly higher odds ratio ( 7 . 75; 95%CI: 1 . 56–38 . 45 ) for poor CSA outcomes only for M nerve in patients with reactions . Poor echogenicity outcomes were more frequent in MB ( 59 . 4% ) compared to PB ( 22 . 2% ) ( p<0 . 05 ) . There was significant association between poor Doppler outcomes and neuritis . Gender , disease duration , and leprosy classification were not significant risk factors for poor outcomes in CSA , echogenicity or Doppler . US nerve abnormalities can worsen after treatment despite the leprosy classification or the presence of reactions .
Leprosy is the leading infectious cause of disability [1 , 2] . Neurological involvement may start before diagnosis or either during or after treatment , leading to functional impairments and deformities [1 , 3–5] . Nerve palpation can detect thickening , but it is examiner-dependent and it demands practical training [6] . One study that evaluated the reliability of nerve palpation detected poor agreement between trained staff [7] . Recently , ultrasonography ( US ) has been used to document anatomical nerve abnormalities in patients with leprosy [8–15] . US provides objective measurements of nerve enlargement and asymmetry [12 , 15] and can identify more extensive involvement than clinical examination [9] . Additionally , leprosy patients can have nerve enlargement detected with US without functional impairment identified in neurophysiological studies and vice versa [10 , 16] . There are prospective studies investigating functional impairment during and after multi-drug treatment ( MDT ) [17–19]; however , to our knowledge , there are no longitudinal studies investigating the evolution of nerve ultrasonographic abnormalities after MDT in leprosy patients . The purpose of this study was to investigate the evolution of US abnormalities in leprosy patients and the risk factors for poor outcomes after treatment . We hypothesize that nerve abnormalities detected by US may not show regression after treatment .
The study was conducted at the Leprosy Reference Center of the Ribeirão Preto Medical School Hospital—University of São Paulo ( HCFMRP-USP ) . The Ethics Committee of the HCFMRP-USP approved the study ( process n°02663112 . 0 . 0000 . 5440 ) . Written informed consent was obtained from all participants . The parents provided written consent on behalf of the minor participants . The present paper shows the results of the prospective US evaluation after treatment . The patient flowchart reporting numbers of individuals at each stage of the study is shown in Fig 1 . One hundred patients underwent bilateral high-resolution US of the peripheral nerves before starting the World Health Organization ( WHO ) MDT . The pre-treatment results have been published [15] . Seventy-three leprosy patients repeated US after completion of MDT ( 31 women and 42 men , age range 8–86 years , age mean 45 . 3±17 . 3 ) . To reduce bias , patients underwent post-treatment US approximately 2 years after the pre-treatment exam , regardless of their classification . The group of paucibacillary patients ( PB ) underwent post-treatment US an average of 27 . 6 months after the pre-treatment exam , and the multibacillary patients ( MB ) underwent post-treatment US an average of 21 . 8 months afterwards ( p>0 . 05 ) . Leprosy diagnosis was established based on clinical signs and symptoms , skin smears , skin biopsy , and neurophysiological examination when necessary . Patients were classified into the following five groups according to the Ridley-Jopling classification [20]: tuberculoid ( TT ) , borderline-tuberculoid ( BT ) , borderline-borderline ( BB ) , borderline-lepromatous ( BL ) , and lepromatous ( LL ) . Patients with the indeterminate form ( I ) of leprosy were also included . Cases of I and TT leprosy were classified as PB , whereas the other forms were classified as MB according to the WHO operational classification [21] . The patients’ medical charts were reviewed for the identification of leprosy reactions . Type 1 cutaneous reactions were defined as the presence of erythema and edema of skin lesions associated or not with new lesions . There may be accompanying neuritis and edema of the hands , feet , or face . Type 2 cutaneous reactions ( erythema nodosum leprosum ) were defined as the presence of tender subcutaneous skin lesions . There may be accompanying neuritis , iritis , arthritis , orchitis , dactylitis , lymphadenopathy , edema , and fever . Neuritis was diagnosed if the patient presented with acute swelling and/or functional impairment of peripheral nerves with spontaneous pain or tenderness on palpation . Anti-reaction treatment was started as soon as the reaction was detected . Type 1 reactions and neuritis were treated with corticosteroids ( initial dose 0 . 5 to 1 . 0mg/kg/day ) for at least 16 weeks . Type 2 reactions were treated with thalidomide ( 100 to 300mg/day ) and/or corticosteroids . Two patients with severe and recalcitrant neuritis received azathioprine associated with corticosteroids . For the statistical analysis of quantitative variables ( CSA , ΔCSA and ΔUtpt ) patients were classified according to the presence of any type of leprosy reaction or absence of reactions during the study . We analyzed the results dividing the patients in these two groups because the presence of reactions , independently of its type , could lead to additional nerve inflammation . Besides , as patients can have concomitant reactions ( for example , neuritis and type 1 cutaneous reaction ) , the separated analysis of reactions could lead to bias . Patients were also classified according to the disease duration , which was defined as the interval between the first symptoms and the realization of the pre-treatment US . The disease duration was considered short ( less than 2 years ) , moderate ( between 2 and 5 years ) or long ( more than 5 years ) . The control group included 41 healthy volunteers ( 27 women and 14 men , age range 12–80 years , mean age 37±17 . 4 years ) who were submitted to the same US exam protocol as the leprosy patients . The results of the control group were used to calculate the upper limits for the nerve cross-sectional areas ( CSAs ) ( mean plus 2 standard deviations ) . Leprosy patients with diabetes mellitus , hypothyroidism , human immunodeficiency virus infection , trauma-related peripheral nerve disease , hereditary neuropathies , autoimmune diseases or alcoholism were not included in the study . The control group comprised healthy volunteers without household contact with leprosy patients and without other potential causes of peripheral neuropathies ( diabetes mellitus , hypothyroidism , human immunodeficiency virus infection , trauma-related peripheral nerve disease , hereditary neuropathies , autoimmune diseases , alcoholism ) . The US exam was performed as previously described [15] . Musculoskeletal radiologists with previous fellowship training in nerve imaging performed all US sessions using a 12-MHz linear transducer model HDI-11 ( Philips Medical Systems , Bothell , Washington , USA ) . The ulnar ( at the cubital tunnel area—Ut—and proximal to the tunnel—Upt ) , median ( M ) and common fibular ( CF ) nerves were systematically scanned along the transverse and longitudinal axes . Ulnar nerves were scanned from the middle third of the arm to the middle third of the forearm . M nerves were evaluated at the middle and distal thirds of the forearm . CF nerves were evaluated from the distal third of the thigh to the knee at the fibular head . In some cases , it was not possible to examine nerves bilaterally due to amputation , cutaneous ulcers or other cutaneous alterations at the site of examination . Nerve CSAs were measured by freehand delimitation at the inner borders of the echogenic rims of the nerves at the level of maximum thickening . The CSA measurements were used to calculate asymmetry . These following measurements were determined: ( 1 ) CSA index ( ΔCSA ) , absolute difference between CSAs for each nerve point from one side to the contralateral side; ( 2 ) Ut-Upt index ( ΔUtpt ) of the ulnar nerve , absolute difference between the largest and smallest CSAs of the Upt and Ut points of the ulnar nerves on the same side . The color Doppler settings were chosen to optimize the identification of weak signals from vessels with a slow velocity and to avoid artifacts . To increase the vascular depiction , the power Doppler mode was used with a PRF of 0 . 7 to 1 kHz . The detection of intraneural or epineural Doppler signal was considered indicative of nerve hypervascularity and therefore an abnormal finding . The nerve echogenicity was also classified as normal or abnormal . Nerves were classified as abnormal if they showed hypoechoic or hyperechoic areas or focal thickening with loss of the normal fascicular pattern . A pilot study with 15 leprosy patients and 5 healthy volunteers was performed to evaluate the inter-observer reliability of US . Two radiologists , who were blinded to the patient diagnosis and to the measurements of the other radiologist , performed consecutive CSA measurements . The 90th percentile of the inter-observer variation ranged between 33 . 3–35 . 6% at Upt , 31 . 4–37 . 9% at Ut , 25 . 6–30 . 9% at M , and 35 . 4–45 . 2% at the CF nerve . The intraclass correlation coefficient ( ICC ) was above 0 . 77 for all nerves examined , which is considered a strong inter-observer agreement . ICC was classified as follows: poor ( 0–0 . 2 ) , fair ( 0 . 3–0 . 4 ) , moderate ( 0 . 5–0 . 6 ) , strong ( 0 . 7–0 . 8 ) , and almost perfect ( >0 . 8 ) [22] . Considering the 90th percentile of the inter-observer variation , we defined a significant change in the nerve CSA between pre- and post-treatment exams as a >30% difference from baseline for the Upt , Ut and M nerves and >40% difference for the CF nerve . Poor CSA outcomes were defined as a post-treatment CSA above the normal limits ( mean plus 2 standard deviations of the control group measurements ) and with less than a 30% ( Upt , Ut , M ) or 40% ( CF ) reduction from the baseline . Poor echogenicity outcome was defined as the presence of echogenicity abnormalities in at least one nerve in the post-treatment study . Poor Doppler outcome was defined as the detection of intraneural or perineural Doppler signal in at least one nerve in the post-treatment study . Statistical analysis was performed using SAS software version 9 . 0 ( SAS Institute Inc . , Cary , NC ) . We performed linear regression controlling for the effects of confounding factors ( operational classification , reactions , disease duration , and gender ) . The logistic regression was used to estimate the odds ratio ( OR ) and we considered the following risk factors for poor outcomes: operational classification , reactions , disease duration , and gender . The statistical analysis also included Chi-square and McNemar tests . Probability ( p ) values less than 0 . 05 were considered significant .
The clinical classifications and incidences of reactions are presented in Table 1 . Some patients presented with cutaneous reactions ( types 1 or 2 ) associated with neuritis . For clinical data and pre- and post-treatment US findings of each patient see supporting information ( S1 Table ) . The prospective analyses of CSA measurements for the PB and MB patients are shown in Table 2 . No significant differences between pre and post-treatment asymmetry measurements ( ΔCSA and ΔUtpt ) were observed in the PB and MB . The majority ( 77 . 8% ) of MB patients had at least one nerve with a poor CSA outcome compared to 40 . 6% of the PB ( p>0 . 05 ) . The analysis of each nerve revealed that none of the PB had poor CSA outcomes for the Upt , Ut , and M nerves . The frequency of poor CSA outcomes in the MB for the right and left side nerves were 22 . 6% and 27 . 9% for the Upt , 25% and 20 . 6% for the Ut , and 25% and 27 . 4% for the M nerve , respectively . For the CF nerve , we observed similar frequencies of poor CSA outcomes in the PB ( 11 . 1% right and 22 . 2% left ) and MB patients ( 20 . 3% right and 17 . 4% left ) . The OR revealed similar risks for poor CSA outcomes between PB and MB at the CF nerve . For the other nerves it was not possible to calculate the OR because none PB patient had poor CSA outcomes . The pre- and post-treatment CSA , ΔCSA and ΔUtpt measurements for the groups with and without reactions are shown in Tables 3 and 4 . We observed higher frequencies of poor CSA outcomes in at least one nerve in patients with reactions ( 66 . 7% ) compared to patients without reactions ( 38 . 7% ) ( p<0 . 05 ) . The results for each nerve are shown in Table 5 . Considering the entire group of patients ( n = 73 ) , we observed a higher frequency of echogenicity abnormalities in the post-treatment exam ( 54 . 8% ) compared to the pre-treatment ( 42 . 5% ) ( p<0 . 05 ) . Only two patients who had echogenicity abnormalities before treatment showed improvement after treatment and 11 patients who had no echogenicity abnormalities before treatment developed abnormalities in the post-treatment US . We observed a lower frequency of Doppler detection after treatment ( 19 . 2% pre-treatment and 8 . 3% post-treatment , p>0 . 05 ) . Among the 14 patients who had Doppler signal detection in at least one nerve before treatment , only one maintained Doppler detection and 13 showed improvement . Considering the presence of each type of reaction , we observed that none of the patients with type 2 cutaneous reactions presented intraneural Doppler detection both on pre-and post-treatment exams . Six patients had Doppler signal detection after treatment: 2 patients had type 1 reaction associated with neuritis , 3 patients had neuritis without cutaneous reactions , and 1 patient had no clinical signs leprosy reactions during the post-treatment US and evolved with neuritis a couple of months later . We observed higher frequencies of poor echogenicity outcomes in MB ( 59 . 4% ) compared to PB ( 22 . 2% ) ( p<0 . 05 ) , without a significant increment in the OR ( OR: 4 . 42; CI95%: 0 . 81–24 . 24; p>0 . 05 ) . 9 . 4% of the MB had poor Doppler outcome and none of the PB patients showed Doppler detection after treatment ( p>0 . 05 ) , precluding the estimation of the OR . Patients with reactions presented higher frequencies of poor echogenicity outcomes ( 61 . 9% ) compared to the patients without reactions ( 45 . 2% ) ( p>0 . 05 ) , without increment in the OR ( OR: 1 . 97; CI95%: 0 . 74–5 . 27; p>0 . 05 ) . There was no significant association between poor echogenicity outcomes and the presence of type 1 reactions , type 2 reactions or neuritis . The frequency of poor Doppler outcomes was also non-significantly higher in the patients with reactions ( 11 . 9% ) compared with those without reactions ( 3 . 2% ) . It was not possible to calculate the OR for this variable due to quasicomplete separation [23] . Considering each type of reaction separately , we observed significant association between Doppler detection and the presence of neuritis ( p = 0 . 02 ) . There was no significant association between Doppler signal and types 1 and 2 cutaneous reactions . Gender and disease duration were not significant risk factors for poor outcomes for CSA , echogenicity or Doppler .
In this first prospective study investigating ultrasonographic nerve abnormalities in leprosy we found that the neural involvement may not improve after treatment , in accordance with the findings of previous clinical and electrophysiological studies [17–19 , 24–27] . Furthermore , US findings can worsen despite the operational classification and the presence of reactions , as demonstrated by the CSA increase in the CF nerve in PB and in patients without reactions and by the increased frequency of echogenicity abnormalities detected in the post-treatment evaluation . We expected to detect higher odds for poor outcomes in the MB and in the patients with reactions , but we observed increments in the OR only for the right median nerve CSA in patients with reactions; the other analyses of the OR ( CSA , echogenicity , and Doppler ) revealed similar odds between PB and MB and between patients with and without reactions . These results are very important because there are few studies that have investigated imaging findings in leprosy neuropathy , and all of them were transversal studies [8–10 , 12 , 15 , 16 , 28] . In our study , we did not investigate nerve function; we only aimed to describe anatomical nerve changes detected by US . Nevertheless , our findings can be compared to those of previous electrophysiology studies . We found that PB patients had a significant increase in the CF nerve CSA , with a pronounced difference between the pre- and post-treatment mean and median values . This nerve also presented an expressive percentage of poor CSA outcomes in the PB , with similar OR between PB and MB . The only published study that has investigated the evolution of electrophysiological findings before and after treatment in PB and MB ( 15 and 17 patients , respectively ) [19] found that 3 PB and 2 MB patients had clinical and/or electrophysiological signs of deterioration and the lower extremity nerves were more frequently and severely affected than the upper extremity nerves in both groups . These results indicate that PB patients can have deterioration of imaging and functional findings after treatment , suggesting that the neural inflammatory process may continue after healing the M . leprae infection . The CF nerve was also the most frequently enlarged nerve in PB patients before treatment [15] , emphasizing the importance of the investigation of the involvement of lower extremity nerves ( anatomical and/or functional ) . The group of patients who did not have reactions during the study also presented significant increases in the CF nerve CSA after treatment . One previous prospective study [17] that investigated the electrophysiological parameters of 365 MB patients who were divided in two groups ( with and without reactions ) revealed that deterioration of nerve function was more frequent than improvement in both groups . Similar to the findings of Capadia et al . , our results revealed high frequencies of poor outcomes for CSA ( up to 25% ) and echogenicity ( 45 . 2% ) in the patients without reactions , sometimes with greater percentages of abnormalities than patients with reactions . These results show that leprosy neuropathy can deteriorate even in patients that received the recommended doses of the WHO-MDT and in patients without reactions . The MB and the patients with reactions showed the opposite tendency; they presented reductions in CSA , ΔCSA and ΔUtpt for some nerves . Although we observed statistically significant reductions in the means of these measurements , the clinical significance is uncertain , as the magnitude of the reduction was small . We observed a higher frequency of poor CSA outcomes in MB compared to PB ( p>0 . 05 ) and in patients with reactions compared to those without reactions ( p<0 . 05 ) , suggesting no significant improvement . Some of the patients of the group with reactions were still receiving anti-reaction treatment at the time of post-treatment US; therefore , it is possible that the presence of poor CSA outcomes were partially due to the persistence of the inflammatory process in this group . In addition , the reduction in nerve diameter is not necessarily a signal of improvement; it can represent the evolution to nerve fibrosis and atrophy . In the US exams , the fascicular changes associated with nerve inflammation , fibrosis and atrophy are represented by the echogenicity abnormalities [8–10 , 13 , 29] . We observed higher frequencies of echogenicity abnormalities in MB ( p<0 . 05 ) and in patients with reactions ( p>0 . 05 ) , indicating that the reductions in nerve measurements might have been due to these processes . The analyses of risk factors for poor US outcomes revealed that the odds for deterioration of anatomical changes detected by US seem to be only slightly higher in patients with reactions , whereas MB and PB patients had similar odds . Although previous studies have demonstrated that MB patients have a higher risk of developing or worsening nerve function impairment [24 , 25] , the differences between our study design and previous studies preclude definite conclusions . Intraneural or perineural Doppler detection is considered a marker of active neuritis [8 , 9 , 13 , 16 , 29] and our results confirm the association between Doppler signal and neuritis . We did not observe higher odds for having Doppler signal in the group with reactions and we detected a smaller frequency of Doppler signal after treatment ( p>0 . 05 ) . As all the patients with reactions received MDT and anti-reaction treatment , our results indicate that adequate treatment can diminish the acute inflammatory process caused by the bacillus and/or by immune reaction [8] . The detection of intraneural Doppler in one patient without clinical signs of reaction that developed active neuritis afterwards indicates that Doppler may detect subclinical reactions allowing for prompt treatment . Intraneural Doppler can be the first sign of nerve damage and it may have a role predicting reactions [8 , 9] . It has been demonstrated that patients with long standing disease can have important nerve echogenicity abnormalities without significant enlargement [8] . In our study we did not observe significant differences among the three groups for disease duration regarding nerve enlargement; however , we also did not observe differences in the frequencies of echogenicity abnormalities . We defined as long disease duration the presence of symptoms for more than 5 years . Martinoli et al . studied patients with disease durations ranging from 20 to 51 years; therefore , it is possible that the time frame defined as “long disease duration” in our study was not long enough to reveal differences between groups and future US exams ( 5 to 10 years after completion of WHO-MDT ) should be done in these patients to investigate late nerve changes . One major limitation of our study was the absence of correlation between nerve anatomical changes and nerve function abnormalities . Although neurophysiology studies could provide important information , we consider that our results can improve the understanding of the evolution of anatomic nerve changes detected by US . Similarly , the correlation between US findings and clinical symptoms and incapacity grade should be explored in future studies . Another limitation of the study was the relatively small sample size given that the standard deviation was great for some measurements . As the study was performed in a Leprosy Reference Center , we included a small number of PB patients , which can weaken the generalization of conclusions for this group . However , this limitation is observed in the majority of studies investigating leprosy neuropathy . The most feared consequences of leprosy are due to nerve damage . Previous studies have shown that MDT does not stop the progression of nerve function impairment [5 , 17 , 26 , 27] . The present study shows that , similar to the results concerning nerve function , the anatomic nerve changes caused by leprosy may not improve significantly with treatment . Furthermore , they can worsen even in the PB and in the patients without reactions . US is an accurate method for detecting nerve enlargement , as was demonstrated by the high ICC values in our study and in previous studies [10 , 13 , 30] . In addition , it provides useful information about active inflammatory process ( Doppler ) and fascicular abnormalities ( echogenicity ) [8 , 9 , 11 , 16 , 13 , 14 , 29] . As the stigma related to leprosy is due to the consequences of neuropathy , it is essential the improvement of diagnostic and therapeutic procedures focusing on peripheral nerve involvement .
|
Leprosy can lead to functional and anatomical changes in the peripheral nerves . Previous studies have shown that the anti-bacterial treatment cannot stop the progression of nerve function abnormalities; to our knowledge , there are no previous prospective studies investigating the evolution of nerve anatomical abnormalities . Seventy-three leprosy patients underwent bilateral ultrasonography of the ulnar , median and common fibular nerves before and after treatment . We analyzed thickening and asymmetry measurements , nerve architecture ( echogenicity abnormalities ) and vascularization ( Doppler signal ) . We observed worsening of thickening in the common fibular nerve in the paucibacillary and in the patients without reactions . On the other hand , we observed reduction in the thickening and asymmetry measurements for some nerves in the multibacillary and in the patients with reactions . Although we observed reduction in the measurements of these two groups , they presented higher frequencies of echogenicity abnormalities , indicating that these reductions might have been due to evolution to nerve fibrosis and atrophy . The frequencies of Doppler signal were similar between the pauci and multibacillary patients . We observed association between Doppler signal and the presence of acute neuritis . Our results indicate that the nerve anatomic abnormalities assessed by ultrasonography can worsen after treatment despite the patient classification and the presence of reactions .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"medicine",
"and",
"health",
"sciences",
"diagnostic",
"radiology",
"ultrasound",
"imaging",
"pathology",
"and",
"laboratory",
"medicine",
"nervous",
"system",
"immunology",
"tropical",
"diseases",
"neuroscience",
"diabetes",
"mellitus",
"bacterial",
"diseases",
"research",
"design",
"endocrine",
"disorders",
"signs",
"and",
"symptoms",
"neglected",
"tropical",
"diseases",
"research",
"and",
"analysis",
"methods",
"neuropathy",
"infectious",
"diseases",
"inflammation",
"imaging",
"techniques",
"endocrinology",
"metabolic",
"disorders",
"immune",
"response",
"radiology",
"and",
"imaging",
"diagnostic",
"medicine",
"neuroanatomy",
"anatomy",
"nerves",
"neurology",
"leprosy",
"biology",
"and",
"life",
"sciences",
"prospective",
"studies"
] |
2016
|
Ultrasonography of Leprosy Neuropathy: A Longitudinal Prospective Study
|
The exacting nutritional requirements and complicated life cycles of parasites mean that they are not always amenable to high-throughput drug screening using automated procedures . Therefore , we have engineered the yeast Saccharomyces cerevisiae to act as a surrogate for expressing anti-parasitic targets from a range of biomedically important pathogens , to facilitate the rapid identification of new therapeutic agents . Using pyrimethamine/dihydrofolate reductase ( DHFR ) as a model parasite drug/drug target system , we explore the potential of engineered yeast strains ( expressing DHFR enzymes from Plasmodium falciparum , P . vivax , Homo sapiens , Schistosoma mansoni , Leishmania major , Trypanosoma brucei and T . cruzi ) to exhibit appropriate differential sensitivity to pyrimethamine . Here , we demonstrate that yeast strains ( lacking the major drug efflux pump , Pdr5p ) expressing yeast ( ScDFR1 ) , human ( HsDHFR ) , Schistosoma ( SmDHFR ) , and Trypanosoma ( TbDHFR and TcDHFR ) DHFRs are insensitive to pyrimethamine treatment , whereas yeast strains producing Plasmodium ( PfDHFR and PvDHFR ) DHFRs are hypersensitive . Reassuringly , yeast strains expressing field-verified , drug-resistant mutants of P . falciparum DHFR ( Pfdhfr51I , 59R , 108N ) are completely insensitive to pyrimethamine , further validating our approach to drug screening . We further show the versatility of the approach by replacing yeast essential genes with other potential drug targets , namely phosphoglycerate kinases ( PGKs ) and N-myristoyl transferases ( NMTs ) . We have generated a number of yeast strains that can be successfully harnessed for the rapid and selective identification of urgently needed anti-parasitic agents .
Parasitic diseases such as malaria , schistosomiasis , leishmaniasis , sleeping sickness , and Chagas disease affect millions of people every year , leading to severe morbidity and death . For example malaria , caused by parasites of the genus Plasmodium , kills 1–3 million people every year ( http://www . who . int/mediacentre/factsheets/fs094/en/index . html ) . The disease is primarily treated by chloroquine , artemisinin , and antifolates ( e . g . pyrimethamine ) . However , Plasmodium spp . have become resistant to all of these drugs [1] . Schistosomiasis , caused by the blood fluke Schistosoma , is the second most important parasitic disease worldwide , affecting ca . 207 million people ( http://www . who . int/mediacentre/factsheets/fs115/en/index . html ) [2] , and causing the death of greater than 300 , 000 individuals per annum [3] . Schistosomiasis is commonly treated through administration of praziquantel , but the sole use of this drug against all human infective schistosome species raises the concern that drug-resistant parasites may develop [4] . Sleeping sickness ( or African trypanosomiasis ) , caused by Trypanosoma brucei , infects about 300 , 000 people each year leading to about 40 , 000 deaths [5] , [6] . Chagas disease , caused by T . cruzi , is endemic to Latin America , where ca . 17 million people are infected , leading to about 21 , 000 deaths reported each year [7] . Very few drugs are available for the treatment of patients infected with Trypanosoma spp . and none is satisfactory due to low efficacy , high cost , or unacceptable side effects [8] , [9] , [10] . Leishmaniasis , caused by Leishmania spp . is endemic in 88 countries affecting 12 million people ( http://www . who . int/leishmaniasis/en/ ) . It is traditionally treated with antimony compounds , but resistance to this class of drugs is increasing and very few novel drugs are under development . All of this demonstrates that the design or discovery of novel antiparasitic drugs is a global imperative . For the past 30 years , the yeast Saccharomyces cerevisiae has been used successfully as a vehicle for the expression of heterologous proteins with the aim of understanding their function , producing high levels of recombinant protein , or studying the effect of drugs on defined targets ( for review see [11] ) . By performing genome-wide drug sensitivity screens ( chemogenomic profiling ) [12] , [13] , [14] , [15] , [16] , [17] of yeast mutants with the antimalarials quinine [18] , St . John's Wort [19] and artemisinin [20] , researchers were able to identify their primary targets as well as identify potential side effects . Furthermore , several groups have been able to complement yeast loss-of function mutations by expressing coding sequences from parasites such as Plasmodium [21] , [22] , Schistosoma [23] , [24] , [25] , Leishmania [26] or Trypanosoma [6] , [27] , [28] , [29] , [30] , [31] . Yeast cells expressing parasite proteins potentially provide a well-characterised platform for functional studies of heterologous proteins as well as for screens attempting to identify novel drugs including antiparasitics [32] , [33] , [34] , [35] . For instance , Geary and co-workers expressed potential drug targets for Haemonchus contortus ( wireworm , an important parasite of ruminants ) in Escherichia coli [36] , [37] and later in Saccharomyces cerevisiae [38] , and pioneered the development of high-throughput drug screens for antiparasitics in yeast [38] . A well-characterised anti-parasitic drug target is dihydrofolate reductase ( DHFR ) . DHFR is the enzyme responsible for converting dihydrofolate into tetrahydrofolate , an intermediate in the synthesis of purines , thymidylic acid , and certain amino acids ( e . g . methionine ) . DHFR is present in organisms ranging from bacteria to humans and is the target of pyrimethamine treatment of malaria and human tumours , since rapidly growing cells require folate to produce thymine [39] . Sibley and co-workers [22] , [40] , [41] have done extensive work on the complementation of yeast dfr1 mutations by overexpression of human and Plasmodium DHFRs and demonstrated the suitability of the strains for drug screens in plate assays . Phosphoglycerate kinase ( PGK ) is a central enzyme in glycolysis and gluconeogesis; it catalyzes the transfer of high-energy phosphoryl groups from the acyl phosphate of 1 , 3-bisphosphoglycerate to ADP to produce ATP . PGKs are essential for the blood stages of parasites but the human enzyme is not expressed in erythrocytes; therefore , the enzyme has been proposed as a drug target [42] , [43] . N-myristoyltransferase ( NMT ) is an enzyme responsible for the co- and post-translational modification of proteins by transferring myristate groups to N-terminal glycine residues , allowing their targeting to various membranes [5] , [44] . NMTs are essential enzymes conserved from kinetoplastid parasites to humans and have been successfully demonstrated as drug targets [5] . In spite of the pressing need for new treatments targeting neglected diseases , pharmaceutical companies have had little interest in the research and development of new medicines towards diseases affecting overwhelmingly or exclusively developing countries [45] , [46] . Thankfully , with the investment of funds from organizations such as the Bill and Melinda Gates Foundation , Medicines for Malaria Venture ( MMV ) , the Drugs for Neglected Diseases initiative ( DNDi ) , and the Institute for One World Health ( IOWH ) this scenario is changing [47] , [48] . Now , pharmaceutical giants such as Novartis [49] , GSK [50] , Pfizer ( just to name a few ) investing great efforts in the development of novel antimalarials ( www . mmv . org/research-development/science-portfolio ) . However , due to the diversity of parasite species and their complex life cycles , the development of inexpensive and rapid drug screening methods is a constant challenge . Various groups have developed efficient high-throughput drug screening methods based on intact parasites [51] , [52] , [53] , [54] . However , these are specific to one or a few parasite species and do not always provide information concerning the target of the hit compound . Conversely , the standard alternative of using pure proteins as targets can be unsatisfactory because the assay neglects all other biological interactions of the candidate compounds [35] . To meet this challenge , we have developed a series of yeast strains that can be used to screen for drugs against multiple drug targets from multiple parasites with a single experimental set-up .
We constructed plasmid maps and Genbank files of the constructs expressing heterologous protein using the program CLC Genomics Workbench . We translated the coding regions of each of the proteins and performed protein alignments of them against Saccharomyces cerevisiae Dfr1p , Nmt1p and Pgk1p . We constructed the similarity tree using the standard settings from CLC Genomics Workbench ( neighbor-joining , bootstrap analysis , 100 replicates ) . The DHFR/PGK/NMT-coding regions of Schistosoma mansoni ( Sm ) , Saccharomyces cerevisiae ( Sc ) , Leishmania major ( Lm ) , Trypanosoma cruzi ( Tc ) and T . brucei ( Tb ) were PCR amplified from genomic DNA templates and cloned into pCM188 [55] . The following plasmids were constructed: pCMSmDHFR ( pCM188 with the complete open reading frame of Schistosoma mansoni DHFR under the control of the tetracycline regulatable promoter TetO2 ) , pCMScDHFR , pCMLmDHFR , pCMTcDHFR , pCMTbDHFR , pCMSmPGK , pCMLmPGKB , pCMTcPGK , pCMTbPGK , pCMSmNMT , pCMLmNMT , pCMTcNMT , pCMTbNMT ( Table S1 ) . Human ( Hs ) PGK and NMT2 were PCR amplified from a human cerebellum cDNA library and cloned into pCM188 to produce pCMHsPGK and pCMHsNMT2 ( Table S1 ) . The DHFR coding sequences from human , Plasmodium falciparum ( Pf ) , and P . vivax ( Pv ) , as well as PGK and NMT from P . vivax , were synthesised by GENEART with a codon usage suitable for expression in yeast . These synthetic constructs were sub-cloned into pCM188 to generate pCMHsDHFR , pCMPfDHFR , pCMPvDHFR , pCMPvPGK and pCMPvNMT ( Table S1 ) . The DHFR mutations N51I , C59R and S108N confer resistance to antifolates in wild Plasmodium falciparum populations; we designate such resistant alleles by the preceding superscript Pfr and give the amino-acid changes in parentheses in a succeeding superscript . Two rounds of site-directed mutagenesis were performed to introduce these mutations into pCMPfDHFR to generate pCMPfRdhfr ( 51I , 59R , 108N ) ( Table S1 ) . All constructs were verified by sequencing ( Figures S1 , S2 , S3 , S4 , S5 , S6 , S7 , S8 and Text S1 , S2 , S3 , S4 , S5 , S6 , S7 , S8 , S9 , S10 , S11 , S12 , S13 , S14 , S15 , S16 , S17 , S18 , S19 , S20 , S21 ) . Deletions of the yeast PDR5 coding sequence ( specifying the major drug efflux pump ) from dfr1 , pgk1 and nmt1 heterozygous mutant strains were performed as previously described [56] . pCM-DHFR constructs were transformed into dfr1Δ::KanMX/DFR1 pdr5Δ::HisMX/PDR5 strains ( BY4743 background [57] ) . pCM-PGK constructs were transformed into pgk1Δ::KanMX/PGK1 pdr5Δ::HisMX/PDR5 strains ( BY4743 background ) . pCM-NMT constructs were transformed into nmt1Δ::KanMX/NMT1 pdr5Δ::HisMX/PDR5 strains ( BY4743 background ) . Heterozygous diploid strains were then sporulated and dissected using a micromanipulator ( Singer MSM ) . Derived haploids with the genotype dfr1Δ::KanMx MET15 lys2Δ0 MATα , pgk1Δ::KanMx MET15 lys2Δ0 MATα or nmt1Δ::KanMx met15 LYS2 MATa were selected for drug screens ( Table S2 ) . Standard growth conditions and either YPD ( 2% peptone , 1% yeast extract , 2% glucose ) or YNB-glucose ( 0 . 68% yeast nitrogen base , 2% ammonium sulphate , 2% glucose ) with the relevant supplements were used for all assays . Cultures of wild type ( BY4741 ) and transformant yeast strains were inoculated into 200 µl of YPD in the wells of 96-well microtiter plates and grown for 40 hours at 30°C . Growth was monitored with the BMG Optima multiplate reader . Each sample was present in quadruplicates , distributed randomly throughout the plate and OD595 measurements were made ever 10 minutes . A 384-well plate ( with 70 µl of YPD + doxycycline ) was prepared in a similar manner . The OD readings from the plate reader were log-transformed and 7 consecutive readings were used to calculate exponential growth rate of the culture in each well . These exponential growth rates were then normalized by dividing by the average exponential growth rate of the wild type culture grown on the same plate . Normalized growth rates were averaged across the 4 or 5 replicates and standard deviations calculated . Serial dilutions ( 5x ) of stationary phase cultures were prepared in 96-well plates and replicated onto agar plates manually or in quadruplicate onto agar plates using a robot ( Singer RoToR ) . Cells were allowed to grow for 2 days at 30°C . Strains were spotted in four replicates ( Figure S10 ) . Each pair of rows on the image corresponds to a strain , spotted in increasing dilutions ( 1∶1 , 1∶5 , 1∶25 … ) in blocks of four . The images of the plates were produced in Gel Doc 2000 ( BioRad ) and saved as . jpg files . Matlab was used to convert the images to three-dimensional matrices , the first and second dimensions were the vertical and horizontal dimensions of the image and third dimension was the colour channel in RBG 24-bit format . As the images were saved in grey scale , choice of channel did not make a major impact on quantification of colonies; third channel ( Green ) was used . The average intensities of pixels on each column and row in the image were calculated , intersection points of rows and columns with highest average intensity among their neighbours ( in a window of 5×5 ) were set as the spot or colony centres . Once the colony centres were set , a 16 pixels×16 pixels diamond shape colony frame was set around each colony centre , and average intensity was calculated for each colony frame . Then the pixels with 50% larger intensity than the average of the colony frame were counted and recorded as the colony size .
We have defined our candidate drug targets based on the following criteria: ( i ) the target should be an enzyme that is essential in yeast ( this permits verification of the functional expression of the target ) ; ( ii ) the target should be essential , or predicted to be essential , in most parasites; ( iii ) it may be present in human ( a ‘humanized’ yeast strain will be used as a control in the screens ) ; ( iv ) there should be a low similarity between human and parasite proteins; ( v ) the target should be one suggested by the TDR targets database ( http://tdrtargets . org/ ) . Combining these criteria , we selected 4 drug targets for expression in yeast: dihydrofolate reductase ( DHFR , EC:1 . 5 . 1 . 3 ) , phosphoglycerate kinase ( PGK , EC:2 . 7 . 2 . 3 ) , N-myristoyl transferase ( NMT , EC:2 . 3 . 1 . 97 ) , and farnesyl pyrophosphate synthetase ( FPS , EC:2 . 5 . 1 . 1 , EC:2 . 5 . 1 . 10 ) . As explained in the Introduction , all of these enzymes have previously been proposed as useful targets for antiparasitic drugs , and many have been used in drug screens [5] , [22] , [41] , [42] , [43] , [44] , [58] . We have constructed a series of four plasmids containing the coding sequences ( cds ) for human DHFR , PGK , NMT , and FPS and transformed these into diploid yeast strains that are heterozygous deletion mutants for the gene encoding the corresponding essential enzyme , namely: DFR1 ( ScDHFR ) , PGK1 ( ScPGK ) , NMT1 ( ScNMT ) or ERG20 ( ScFPS ) . The transformed diploids were then sporulated . Two of the four haploid spores in the tetrad will carry the deletion for the essential yeast gene and will only grow if that mutation is complemented by the orthologous human cds . We observed that the overexpression of the cds for human DHFR ( a synthetic gene codon-optimised for expression in yeast was used ) , PGK or NMT2 could complement the essential function of the yeast deletions ( Figure 1 ) , whereas human FPS could not complement the deletion of the yeast FPS ( data not shown ) . We then constructed plasmids expressing codon-optimized cds for Plasmodium vivax DHFR , NMT and PGK , P . falciparum DHFR and drug resistant P . falciparum dhfr , and verified that these heterologous sequences could also complement the yeast deletions . We constructed strains expressing cds for Schistosoma mansoni , Leishmania major , Trypanosoma cruzi and T . brucei DHFRs and NMTs , and found that each of them could complement yeast deletions , albeit with varying efficiencies ( Figure 1 ) . The same was true for each of the Schistosoma mansoni , Trypanosoma cruzi , and T . brucei PGKs tested . We tested whether cds for the different Leishmania major PGKs – the cytosolic PGKB , the glycosomal PGKC , and the putative PGK LmjF30 . 3380 - could complement a yeast pgk1 deletion mutation . We found that only PGKB could complement the yeast deletion and resulted in a slow-growth phenotype ( approximately 64% of the maximum growth rate of a wild-type strain ) even with full expression of the heterologous protein ( Figure 1 ) . Plasmid maps for all of these plasmids are shown in Figures S1 , S2 , S3 , S4 , S5 , S6 , S7 , S8 and the full sequences of the plasmids can also be found in the Supplementary Material ( Text S1 , S2 , S3 , S4 , S5 , S6 , S7 , S8 , S9 , S10 , S11 , S12 , S13 , S14 , S15 , S16 , S17 , S18 , S19 , S20 , S21 ) . We cloned cds for all of the heterologous proteins into pCM188 plasmids , under the control of a TetO2 promoter , to allow the modulated expression of each drug target in yeast . TetO2 is a powerful constitutive promoter but can be repressed by the addition of a tetracycline analogue ( doxycycline ) to the growth medium . Thus , normal growth conditions result in the full expression on the cloned cds; , however , by adding doxycycline , we can reduce the expression of the parasite or human cds to levels that should be inversely proportional to the concentration of doxycycline added . We measured the growth rate of the parasite mimetic or human mimetic yeast strains in the presence of 2 , 5 , 10 or 20 mg/L doxycycline and found that the reduction in growth rate of yeast expressing human or parasite cds for either DHFR or NMT were relatively insensitive to the addition of doxycline ( ≤ 25% ) growth-rate inhibition ( Figure S9 ) . In contrast , addition of doxycline to strains express cds for PGK displayed up to 75% growth-rate upon doxycycline treatment ( Figure S9 ) . While these results indicate that only small amounts of the heterologous enzymes are often sufficient to achieve full complementation of the yeast deletion mutation , we demonstrate below that reduction in enzyme concentration achieved by inhibiting promoter activity with doxycycline is sufficient to enhance the sensitivity of yeast growth to the action of drugs that inhibit the activity of target enzymes to a useful extent . PGKs and NMTs are fairly novel drug targets and therefore we had no access to specific inhibitors for the different species tested . However , Plasmodium DHFRs are classic drug targets , which can be specifically inhibited by the antimalarial drug pyrimethamine . Therefore , we performed drug sensitivity assays ( on agar plates ) of the different yeast transformants expressing parasite mimetic or human DHFRs in the presence of various concentrations of pyrimethamine . We found that yeast strains expressing Plasmodium vivax DHFR ( yPvDHFR ) showed sensitivity to concentrations of pyrimethamine ranging from 10 to 500 µM ( Figure 2 , upper panels ) . Yeast strains expressing Plasmodium falciparum DHFR ( yPfDHFR ) showed sensitivity to concentrations of pyrimethamine ranging from 100 to 500 µM ( Figure 2 , upper panels ) . No other strain showed sensitivity to the pyrimethamine concentrations tested . Importantly , by modulating the expression of the heterologous cds by adding 5 mg/L doxycycline to the growth media , it was possible to increase the sensitivity of the strains to pyrimethamine by approximately 50-fold ( Figure 2 , lower panels ) . With the aim of further increasing the sensitivity of our assays , we deleted the PDR5 gene , which encodes the major yeast multidrug export pump , from all of our strains and tested the effect of such a deletion on the pyrimethamine sensitivity of strains expressing heterologous DHFRs . We spotted ( using a Singer RoTor robot ) serial dilutions of PDR5 and pdr5Δ strains onto rectangular agar plates containing a series of concentrations of pyrimethamine with or without doxycycline ( 5 mg/L ) . Following quantification of the data ( Figure S10 ) , we observed that removal of the multidrug export pump by the pdr5Δ deletion significantly increased the strains' sensitivity to pyrimethamine , such that ( in the presence of doxycycline ) the sensitivity of the Trypanosoma brucei DHFR could be observed ( Figure 3 ) .
Yeast cells are suitable hosts for the expression of heterologous proteins from various species , including enzymes essential to the different life-cycle stages of parasites . Some of these parasite life stages cannot be propagated under laboratory conditions , and thus yeast provides a practical and flexible platform for in vivo drug screens . Here , we have reported the construction of a series of yeast strains that are identical apart from the cds for different heterologous drug targets ( DHFR , PGK or NMT ) which they express . For all of these targets , we also constructed yeast strains expressing the cds for the equivalent human enzyme , thus permitting the design of screens for agents that discriminate between the parasite and human targets . Sibley and co-workers [22] , [40] , [41] have done extensive work on the complementation of yeast dfr1 mutations by overexpression of human and Plasmodium DHFRs and demonstrated the suitability of yeast strains for drug screens in plate assays . Now , we have demonstrated that the DHFR cds from Schistosoma mansoni , Leishmania major , Trypanosoma cruzi and T . brucei can also successfully complement a dfr1 null mutant of yeast . Furthermore , we have found that yeast cells expressing T . brucei DHFR are partially sensitive to pyrimethamine ( Figure 2 ) and that strains expressing T . cruzi or S . mansoni DHFRs are sensitive to the chemotherapeutic agent methotrexate ( data not shown ) . In addition , we have also successfully complemented yeast deletion mutants with cds for two new potential drug targets from parasites: N-myristoyl transferase ( NMT ) and phosphoglycerate kinase ( PGK ) . To the best of our knowledge , this is the first evidence for the functional expression of these parasite enzymes in yeast . While all of the data presented in this paper relate to screens against yeast cells spotted onto agar plates , we have confirmed all the results for DHFR using yeast strains grown in liquid cultures in 384-well microtiter trays . Indeed , we intend to use such liquid cultures in HTP screens based on the series of yeast strains reported here . Yeast cells may be refractory to chemotherapeutic drugs due to their thick cell walls and the high levels of expression of multiple drug efflux pumps . Thus the genetic manipulation of yeast strains for use in screens in order to increase their drug sensitivity is highly desirable . We have improved the drug sensitivity of our yeast constructs in two ways . First , all cds encoding heterologous enzymes that represent drug targets were placed under control of the TetO2 promoter , so that their expression could be down-regulated by addition of the tetracycline analogue , doxycycline . We demonstrated that additions of doxycycline to the growth medium enhanced the pyrimethamine sensitivity of parasite DHFR enzymes expressed in yeast ( Figure 2 ) . Second , we demonstrated ( again using DHFR and pyrimethamine ) that the deletion of the gene PDR5 , which encodes the major yeast ABC transporter , can increase the sensitivity of drug screens ( Figure 3 ) . Moreover the use of pdr5Δ mutants in conjunction with pyrimethamine has a synergistic effect on drug sensitivity ( Figure 3 ) . The ease of genetically manipulation in yeast should allow further improvements in drug sensitivity in the future; for instance , by modifying the sequence or expression of genes ( such CWP1 , CWP2 , or PSA1 ) that encoding cell wall proteins [59] . We believe that many more enzyme targets and parasite species can be studied using a similar approach and are currently employing some of our strains in high-throughput drug screens using a Robot Scientist [60] , [61] , [62] in search of novel antiparasitic agents .
|
Parasites kill millions of people every year and leave countless others with chronic debilitating disease . These diseases , which include malaria and sleeping sickness , mainly affect people in developing countries . For this reason , few drugs have been developed to treat them . To make matters worse , many parasites are developing resistance to the drugs that are available . Thus , there is an urgent need to develop new drugs , but this is hampered by the fact that most parasites are difficult or impossible to grow in the laboratory . To address this , we have engineered baker's yeast to be dependent on the function of enzymes from either parasites or humans . In all , our engineered yeast constructs encompass six parasites ( causing malaria , schistosomiasis , leishmaniasis , sleeping sickness , and Chagas disease ) and three different enzymes that are known or potential drug targets . Further , we have increased yeast's sensitivity to drugs by deleting the gene for its major drug efflux pump . Because yeast is robust and easy to grow in the laboratory , we can use a robot to screen for drugs that will kill yeast dependent on a parasite enzyme , but not touch yeast dependent on the equivalent human enzyme .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"biotechnology",
"medicine",
"biochemistry",
"infectious",
"diseases",
"model",
"organisms",
"biology",
"microbiology"
] |
2011
|
Functional Expression of Parasite Drug Targets and Their Human Orthologs in Yeast
|
Morphology typically enhances the fidelity of sensory systems . Sharks , skates , and rays have a well-developed electrosense that presents strikingly unique morphologies . Here , we model the dynamics of the peripheral electrosensory system of the skate , a dorsally flattened batoid , moving near an electric dipole source ( e . g . , a prey organism ) . We compute the coincident electric signals that develop across an array of the skate's electrosensors , using electrodynamics married to precise morphological measurements of sensor location , infrastructure , and vector projection . Our results demonstrate that skate morphology enhances electrosensory information . Not only could the skate locate prey using a simple population vector algorithm , but its morphology also specifically leads to quick shifts in firing rates that are well-suited to the demonstrated bandwidth of the electrosensory system . Finally , we propose electrophysiology trials to test the modeling scheme .
Sensor placement and arrangement are crucial for any sensory system used to locate the distance and bearing of a stimulus source , and this is true for a variety of modalities . In addition to stereo vision in much of the animal kingdom , echolocation ( e . g . , the barn owl ) and mechanosensory location ( e . g . , the sand scorpion ) uses multiple sensors to locate prey with requisite precision [1–3] . Electrosensitive vertebrates , including monotremes ( e . g . , the platypus ) , siluriforms ( e . g . , the catfish ) , osteoglossomorphs ( e . g . , the knifefish ) , and chondrostreans ( e . g . , the sturgeon and the paddlefish ) , typically possess large populations of independent electrosensitive organs . Obviously , multiple organs enhance signal-to-noise ratio by an increased number of simultaneous measurements since environmental electrical signatures of biological relevance are often weak , even at close range . However , these animals may also use the sensor population to determine the location of electric sources . In the case of the black ghost knifefish , Apteronotus albifrons , the surface receptors simply report magnitudes of electric field , and the contrast of signal magnitudes across the body facilitate the location of nearby objects and the capture of prey [4 , 5] . The elasmobranch fishes ( sharks , skates , and rays ) possess an electrosensory system used to detect prey , possibly navigate with respect to magnetic fields , and locate mates [6–9] . The system includes hundreds or thousands of separate electrosensor units known as the ampullae of Lorenzini . The ampullae are often tightly clustered , but each is linked to an individual pore on the surface of the body via a long , gel-filled canal ( see Figure 1 ) , and the pores are widely distributed ( on the head and pectoral fins in skates and rays ) . Each ampulla is able to code minute electrical fluctuations into discharge patterns of primary afferent nerves [10] . The sensing cells of an ampulla's epithelium detect electric potential differences between its apical side within the ampulla , and its basal side outside the ampulla . A sudden drop in apical-side potential with respect to the basal potential leads to an increased firing rate for the associated nerves , while an increase in that potential difference leads to an inhibited firing rate [11] . More recent measurements also delineate organ-specific and frequency-specific gain functions [12 , 13] . The amplification mechanism of the sensory epithelium is not completely understood , but persuasive models exist [11] . While previous studies have focused on the response properties of single electroreceptors , few have explored the simultaneous response of electroreceptor populations to biological stimuli . This aspect is especially relevant since the pores and canals associated with the electroreceptors display great geometrical variation within a given organism and among species [14] . In terms of higher processing , output from principal cells in the dorsal octavolateralis nucleus ( DON ) of the skate hindbrain showed systematic responses to oscillating dipole stimulation that were twice as strong as responses to uniform body-wide fields [15] . Hence , it is important to understand the coincident responses of the electrosensory periphery to electric stimuli to understand how central processing mechanisms integrate peripheral receptor responses and affect behavior . Kalmijn has proposed an algorithm for elasmobranchs approaching stationary prey via the electric sense [16] . In this model , the hunting elasmobranch simply maintains its orientation with respect to the dipole field of the prey . Geometrically , this means that the elasmobranch will always arrive at the source , even if this sometimes means an inefficient , spiraling path . In an initial simulation effort , one of us showed that the Kalmijn algorithm would typically mean that an elasmobranch moves in such a way to reinforce the signal received by each electroreceptor [17] . However , recent behavioral analyses of sharks approaching artificial electric dipoles show that the animals usually exhibit sharp turns toward the electric field source [18] . Thus , it appears that information regarding the source's location and/or the decision to approach the source is developed rather quickly by the predator . These observations are not necessarily at odds with the Kalmijn algorithm , however . The algorithm could guide a creature for a brief period of time until an overwhelming strength of signal among the sensor population helps it establish the exact location of the source . From the electrosensors , primary afferents project to principal cells of the DON in the medulla [19] . Ascending efferent neurons in the skate have a strong ascending projection from the medulla to the contralateral midbrain , where they converge in the lateral mesencephalic nucleus and the tectum , combining there with other sensory inputs [20] . Self-generated electrical signal subtraction has been convincingly documented in the elasmobranch DON [19 , 21] . While the afferent fibers show vigorous reactions to an elasmobranch's own ventilatory movements , the ascending efferent neurons demonstrate so-called common-mode suppression . In addition , electrical noise created by self-generated movements may be cancelled by an adaptive filter via anti-Hebbian learning of ascending efferent neurons in the hindbrain [22] . ( By anti-Hebbian we mean responses to familiar , repetitive stimuli are suppressed in favor of novel stimuli . ) However , how the combined neural responses from the periphery are processed by the DON , the midbrain , and higher centers to determine prey headings is unexplored in sharks and batoids . Here , we model the system-wide signals and neural responses of an elasmobranch fish . In contrast to a prior modeling effort [17] , we now use precise , biologically relevant morphological measurements and move beyond electrical signals within the ampullae to the population codes of neural information available to the elasmobranch central nervous system . The canals in sharks' electrosensory systems thoroughly map a 3-D space . While this is computationally straightforward , data representations and interpretations are cumbersome . Therefore , we examine several scenarios for the peripheral electrosense of the dorso–ventrally flattened barndoor skate , Raja laevis , moving near an electric dipole . Adults are typically found at significant benthic depths along mud floor troughs , where they skim along the dark sea floor to feed on large crustaceans , mollusks , worms , and flatfish [23] . We model a barndoor skate moving near an electric dipole to represent the approach to a stationary prey , but these computations can also represent a moving dipole source that approaches a stationary skate . On average , the barndoor skate possesses a greater number of electroreceptors than other skate species [24] . The peripheral electrosensory geometry is precisely mapped [14] , and it presents an almost perfectly 2-D system . Where prior efforts stopped at voltage calculations [17] , here we take two further steps: we assess afferent spike trains of the peripheral system , and present a basic model that averages these afferent inputs to ascertain the location of a bioelectric field's source .
For the case of the skate moving with respect to a live source , we first compute the realistic voltage signals that develop in the ampullae of Lorenzini using the skate's measured canal geometry . In this work , we concentrate on the bilateral canals of the hyoid clusters that project to the dorsal surface of the animal ( see Figure 2 ) . Of approximately 1 , 400 total ampullary canals in this species , we choose to focus on the dorsal hyoid cluster for two main reasons . First , we narrow the focus for the sake of clarity and lucid data management; we show results for the signals arising from just 132 ampullae . Second , we choose the dorsal hyoid clusters because they exhibit both the greatest canal lengths ( corresponding to the lowest signal thresholds; [12] ) and also the greatest variation of canal orientations when compared with other clusters . These clusters appear to be best suited for long-range detection when compared with the short canals of the rostral surfaces and the mandibular cluster on the jaw that presumably assists positioning for the final stages of feeding . Though some authors have treated each canal as an equipotential or cable-like contact , linking the pore directly to its associated ampulla [25–27] , we agree with the viewpoint that significant potential differences can and will develop within the length of the canals [28] . Here , consistent with an earlier simulation [17] , we emphasize the potential difference that develops between an electroreceptor and its associated pore , due to the electric field originating in the physical electric dipole P of small bioelectric source ( see Figure 3 ) . The origin for our computations sits at the source electric dipole . The potential at all other points , referenced by the position vector r , is computed for an ideal electric dipole [29] , Here , P is the source's dipole moment vector , which includes the orientation direction of the source; r is the distance from the source to the point of interest; er is the unit vector pointing from the source to the point of interest; and ɛ is the static permittivity of seawater ( see Figure 3 ) . Our entire computation is 2-D , approximating the dorso–ventrally flattened skate skimming the ocean floor in search of prey . As originally noted by Murray , maximum excitatory or inhibitory response occurs when an electric field vector is parallel or antiparallel to a canal [10] . This geometrical observation is fundamental to our model , as this alignment maximizes the potential difference between a pore and its associated ampulla . We refer to the geometry of Figure 1 to describe our assumptions and computations of relevant potential differences in the electrosensory periphery . There are five geometric points of interest , labeled A through E . The two ampulla and respective canals belong to the same cluster here . We assume that the basal voltage will be shared for each ampulla in a cluster ( Figure 1 , point E , within the sac that surrounds the cluster ) . Far from a dipole or other electric source , we assume all potentials at pores and ampullae to be more or less equal , and the two organs will exhibit the same resting firing rate . Fluctuations in the ampulla relative to the electric potential at point E will determine excited or inhibited firing rates for each organ . Since we treat the potential at Figure 1 , point E , VE , as a constant for all organs within a given cluster , the crucial quantities become the apical electric potentials within the ampullae ( at points B and D in Figure 1 ) . In what manner would these electric potentials vary ? Will they communicate electric potentials from their pores ( Figure 1 , points A and C ) ? Recent electrical measurements of the gel demonstrate that the gel-filled canals do not provide good electrical contact between pores and ampullae [30] . The gel exhibits much stronger capacitive properties ( charge storage ) than seawater or generic collagen gels , suggesting a canal is better suited to sustaining electric potential differences than it is to bringing its two ends to quick electric equilibrium [30] . In fact , the relative electric resistance along either canal ( e . g . , from points A to B , or from points C to D; Figure 1 ) is more than 100 times greater than the electrical resistance between the pores A and C ( in the semi-infinite seawater medium ) [30] . Based on these measurements , and based on the observations of Murray regarding the crucial nature of geometric alignment between the canal and applied electric fields [10] , we believe that the potential differences evolving along canals will provide crucial voltages in the system , with a driving influence for the apical surface of the sensory epithelium . The electric potentials between nearby pores would approach the same electric potential before the electric potentials at either ends of a canal would equilibrate . Hence , as a skate draws close to an electric source , the sharply inhomogeneous source field will set up significant differences along the canals , depending on their length , angular orientation , and relative distance from the source [17] . Clustered ampullae will thereby be allowed to develop distinct firing responses , as each develops distinct apical potentials . By treating the basal electric potential ( VE , using the geometry of Figure 1 ) as a constant for all ampullae in a given cluster , the model's chief simplification is a focus on the voltages that arise along the interior length of the canals . These lead to variations at the apical surface of the sensory epithelium , and with the basal potential treated as a constant , the apical variations drive the resulting firing responses . Our model treats the sensory epithelium as a “black box” that translates transepithelial electrical fluctuations into firing rate alterations ( see Analysis below ) . We do not imply that it plays a minor role or that its electrical characteristics are trivial . The epithelium's behavior is best described by multiple circuit elements acting in concert , with some functioning as a strong impediment to current flow; as a distinct electrical entity , its measured impedance exhibits great variation , presumably resulting from the behavior and condition of epithelial ion channels [11] . These details are not germane to our model , and we rely on the epithelium's measured and predictable response to electrical stimuli . At successive moments in time , as the skate moves with respect to the source dipole in close range ( within 1 m of distance ) , we assign an electrical signal Vsignal to each electrosensor by computing the electric potential at the various pores and associated ampullae , and then using the expression in which Vpore is the potential at the pore and Vampulla is the potential in the internal ampulla chamber ( apical side of the sensing cells ) . For instance , using Figure 1 , the electric signals are computed as the relative potential differences respectively from points B to A and from points D to C . Here , both Vampulla and Vpore are computed using the classical expression for the dipole electric potential . This is exactly equivalent to the path integral treatment used in a previous effort [17] . Given the simplifications described above , we do not pretend that our data will be the precise signals leading to firing alterations in the skate—however , we believe our computations capture the dominant component of firing alterations that arise when the skate moves with respect to an electric field source . For the source dipole we use a magnitude that is consistent with bioelectric fields measured for small prey , with dipole values in the 10−15−10−16 Cm range [28] . While actual bioelectric fields do not conform exactly to an ideal dipole field , using such fields makes for an excellent approximation for distances greater than the physical extent of the source itself . When comparing the ideal dipole field magnitude to the slightly more realistic physical dipole ( separating the positive and negative charge centers ) , we find that the ideal case gives values nearly identical to the physical dipole as long as the skate-to-source separation is more than two to three times the size of the source . For instance , if the source were a bivalve with 3 cm separating its relatively positive and negative charge centers , the ideal dipole approximation would be quite accurate as long as a skate was more than 6–9 cm away . This approximation will be valid for the cases explored here . In early simulations , one of us used a dipole moment strength of 5 . 6 × 10−16 Cm [17] . This produced values for Vsignal in the range of tens to hundreds of nanovolts for the skate–source distances of interest . These low values congregate at the very center of the empirical gain function for computing afferent firing rates ( see Analysis ) . In this region , the empirical fitting is most speculative . Therefore , we increased the dipole moment strength for the current simulations , using a value of 3 × 10−15 Cm ( i . e . , around five times the original choice ) . This is not far off from measured values for small prey sources [28] , and renders values for Vsignal in the low microvolt range , where the translation to firing rate alterations is well mapped . We translate ampullary signals from Equation 2 into firing rate alterations of the primary afferent neurons associated with the ampullae . Again , treating the sensory epithelium as a “black box , ” the firing process is modeled with an ad hoc firing rate gain function based on the known electrophysiology literature . The model consists of a firing rate function ri ( t ) generated in a given ampulla by the corresponding Vsignal . The index i refers to a specific ampulla–canal system , and t represents time . Firing rate functions ri ( t ) are obtained by multiplying the indexed Vsignal values by a universal empirical gain function . To build this gain function , we note that the resting tonic rate in the absence of electric fields is known to be in the range of 30 to 40 Hz [11] . Data from the thornback ray suggest 100% change in rate for each 5 μV of apical voltage fluctuation [19] , where negative drops are excitatory and positive fluctuations are inhibitory , with a somewhat sigmoidal overall shape . While those points provide anchors for the ampullary gain function , the shape was derived by digitizing the experimental points found in electrophysiology experiments with skates [11] , to which we applied the standard Levenberg-Marquardt algorithm for nonlinear fitting . The resulting canal gain function can be seen in Figure 4 . In this treatment , we factor neither the natural relaxation of voltage signals within the ampullary system nor the natural relaxation of firing rate alteration in the primary afferents . These relaxations will be dominated by the rapidly developing electric potential changes over the short time scales of the skate–dipole interaction ( the most dramatic activity in the results that follow include 1–2 s of closest approach ) . As noted for weakly electric fish , electric perception is limited to such short range that instantaneous data are presumably of fundamental importance [31] . In addition to raw firing rate data , and to further explore the potential fates of firing rates entering the DON , we use the concept of firing-rate population vector , [32] which was used successfully in the prey location analysis for the sand scorpion and the clawed frog [2 , 3] . We use a basic form of a population vector , where its heading ( or yaw ) in the horizontal plane of the skate is determined by a weighted average of the headings of the various canals . The weighting of each canal orientation is simply based on its electrosensor's firing rate . We thus consider a canal with heading θ to correspond to a unit vector ( cos θ , sin θ ) on the canal plane . We do this for both the right and left canal clusters . The population vector ( px , py ) is then given by the weighted sum of individual unit vectors: We take this vector to be a kind of compass heading that can bias the animal's orientation . In other words: While we cannot confirm that the DON actually computes such a population vector , it shows , at the very least , what information is available to the skate from its peripheral electrosense .
To illustrate the first step of the calculations , we present an example “snapshot” of signal voltages for the skate near a source . Figure 5 displays potential differences developing along the canals in Figure 2 for a skate approaching a source that is 30 cm to its left and 10 cm in front of it ( as in Figure 3 ) . Naturally , canals in the left hyoid ampullary cluster show the stronger signals . We now present two example “swim-by” scenarios , with a skate swimming in a straight path past a nearby source ( Figure 6 ) . The responses presented here do not assess orientation behaviors; instead , they monitor the sensory information available to the skate as it moves past a bioelectric signal . In the first scenario , a skate swims at 0 . 5 m/s past a source dipole with a closest approach distance of 0 . 15 m , in a direction parallel to the orientation of the source's dipole . In the second scenario , a skate swims again at 0 . 5 m/s , but with a direction no longer parallel to the source's dipole ( 45° ) . For comparative purposes , the closest approach distance is again set to 0 . 15 m . As the skate moves in relation to the source in each of the two swim-by scenarios , the electric signal Vsignal associated with each electrosensory canal will change . Consequently , the associated firing rates also become functions of time ( i . e . , a function of environmental variables such as skate–source distance and relative orientation ) . Figures 7 and 8 present firing rate snapshots taken at different moments in each scenario . As seen in Figure 6 , canals in the right cluster for scenario 1 are closer to the source during the entire swim trajectory , and their associated firing rates show the largest variations , as depicted in Figure 7 . Note that at skate–source distances greater than 0 . 30 m , each firing rate is essentially at the baseline value of around 34 spikes/s . As the separation distance decreases , the canals' electric signals change in a nonuniform way . This leads to nonuniform firing rate profiles associated with each canal , as dictated by the firing rate gain function of Figure 4 , with the maximum change from the baseline firing occurring around the closest skate–source approach . Analogous to the previous case , canals in the left cluster for scenario 2 are closer to the source during the entire swim-by , and their associated firing rates show the largest variations , as depicted in Figure 8 . We again note that at skate–source distances larger than 0 . 30 m , each firing rate is essentially at the baseline value . To further explore how much neural information is available to the skate , we compute a firing rate population vector for each dorsal hyoid ampullary cluster ( Equation 3 ) . Figures 9 and 10 depict the resulting population vector magnitudes versus time for each scenario , charting the 4 to 5 s surrounding the point of closest approach . We depict the global or net population vector ( i . e . , summing for all hyoid canals ) as well as the bilateral left and right clusters individually . Inputs from left and right are not combined in the neural architecture until they reach the skate's midbrain , so the left and right vectors may be relevant to the DON ( e . g . , processing common mode rejection ) . Also , note that the most significant variations occur within a total time of approximately 1 s . Like any vector , the population response includes angular ( or heading ) information , which we relate to the actual heading of the source ( defined here as the angle between the skate's direction of motion and the vector pointing to the source position ) . Relevant angles are depicted in Figure 11 . It is important to note that the dipole angle and the direction of the skate's motion remained fixed in each of the swim-by scenarios , while the heading of the source would naturally vary given the skate's motion . Figures 12 and 13 depict heading information from the global or net population vector . We plot the vector headings and the actual heading of the source dipole relative to the skate over time during each encounter . Of note are the abrupt discontinuities when the skate reaches the point of closest approach to the source , and the near perfect match ( up to a constant phase ) to the actual heading . In all of our simulations , population vector headings were defined with respect to the skate's longitudinal axis , following the usual counterclockwise mathematical convention . We note that such a human-centric system of angle definition need not bear any relation to the skate's neural processing of its electrical landscape . Hence , we believe constant offset phases ( e . g . , 90° alternately added to Figure 12 and subtracted from Figure 13 ) , can be considered without decreasing the predictive nature of the modeling data . In other words , we do not assert that the net population vector should act like an orientation compass toward the dipole source . Rather , we point to relevant angular information encoded by the neuronal population's spiking activity . Finally , we suggest experimental tests of the model and its assumptions . The following analysis can be tested directly by benchtop electrophysiology measurements in which investigators use dipoles in various positions to map receptive fields in anesthetized skates ( e . g . , [15] ) . We note that existing data do not track dipole orientation . In this simulation , a dipole is placed near a stationary skate , and the dipole is simply rotated in the horizontal plane of the skate without changing its position ( Figure 14 ) . This experiment has not been conducted , to the best of our knowledge . By monitoring the resulting firing rate activity and computed population vectors , we can predict “sweet spots” or regions in which the electrosensory system of the skate should show specific and varied reactions to relatively small changes in dipole orientation . Given the sharp dependence of our modeling system on the interplay of the canal geometry with the dipole field , a significant variation in neural response results . The firing rate snapshots of Figure 15 show differences that arise in the ampullae simply by changing the relative angle between the dipole and the skate . Such sharp contrast would not follow from a model where pore potentials were communicated directly to the apical side of sensing cells [25–27] . Slight rotations do not change pore magnitudes of electric potential as appreciably as they change the canal potential differences considered in our model . Another way to exhibit the spiking activity dependence on geometry is to plot the magnitude of the population vector for each canal cluster versus the external dipole angle ( Figure 16 ) . Here , as in Figures 9 and 10 , the population vectors for both canal clusters have a magnitude of around 13 . 7 Hz when the source is far away or absent ( i . e . , Vsignal for each canal is essentially zero ) . The maximum firing rate variation corresponds to an external dipole angle around 60° for the right cluster and 120° for the left cluster . The opposing nature of the geometric relationship between the skate and the dipole creates a predominance of positive electric potential signals , which have an inhibitory effect on firing rates ( see Figure 4 ) . A similar but excitatory result would be obtained if the external dipole were behind the skate , rather than in front of it .
Though bilaterally symmetric , the canal arrays are asymmetric on either side as one moves from the anterior to posterior regions of the skate . This has a dramatic effect on the encoding of neural data ( e . g . , Figure 5 ) , and we wish to emphasize two features in particular . First , the suppressed and fairly uniform signals exhibited by ampullae in the right cluster ( farther from the dipole ) are dramatically different from those of the left cluster ( closer to the dipole ) . This type of uniformity in the contralateral signals would not emerge if the ampullae responded directly to pore potentials , and it does not emerge in data from a homogeneous array ( see Figures 17 and 18 below ) . Second , there is an abrupt reversal of signal polarity for the short groups of canals on each side that project medial on the body ( canals 55–66 and 121–132 in Figure 2 ) . These subgroups effectively mimic the opposite cluster; such a feature could presumably be of great use to the skate in common mode suppression or signal amplification at higher levels of neurosensory processing . For the firing rates computed , it is important to note that the variations revealed by our simulations are of significant magnitudes for even a simple neuronal system to respond accordingly . In a strict mathematical sense , this nonuniform firing rate profile encodes enough information for a precise location of the source , although empirical neurophysiological evidence is required to determine the processing of this information by the skate central nervous system . The two main swim-by scenarios present fundamentally different cases . In the first , the source and the skate are aligned , and the source sits to the right of the skate's path; in the second , the source and the skate are not aligned , and the source sits to the left of the skate's path . While the skate clearly receives different information in these two scenarios , the global qualitative similarities of the firing rate snapshots for scenarios 1 and 2 overwhelm their differences . First we note that firing from ampullae on the side opposite of the source are nearly unaffected , even when the skate passes within 15 cm of the source . This corresponds to the left cluster in Figure 7 and right cluster in Figure 8 , and is consistent with the relative potential data shown for the right cluster in Figure 5 . Presumably , such consistent lack of excitation or inhibition would be very beneficial to the skate's source location because it will provide a reference for contralateral excitation and inhibition by the source bioelectric field . Again , this is not the case for an artificial array of homogeneous canals . Next , we note the dramatic changes for steps D and E for each trial . A sharp pattern of excitation and inhibition arises for the cluster nearest the source during the approach ( Figures 7D and 8D ) . While some of the more anterior organs ( e . g . , canals 1–15 in Figure 7 ) show firing excitations , ampullae corresponding to the more posterior canals exhibit nearly complete inhibition ( e . g . , canals 35–45 in Figure 7 ) . As the skate passes and moves away from the source , consistent excitation emerges across the ipsilateral side . In particular , note that the organs experiencing inhibition on approach shift abruptly to excitation . In Figure 7D and 7E , for instance , organs 35–45 shift from near total inhibition to near 100% excitation over a travel distance of only a few centimeters by the skate . Not only would sharp changes provide for simple interpretations , the rate of change demonstrated in these simple trials fits the response characteristics of the ampullae . Given the skate's speed of 0 . 5 m/s , the organs in question experience this dramatic change over about 0 . 1 s , or at a frequency of 10 Hz . The time signature of this signal is within the ideal ampullary bandwidth of these organs [11 , 19] . And when considering the natural relaxation of firing rate alterations—afferent rates typically accommodate to a constant stimulus in less than 5 s [10]—the relatively quick changes shown in our trials again are within these neurophysiological constraints of the skate primary afferents . Both trials demonstrate asymmetry between approach and retreat situations , thus providing clear information concerning the anterior versus posterior source location . We note with interest that the largest excitations were observed for sources in anterior locations to the body and after the skate passed the dipole source . This observation is consistent with the functional subunit hypothesis [17] , and indicates that the dorsal hyoid clusters may best detect and encode information concerning anterior bioelectric sources . As skates are not just predators but are also potential prey , our results suggest that these clusters may be used to detect the approach of a large predator or conspecific . The dorsal location adds weight to this speculation . In terms of electric source localization , the net population vector headings appear in Figures 12 and 13 . Despite very different geometric scenarios in scenarios 1 and 2 ( e . g . , opposite relative source-to-skate position and orientation ) , the data relative to the actual source headings are virtually identical . The population vector headings follow the change of the actual source headings with precision , with two notable deviations . First , the population vector headings do not match the exact values of the source headings; second , an abrupt shift of direction occurs at the point of nearest approach . The fact that the precise angle values of the computed vectors do not match the apparent source headings is of very little concern , since we have imposed a human-centric system of angle definition ( e . g . , Figure 11 ) . Moreover , a skate could presumably compensate for such regular offsets , much as visual processing inverts the actual image on the retina . The abrupt shift at the point of closest approach ( at the 2-s mark ) coincides with the magnitude of the net vector falling briefly to zero . ( See solid lines for net vector magnitudes in Figures 9 and 10 . ) We have confirmed that these shifts are absolutely independent of our choice of reference frame—the shifts are not an artifact of our angle definitions . As the skate moves from one side of the source dipole to the next , such a shift could be of significant biological benefit . The abrupt changes of vector magnitude and direction presumably give the skate a very clear signal as it comes very close to a prey , predator , or mate , and transitions from approach to moving away . In summary , the skate electrosensory array can provide a wealth of information on the heading of a nearby bioelectric source by using a simple population vector scheme . The morphology is well suited to tracking the source location , including information concerning whether the skate is moving toward or retreating from a source . In addition , population vector magnitudes vary sharply at the position of closest approach . This presumably supplies clear neural information that the source is within critical , minimum distance associated with the given swimming trajectory . Moreover , the changes in firing activity happen over a time scale that is ideally suited to the frequency response of the electrosensors . It is not unreasonable to suggest that the overall shape of R . laevis evolved at least in part for the tuning of electrosensory tasks , especially as the dorsal hyoid canal pores extend to the periphery of the skate's wings ( Figure 2 ) . The location of the hyoid pores follows this pattern for the canal systems of most skates [24] . Does the skate morphology confer an obvious advantage over , for instance , a simple homogeneous array of canals ? We ran the same simulations for an artificial set of dorsal hyoid electrosensors with associated canals of uniform 10-cm lengths and equal angular spacing covering 360° of horizontal arc . Figure 17 depicts a homogeneous array of 132 canals , with 66 per cluster , as in the skate ( compare with Figure 2 ) . The two clusters have the same lateral spacing as those in the skate . The canal lengths of the artificial array are set to 10 cm , the average length of those for the skate's dorsal hyoid cluster; the angular distribution is similarly uniform , with canals placed every 2 . 78° , as opposed to the skate's actual canals , which have some densely spaced canals in terms of orientation , while some others are spaced more broadly . We compute a snapshot of Vsignal , according to Equation 2 , for each of the canals , exactly as we did for the skate array . The results are shown in Figure 18 ( compare with Figure 5 ) . The significant differences for the skate versus the homogeneous , artificial array are as follows . The skate signals are , naturally , of much greater range , reflecting the range of canal lengths—the signals of the skate's system are three times that of what a homogeneous array would offer . Even if one adjusts the canal lengths of this artificial array to the maximum skate canal length , however , a more fundamental difference remains . The canal-to-canal signal differences vary in a moderate fashion for the artificial array , while those in the skate vary dramatically , even for some nearly adjacent canals . Figure 5 exhibits signals that change by approximately 50% over as few as three canal spacings , while Figure 18 shows that a similarly dramatic difference can only be obtained over 15–20 canal spacings . Further computations of firing profiles and population vector magnitudes exhibit the same fundamental differences shown in Figure 18 . The size of signals are typically smaller for the artificial canal array , and in all cases the signals change more gradually , as shown both over the array of canals , and also over time , in the case of the population vector magnitudes . These trials confirm that a homogeneous array can also track the source dipole , but at least two distinct advantages for the skate's varied canal morphology emerge . First , the “snapshot” canal voltages ( see Figure 5 and Figure 18 ) are much less dramatic for the artificial array; the contralateral cluster does not exhibit a homogenous set of voltages , and the variation of signals within the near-side cluster exhibits a much less pronounced change . In essence , the homogeneous , artificial array would not provide the same sharp contrast of signals between organs . Second , though firing rates and population vectors change for the artificial array , they change more slowly over time . As noted previously , referencing Figure 7 , a subset of canals can change their firing rate from near 0 to 50 Hz over about 0 . 1 s as the skate moves past the dipole . Similarly , the skate's morphology provides population vector magnitudes that nearly double in less than 0 . 25 s , while a similar change in the artificial array takes a second or more . This is crucial when one considers that the electrosensory system of the skate is finely tuned to changes in a narrow bandwidth peaked between 2 and 10 Hz [10–12 , 19] . Hence , an abrupt change evolving in a fraction of a second would be much more useful than a change that takes a second or more . To correctly model the function of the electrosensory periphery , it is critical to understand the excitation of canal arrays by bioelectric fields . We have proposed a simple , specific test for the modeling scheme advocated here . We identify an area of specific opposite predictions from our model and that of cable-like , equipotential canal function . As the dipole of Figure 14 turns between 0°–90° , the anterior pore potentials ( e . g . , for canals 0–50 in the left cluster ) become increasingly positive , via the dot product of Equation 1 . Hence , a cable-like canal function , conveying the pore potential to the ampullae , would predict inhibited firing rates for increasing apical potentials . However , our model predicts the opposite , as can be seen in Figure 15A–15C . The turn between 0° and 90° yields negative Vsignal values and excited firing rates . Furthermore , according to the data shown in Figures 15 and 16 , a quick dipole rotation from just 0° to 45° would appreciably alter not only the firing rates associated with the anterior canals but also the population vector magnitudes . These changes could be exhibited in primary afferent rates , ventilatory activity , heart rate , or even DON activity . A traditional theory by which canals simply convey pore voltage signals to the ampullae would not be as sensitive to the canal's overall alignment with the dipole field vectors . We have modeled the neural responses of only a small subset of the approximately 1 , 400 ampullary canals in the barndoor skate . While we chose the approximately 132 hyoid canals that project to the dorsal surface of the skate's body for their apparent sensitivity and to present clear and finite data , future efforts incorporating all clusters are required to understand central processing and integration of receptor population data . Even more dramatic and discontinuous rostral and caudal projections are found in the dorsal canals of the skate superficial ophthalmic cluster [14] . In addition , there are approximately 700 total hyoid canals that project to the ventral surface . These canals are more continuous in distribution , and have prominent contralateral projections that cover a wider angular field . Even more dramatic differences in canal projection vectors exist among the dorsal and ventral buccal clusters . Application of the electrodynamic model that integrates responses of these ampullary groups would provide a more comprehensive model of electrosensory processing of environmental fields in feeding , social behavior , and navigation . In addition , it would permit development of models for testing the functional subunit hypothesis [14] , in which information from subgroups of canals in different clusters that have similar directional projections may be integrated to maximize direction computations . Future comparative work that assesses electrosensory-processing mechanisms across taxa will also provide important clues on the evolution and diversity of ampullary arrays seen in elasmobranch fishes .
|
The electric sense appears in a variety of animals , from the shark to the platypus , and it facilitates short-range prey detection where environments limit sight . Typically , hundreds or thousands of sensors work in concert . In skates , rays , and sharks , each electrosensor includes a small , innervated bulb , with a thin , gel-filled canal leading to a surface pore . While experiments have mapped single electrosensor activity , the mechanisms that integrate neural input from multiple electrosensors are still largely unknown . Here , we model the response of a precisely mapped subset of electrosensors responding in concert for a skate moving near stationary prey . Just as two ears help locate sound via time and intensity differences , we ask how a bilateral electrosensor array can contribute to electrical scene analysis . Our results show that the sensor array provides rich data for precise prey location , tuned by the morphology to render certain events , like the point of closest approach , “loud and clear . ” This proof of principle makes a significant step in understanding the electric sense processing , and we recommend future experiments to compare and assess functions for the diversity of arrays found in other sharks and rays .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"Methods",
"Results",
"Discussion"
] |
[
"vertebrates",
"physics",
"eukaryotes",
"computational",
"biology",
"evolutionary",
"biology",
"biophysics",
"neuroscience",
"animals"
] |
2007
|
From Morphology to Neural Information: The Electric Sense of the Skate
|
Systemic inflammation and sequestration of parasitized erythrocytes are central processes in the pathophysiology of severe Plasmodium falciparum childhood malaria . However , it is still not understood why some children are more at risks to develop malaria complications than others . To identify human proteins in plasma related to childhood malaria syndromes , multiplex antibody suspension bead arrays were employed . Out of the 1 , 015 proteins analyzed in plasma from more than 700 children , 41 differed between malaria infected children and community controls , whereas 13 discriminated uncomplicated malaria from severe malaria syndromes . Markers of oxidative stress were found related to severe malaria anemia while markers of endothelial activation , platelet adhesion and muscular damage were identified in relation to children with cerebral malaria . These findings suggest the presence of generalized vascular inflammation , vascular wall modulations , activation of endothelium and unbalanced glucose metabolism in severe malaria . The increased levels of specific muscle proteins in plasma implicate potential muscle damage and microvasculature lesions during the course of cerebral malaria .
Human malaria is a life-threatening disease causing an estimated 655 , 000 deaths in 2010 [1] . Although the mortality rates have decreased during the last decade , deaths in Africa due to childhood malaria are still elevated with Plasmodium falciparum attributable to a third of the childhood deaths accounted in Nigeria [2] . Complications may develop abruptly and may be fatal . Although the most common severe syndromes , i . e . cerebral malaria , severe malaria anemia or respiratory distress , have been widely investigated , many aspects of their pathogenesis remain elusive . Furthermore , it is yet unknown what predetermines which children are at risk of developing complications . Parasitized red blood cells ( pRBC ) are specifically withdrawn from the peripheral circulation during severe malaria infection through binding to and activation of vascular endothelial cells , erythrocytes , leukocytes and platelets , which may obstruct the blood flow . It is known that increased micro-vascular congestion accompanies coma in cerebral malaria and the depth of coma is correlated to the extent of the sequestration of the pRBC [3] . Plasma proteins are involved in the adhesive events of pRBC [4]–[6] and an electron-dense fibrillar material composed of immunoglobulins , fibrinogen and albumin was found deposited on vessels at autopsy and also involved in mediating adhesion of pRBC [6] , [7] . In the case of severe malaria anemia , increased destruction of pRBC and non-pRBC , splenic sequestration of RBC and dyserythropoiesis contribute to anemia and free-heme-induced oxidative stress [8] . In addition , there is compelling evidence that prolonged pro-inflammatory response and an inadequate anti-inflammatory response might contribute to persistent anemia [8] . However , due to different observations in different cohort studies in cerebral malaria [9] , [10] , the role of circulatory inflammatory cytokines in malaria physiopathology remains elusive . Despite the current plethora of new technologies available for analyzing and profiling proteins in body fluids , the yield of validated biomarker molecules remains low [11] . Previous studies have tried to detect protein signatures specific to malaria disease but the wide dynamic range of plasma proteins has been a limiting factor [12]–[14] . The technical issues have not yet allowed for comprehensive studies of circulating proteins since this proteome has many members and their identification is laborious . Here , we have overcome some of these challenges by applying a single-antibody microsphere-based multiplex assay utilizing more than 1 , 000 antibodies from the Human Protein Atlas ( HPA ) project [15] . For the generation of antibody suspension bead arrays , HPA antibodies are coupled to color-coded magnetic microspheres and combined to create a 384-plex-bead array . After combination with biotinylated samples , bead identity and captured plasma proteins are detected using a flow cytometric analyzer . Previously , we have shown that limits of detection reach into lower ng/ml or higher pg/ml ranges while consuming less than 1 µl of plasma sample [16] for the profiling of 384 proteins [17] . The potential to screen hundreds of analytes in hundreds of patient samples simultaneously in one parallel assay allows for an effective exploration of potential candidates in a time-efficient manner . Here , we have employed an affinity proteomics approach to study 1 , 015 human proteins with 1 , 132 antibodies in the peripheral blood of children from Nigeria with different syndromes from severe to uncomplicated malaria as well as non-diseased parasite-negative children from the same community . Protein markers of inflammation , endothelial activation , platelet adhesion , vaso-modulation , glucose metabolism , oxidative stress and muscle-damage were found in relation to severe malaria . In particular , three muscle derived proteins , creatine kinase , carbonic anhydrase III and myoglobin , were detected in the plasma of children with cerebral malaria suggesting deep lesions into their micro-vasculature including the vascular smooth-muscle cell-layer and extra-vascular striated muscles cells , concurrent with excessive sequestration of pRBC throughout the human body . As a consequence we have identified protein signatures that allowed the distinction between the different presentations of the disease with AUCs up to 0 . 90 in plasma samples from a verification cohort . The data contributes to a deeper understanding of the complex mechanisms that lead to severe disease and may serve as a basis for the development of novel diagnostic strategies that would enable the prediction of the severity of malaria .
First , a set of patients was selected and carefully matched and balanced to contain 356 childhood malaria patients and controls recruited during 2006 to 2011 in Ibadan , Nigeria . This set , denoted as “discovery cohort” in this study , included samples from patients suffering from uncomplicated malaria ( UM , n = 89 ) , severe malaria anemia ( SMA , n = 89 ) and cerebral malaria ( CM , n = 89 ) and from parasite-negative community controls ( CC , n = 89; Table 1a ) . To confirm the validity of the protein signatures identified , a second set of plasma samples was analyzed including 332 independent individuals recruited from the 2009 to 2012 period . This set consisted of 178 UM , 58 SMA , 36 CM , 60 CC and 31 disease controls ( DC ) and was here denoted as “verification cohort” ( Table 1b ) . The discovery and verification cohorts had 24 patient samples in common to assess the technical quality of the data in independent analyses ( Methods S1 and Fig . S1C in Text S1 ) . Affinity proteomic arrays require that the analytes of interest are defined prior to analysis . For the presented approach , an inclusive and generous target selection strategy was employed by thorough literature mining of major processes previously associated to various aspects of malaria . Antibodies towards these targets were obtained based upon availability within the Human Protein Atlas ( HPA ) . In total , a list of 372 antibodies targeting 304 different human proteins was compiled and denoted as ‘targeted array’ . This list comprised primarily plasma proteins associated with acute inflammation ( Fig . S9 in Text S1 ) , iron metabolism , oxidative stress , endothelial activation , coagulation , complement activation , angiogenesis , hematopoiesis and brain injury . Two additional sets of 380 antibodies were used to profile all samples in the discovery cohort and are denoted as ‘random arrays’ . The two additional sets of antibodies were randomly chosen from the routine antibody production within the Human Protein Atlas ( HPA ) where more than 40 , 000 antigen purified and protein microarray validated antibodies have been generated . The antibodies fulfill the quality criteria of having a concentration that is higher than 50 µg/ml and that the specificity is validated on planar protein arrays . These criteria were also true for the antibodies on the targeted array . These two ‘random arrays’ , covering 760 antibodies were corresponding to 711 unique proteins , generating the total sum of 1 , 132 antibodies representing 1 , 015 unique proteins . The technical reproducibility of the assays was assessed in independent experiments ( shown for ‘targeted array’ in Fig . S1A , S1B and Methods S1 in Text S1 ) . To provide further evidence for the validity of the protein profiles , antibodies raised towards different regions of a common protein were compared ( Fig . S7 and Table S3 in Text S1 ) . Principal component analysis ( PCA ) was chosen as a first tool to visualize a separation between control and disease groups globally . The first three principal components indicated a larger spread separation between the samples analyzed by proteins represented on the ‘targeted array’ compared to the ‘random arrays’ ( Fig . S2A in Text S1 ) . A non-parametric test was used to identify proteins that were significantly different ( p<0 . 001 ) between any of the groups , i . e . UM , SMA and CM as well as healthy parasite-negative CC . From the ‘targeted array’ , 29 human proteins were identified in plasma as discriminatory protein targets ( Table 2 , Fig . 2 , Figs . S2B , S3 in Text S1 ) . By using self-organizing tree algorithm ( SOTA ) cluster analysis on candidate proteins from the ‘targeted array’ , four clusters grouped proteins according to disease severity ( Fig . 2A ) . A heatmap built on SOTA clusters demonstrated that all but three proteins were elevated in malaria disease groups compared to controls ( Fig . 2B , Fig . S2C in Text S1 ) . Proteins that differed in abundance in plasma between the three syndromes were further visualized in volcano plots as two-group comparisons ( Fig . S5 in Text S1 ) . A total of 17 proteins were found to discriminate malaria disease from controls . The set of proteins that was denoted ‘Malaria decreased’ ( Fig . 2 ) contained fibulin-1 ( FBLN1 ) and RANTES ( CCL5 ) , both involved in endothelial cell death , whilst glutathione peroxidase ( GPX1 ) protects from oxidative stress . Decreased levels of these proteins may therefore imply endothelial survival and oxidative stress . The 14 proteins denoted ‘Malaria increased’ were primarily inflammatory acute phase proteins and increased with severity of the disease . Notably , the most significant candidates differing between controls and malaria groups were major inflammatory components von Willebrand factor ( VWF ) and C-reactive protein ( CRP ) . Additionally , lipopolysaccharide binding protein ( LBP ) , serpin peptidase inhibitor , member 3 ( SERPINA3 ) and orosomucoid ( ORM ) are known members of the malaria-induced acute inflammatory response ( Fig . S9 in Text S1 ) , while neurofilament-M ( NF-M ) is an intracellular protein not secreted under healthy conditions , but its increased levels may point towards increased tissue lysis . A remarkable group of markers were proteins found to be common for ‘severe malaria’ ( Fig . 2 ) , defined here as common for CM and SMA . Many of these eleven proteins are general inflammatory proteins . Some have specific roles , such as endothelial cell activation ( e . g . VCAM1 ) and coagulation ( factor X ) . In addition , molecules linked to glucose metabolism , as well as anemia ( EMBP4 . 1-L2 ) , were found in this group of proteins . Interestingly , two proteins were increased exclusively in plasma of children with CM: carbonic anhydrase III ( CA3 ) , which has been reported to be strictly tissue specific and present at high levels in skeletal muscle and creatine kinase ( CK ) , which is involved in energy homeostasis and specific to brain , muscle and heart tissues . These two candidates are associated to muscle tissue and indicate a linkage between muscle damage and CM . Overall for the ‘discovery cohort’ , pathway analysis using the Ingenuity Pathway analysis software ( Ingenuity Systems ) outlined acute phase signaling as the most significant pathway affected in malaria disease ( Fig . S9 in Text S1 ) . From the analysis using the ‘random array’ , 12 proteins revealed significant differences using a non-parametric test ( p<0 . 001 , Table 3 , Fig . S4 in Text S1 ) . In summary , an additional major inflammatory protein , CCAAT/Enhancer Binding protein-alpha ( CEBPA ) and three intracellular proteins were found with increased levels in plasma from all malaria-positive groups . In the severe malaria groups , intracellular proteins ( TIPIN , MSRB1 ) , components of glucose metabolism ( ADSSL1 , DNPEP1 ) and an inflammatory protein ( DAPK1 ) had increased levels , further confirming observations made above with the ‘targeted array’ . Moreover , myosin 15A ( MYO15A ) was increased in the CM group similarly to the ‘cerebral malaria’ cluster components of the ‘targeted array’ . In summary , both targeted and blinded selections revealed differential protein profiles in plasma that highlighted several processes in response to the infection and encouraged the subsequent analysis for a classification of malaria disease subtypes . The list of candidate proteins discriminating the three disease groups contained a number of interesting targets ( Table 2 , Table S1 in Text S1 ) . However , a multi-protein signature consisting of an optimal combination of these or additional markers generally achieves a more efficient and robust discrimination . Therefore , we chose a L1 penalized regression model [18] , [19] to identify panels of proteins that would distinguish between each sub-group comparison , as presented below . The best predictor protein to differentiate SMA from UM was insulin-like growth factor binding protein 1 ( IGFBP1 ) with an area under the ROC curve ( AUC ) of 0 . 84 alone ( Fig . 3A , Table S2 in Text S1 ) . The best multi-protein combination was a 3-protein signature consisting of IGFBP1 with von Willebrand factor ( VWF ) and hemoglobin α-subunit ( HBA2 , HBA1 ) , which resulted in a slightly improved AUC of 0 . 87 . Next , Carbonic anhydrase III ( CA3 ) alone discriminated CM from UM with a high AUC of 0 . 90 ( Table S2 in Text S1 ) . Variable selection refinement showed that a panel of 4 proteins , including angiotensinogen ( AGT ) , FBLN1 and 2 , 3-bisphosphoglycerate mutase ( BPGM ) , resulted in a yet improved performance of 0 . 94 ( Fig . 3B ) . The optimal signature identified for the discrimination of CM and UM was a 23-protein signature with an AUC of 0 . 98 . The two best classifier proteins for the comparison of the two severe malaria groups , CM and SMA , were CA3 and CK . The combination of the 2 proteins resulted in an AUC of 0 . 84 ( Fig . 3C ) . Moreover , the most optimal classifier combination comprised 9 proteins resulting in an AUC of 0 . 91 . In summary , protein targets in the multivariate signatures overlapped largely with the targets found in the univariate analysis in the two-group comparisons ( Fig . S5 in Text S1 ) . As expected , ranking of the proteins differed , because multivariate models aim at the identification of the best combination of proteins to maximize the discriminative power . This resulted in protein signatures that contained not only highly significant proteins found with univariate analysis but proteins that carry important discriminatory information when combined with others . Both the single protein classifiers and multi-protein panels were validated with a new and independent set of 363 samples from the same hospital . Using the classifiers determined in the multivariate signatures containing the full list proteins , the AUC for the UM versus SMA ( 3 proteins ) , UM versus CM ( 23 proteins ) and SMA versus CM ( 8 proteins ) comparisons were 0 . 72 , 0 . 90 and , 0 . 86 , respectively ( Fig . 3D ) . To further verify the results technically , additional antibodies were used targeting other regions of the same protein . For example , von Willebrand factor ( VWF ) was repeated using the same antibody ( HPA00282 ) and an additional antibody ( HPA001815 ) , with both showing comparable results in the different disease groups in the verification cohort and the discovery cohort ( Fig . S7 in Text S1 ) , demonstrating reproducibility of the presented results . Additionally , a small cohort of disease control ( DC ) samples ( Table 1b ) from patients suffering from coma or meningitis was also profiled with the same protein panel analyzed in the verification cohort . Changes in levels due to inflammation , as shown with CEBPA , CRP and CCL5 protein levels , were more exacerbated in the CM than the DC samples ( Fig . S10 in Text S1 ) suggesting a stronger inflammatory response in malaria-infected patients than in other diseases . CA3 was identified as the top candidate to differentiate CM from the other two malaria syndromes . In both the discovery and the verification analysis , two antibodies with distinct specificities to CA3 showed concordant performance ( Fig . S7 in Text S1 ) . An additional antibody was acquired against the same target protein and further confirmed the results . Immunohistochemical staining of healthy human tissue confirmed the muscle specificity of both antibodies ( Fig . S8 in Text S1 ) . Similarly , antibodies against creatine kinase ( CK ) revealed increased levels in CM in the discovery phase . The CK antibody used in the discovery phase was raised against a region of the brain-specific form ( CKB ) that is shared with the muscle-specific ( CKM ) isoform ( Table S3 in Text S1 ) . To further evaluate which isoform is detected , an HPA antibody directed towards the muscle-specific isoform ( CKM ) as well as a commercially available antibody against CKM were tested in the verification phase . The results from these CKM antibodies were similar to the trends with the CKB/M antibody used in the discovery phase ( Fig . S7 in Text S1 ) . Furthermore , immunohistochemical staining showed that the CKB/M antibody recognized skeletal and cardiac muscle as well as cerebral tissues ( Fig . S8 in Text S1 ) . Finally , myoglobin ( MB ) , a cardiac and skeletal muscle protein , was also part of the discriminatory profile between CM and UM ( Fig . 3b ) and also had higher plasma levels in the CM group compared to all other groups in the verification sample set . Using the small additional DC cohort , CA3 and CK , previously identified as related to CM syndrome , had levels slightly lower in the DC group compared to CM but were not significantly different ( Fig . S10 in Text S1 ) suggesting that muscle damage might not be specific to cerebral malaria but probably linked to coma . Further studies with a larger cohort of DC comatose group of children with pathologies other than malaria are required to verify these findings .
We have here investigated the levels of human proteins circulating in plasma of children with different forms of uncomplicated or severe malaria and compared the levels with those of parasite-negative community controls . The study comprises a total of 709 plasma samples including 515 from malaria-infected children . Amongst the 1 , 015 host proteins studied , 41 were identified as candidates discriminating between healthy community controls and malaria patients . Protein markers of oxidative stress were found elevated in anemic individuals while markers of endothelial activation , platelet adhesion and muscle- and tissue damage were found linked to cerebral malaria . Taken together , this suggests the presence of a generalized vascular inflammation , an unbalanced glucose metabolism and deep lesions into the micro-vasculature . The chosen bead array technology enabled the generation of protein profiles in unfractionated and biotinylated plasma samples by using combinations of large sets of antibodies as demonstrated by the use of both carefully pre-selected and blindly chosen antibodies . Most previous studies have focused on markers of fatalities in those already with severe malaria [20]–[22] . In one recent analysis [14] , discrimination of different malaria syndromes from each other was suggested possible but only when using extensive protein panels with up to 50 proteins ( AUC of 0 . 7–0 . 8 ) . In contrast , we show herein that , employing small panels of proteins , it is possible to build models that predict with an AUC higher than 0 . 90 which children have severe malaria complications . We also demonstrate a discriminatory signature that reaches superior accuracy for UM vs . SMA with IGFBP1 alone and for UM vs . CM using only four proteins . Higher plasma levels of muscle-derived proteins were found in children with cerebral malaria only including carbonic anhydrase III and creatine kinase , suggesting that smooth muscle-cells of the microvasculature may be injured . The excessive sequestration of pRBC seen in cerebral vessels , the level of which has also been found to correlate with coma [3] is probably one of the reasons for the injury of the muscle cells . This is in concordance with previous histo-pathological , studies where subjects who succumbed to cerebral malaria showed vascular- and microvascular lesions complicated by ring-hemorrhages [3] . The presence of myoglobin in the plasma of the patients with CM , a marker of cardiac- and striated muscles only , also indicates that the vasculature and the muscles outside of the brain are severely affected by sequestration as seen for example in muscle biopsies of Thai adults [23] . Further , a recent study showed that blood flow obstruction might be exacerbated by increased skeletal muscle oxygen consumption in severe malaria [24] , contributing to hypoxic and hyperlactemic conditions in the microvasculature . Lack of oxygen in muscle cells accompanied with hypoglycemia , lactate overproduction , oxidative stress and inflammation are typical consequences of muscle damage [25] , which could further contribute to vascular injuries and subsequent muscle cell death with the release of muscle-specific proteins into the blood circulation . Whether creatine kinase , myoglobin and carbonic anhydrase III release in the plasma exacerbate these deleterious events is not known at this stage and deserves further investigation . For example , increased plasma carbonic anhydrase activity could contribute to the impairment of the acid-base and excess myoglobin in blood circulation could lead to kidney failure if filtrated by kidneys . An additional small cohort of malaria-negative children , with other illnesses involving coma , suggested that coma could either be a cause or a consequence of muscle damage observed in cerebral malaria , similarly to other comatose diseases . Further studies , involving larger cohorts of non-malaria comatose children , will be required to verify this hypothesis . In our study , predicting which samples were from patients diagnosed with cerebral malaria was very accurate in both cohorts due to the discovery of the presence of the muscle-proteins only in the plasma of the children with cerebral malaria . Previous studies have successfully used endothelial cell activation markers to predict severe malaria , notably using angiopoietin-1 and 2 [20] , [21] , [26] , but their specificity to cerebral malaria , as compared to other severe complications , remains unclear . Here we propose that markers of muscle damage accompanied by markers of endothelial cell activation/platelet adhesion in the plasma ( Fig . 4 , orange bars ) are specific to cerebral malaria pathogenesis and distinct to severe malaria anemia . Our data therefore indicate that children with uncomplicated malaria that develop cerebral malaria are likely to have vascular lesions and muscle damage , which can be readily monitored in plasma . Most of the plasma proteins showing differences in between the malaria patients and the community controls were components of the inflammatory response ( Fig . 4 , blue bars ) . Further , multiple novel inflammatory components were identified , including MMP2 , CSF1 , and IL7 ( table 2 ) , demonstrating the presence of a more generalized vascular inflammation in patients with malaria infections . The reliability of the present study was furthermore confirmed by the fact that a number of protein measured herein include acute phase- and inflammatory proteins previously documented to be present in the plasma of malaria infected individuals [27] , [28] ( Table 2 ) . Proteins related to different aspects of the glucose metabolism , including insulin-glucagon modulators and glycolytic enzymes , were also elevated in both SMA and CM patients compared to those with UM ( Fig . 4 , purple bars ) . Further investigation of their potential role in the induction of hypoglycemia , a hallmark of severe malaria associated with fatality [29] , [30] might refine knowledge on the human response mechanisms to malaria infection . Most proteins elevated in plasma from SMA compared to UM were also elevated in CM but to a lower extent . Yet , IGFBP1 and HBA were part of the protein signature for comparing SMA with UM , and had highest levels in patients suffering with SMA . IGFBP1 could be further expressed due to high levels of reactive species [31] , and free-hemoglobin release in blood circulation could be a trigger of free-heme induced oxidative stress , particularly if not properly scavenged by the haptoglobin-hemopexin system [32] . It is noteworthy that predictive protein signatures for SMA have in previous studies mainly included inflammatory cytokines [22] , [33] , [34] . For example , we recently showed in the same Nigerian population , that pro-inflammatory cytokines were more pronounced in SMA than in CM [9] , a finding supported by the fact that pro-inflammatory TNF-alpha has a role in anemia establishment [8] . Due to the sensitivity of the assay , the levels of most of the cytokines tested in the present study were too low to be detected . Consequently , we hypothesize that including oxidative stress-related proteins as well as pro-inflammatory cytokines in future studies in the protein signature could potentially assist to further improve SMA distinction from other malaria complications . In summary , a high-throughput antibody-based protein profiling method and large-scale discovery and verification cohorts , revealed muscle-specific proteins in plasma as potential indicators of cerebral malaria . Our study could therefore provide key elements towards the discovery of distinct mechanisms in the human response to malaria infection between the two most fatal syndromes of childhood malaria .
Parents or guardians of study participants gave informed written consent . This research was approved by the internationally accredited joint ethics committee of the College of Medicine of the University of Ibadan and the University College Hospital , Ibadan . All study participants with illness were recruited under the auspices of the Childhood Malaria Research Group ( CMRG ) at the University College Hospital ( UCH ) in the city of Ibadan , Nigeria . Malaria-negative community control ( CC ) children were recruited from local vaccination clinics as well as during school visits across several Ibadan districts . This case-control study was divided in a discovery cohort that contains those patients recruited during 2006 to 2011 and a verification cohort made up of those recruited in the 2009 to 2012 period . Children were aged from 6 months to 13 years and were screened for parasite detection by microscopy following Giemsa staining of thick and thin blood films as performed routinely at UCH . Clinical definitions used were as defined by the WHO criteria for severe P . falciparum malaria [1] . Uncomplicated malaria ( UM ) cases were defined as febrile children with P . falciparum parasitaemia and PCV ( Packed Cell Volume ) greater than 20% who did not require hospital admission . Severe malarial anemia ( SMA ) cases were defined as conscious children with PCVs less than 16% in the presence of P . falciparum parasitaemia . Cerebral malaria ( CM ) cases were defined as children in unrousable coma for at least one hour in the presence of asexual P . falciparum parasitaemia with normal cerebrospinal fluid and PCV greater than 20% . Community controls were children that did not show any obvious symptoms of illness and seemed healthy . They were screened for parasite presence and were only included in the study if the Giemsa staining of both thick and thin blood films were negative for Plasmodium parasites . They were selected to match age and sex with malaria-infected patients . The clinical data was compiled for each patient and samples were collected as previously described [9] , [14] ( see Methods S1 in Text S1 ) . Antibodies were selected and acquired from the huge antibody collection within the Human Protein Atlas ( HPA , www . proteinatlas . org ) consisting of more than 40 , 000 antigen purified and protein microarray validated antibodies . The selection of antibodies was carried out using two different strategies . Using a ‘targeted’ approach , 380 antibodies were selected against 304 protein targets according to a generous and inclusive literature mining of malaria pathogenesis . The final set was defined by availability and fulfillment of technical validations , such as having a concentration that is higher than 50 µg/ml and that the specificity is validated on planar protein arrays . The two additional sets of 380 antibodies were randomly chosen from the routine antibody production within HPA . These 760 antibodies were directed to 711 unique proteins and fulfilled the same criteria as for the antibodies on the targeted array . For the generation of antibodies suspension beads , antibodies were diluted using a liquid handling system ( EVO150 , Tecan ) and coupled to carboxylated magnetic microspheres ( MagPlex , Luminex Corporation ) as previously described [35] . Briefly , carboxylated beads were activated with 1-ethyl-3- ( 3-dimethylaminopropyl ) carbodiimide ( EDC ) and N-hydroxysulfosuccinimide ( Sulfo-NHS , Thermo Scientific ) and incubated with 1 . 6 µg antibody in a multi-well microtiter plate for 2 h . After the coupling reaction , beads were stored at 4°C in a protein containing buffer ( Blocking Reagent for ELISA , Roche Applied Science ) supplemented with ProClin ( Sigma-Aldrich ) . Before incubation with samples , the different bead identities were combined to create a 384-plex-bead array . Antibody coupling was confirmed with R-phycoerythrin ( PE ) -conjugated donkey anti-rabbit IgG antibody ( Jackson ImmunoResearch ) . The data from single antibody and direct sample labeling assays was judged by technical replication of the experiment , profile concordance of several antibodies raised towards a common target protein , and biological replication of the analysis in new , independent samples . Biotinylation of plasma samples was performed as previously described [36] ( refer to Methods S1 in Text S1 ) . Biotinylated samples were then diluted in PBS containing 0 . 5% ( w/v ) polyvinylalcohol , 0 . 8% ( w/v ) polyvinylpyrrolidone , 0 . 1% casein ( all from Sigma ) supplemented with 0 . 5 mg/ml rabbit IgG ( Bethyl ) using a liquid handler ( SELMA , CyBio ) . Before incubation with bead arrays , samples were heat-treated in a thermocycler for 30 min at 56°C for epitope retrieval . After incubation of samples with beads for 14 h , beads were washed with PBS-T ( pH 7 . 4 , 0 . 05% Tween20 ) on a plate washer ( EL406 , Biotek ) and incubated with 0 . 4% paraformaldehyde for 10 min . Subsequently , beads were incubated with 0 . 5 µg/ml R-Phycoerythrin labeled streptavidin ( Invitrogen ) for 20 min and washed with PBS-T . Bead identities and median fluorescence intensity of R-Phycoerythrin were analyzed simultaneously using a FlexMAP 3D system ( Luminex Corp . ) . Data analysis was performed using R language for statistical computing [37] , [38] . The data from the ‘targeted array’ and the ‘random array’ was normalized using probabilistic quotient normalization ( PQN ) as described before [39] . The non-parametric Kruskal-Wallis test was applied to identify proteins that are different among the different malaria disease groups ( CC , UM , SMA and CM ) . For pairwise comparison of the different malaria disease groups , a Wilcoxon rank sum test was applied ( with continuity correction ) . Bonferroni method was used to control the family-wise error rate . For cluster analysis of the protein profiles self-organizing tree algorithm ( SOTA ) was applied after the medians per protein and subgroup were centered and scaled using the R function scale [40] ( Methods S1 in Text S1 ) . The data from the verification cohort was normalized using a linear mixed modelwhere: y = log2 ( intensity ) , Plate: biotinylation plates , Index: order of assay , and bi: random effect of the ith target . For the identification of protein signatures a logistic regression model was used . We used L1 penalization proposed by Tibshirani [19] , also known as Lasso , which performs parameter estimation and variable selection at the same time . The penalization involves a penalize parameter ( λ ) which is chosen through a cross-validation procedure . Thereby , the dataset was randomly divided into subsets . The first subset ( K ) was designated as the test dataset , while the model was fitted to the remaining training dataset ( subset K-1 ) . This procedure was repeated k times for each subset ( see Methods S1 in Text S1 for details ) . For a first verification of the identified multi-protein signatures , the parameter estimates from the first dataset were used to obtain the prediction based on the second replicate data of the discovery cohort ( Fig . S6 in Text S1 ) .
|
Why do some malaria-infected children develop severe and lethal forms of the disease , while others only have mild forms ? In order to try to find potential answers or clues to this question , we have here analyzed more than 1 , 000 different human proteins in the blood of more than 500 malaria-infected children from Ibadan in Nigeria , a holoendemic malaria region . We identified several proteins that were present at higher levels in the blood from the children that developed severe malaria in comparison to those that did not . Some of the most interesting identified proteins were muscle specific proteins , which indicate that damaged muscles could be a discriminatory pathologic event in cerebral malaria compared to other malaria cases . These findings will hopefully lead to an increased understanding of the disease and may contribute to the development of clinical algorithms that could predict which children are more at risks to severe malaria . This in turn will be of high value in the management of these children in already overloaded tertiary-care health facilities in urban large densely-populated sub-Saharan cities with holoendemic malaria such as in the case of Ibadan and Lagos .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"biochemistry",
"infectious",
"diseases",
"medicine",
"and",
"health",
"sciences",
"diagnostic",
"medicine",
"pathology",
"and",
"laboratory",
"medicine",
"biology",
"and",
"life",
"sciences",
"proteomics",
"malaria",
"biomarkers",
"parasitic",
"diseases"
] |
2014
|
Affinity Proteomics Reveals Elevated Muscle Proteins in Plasma of Children with Cerebral Malaria
|
Visceral leishmaniasis ( VL ) in Brazil is a neglected , vector-borne , tropical parasitic disease that is responsible for several thousand human deaths every year . The transmission route involves sand flies becoming infected after feeding on infected reservoir host , mainly dogs , and then transmitting the Leishmania infantum parasites while feeding on humans . A major component of the VL control effort is the identification and euthanasia of infected dogs to remove them as a source of infection . A rapid , non-invasive , point-of-care device able to differentiate between the odours of infected and uninfected dogs may contribute towards the accurate diagnosis of canine VL . We analysed the headspace volatile chemicals from the hair of two groups of dogs collected in 2017 and 2018 using a bench-top eNose volatile organic chemical analyser . The dogs were categorised as infected or uninfected by PCR analysis of blood samples taken by venepuncture and the number of parasites per ml of blood was calculated for each dog by qPCR analysis . We demonstrated using a robust clustering analysis that the eNose data could be discriminated into infected and uninfected categories with specificity >94% and sensitivity >97% . The eNose device and data analysis were sufficiently sensitive to be able to identify infected dogs even when the Leishmania population in the circulating blood was very low . The study illustrates the potential of the eNose to rapidly and accurately identify dogs infected with Le . infantum . Future improvements to eNose analyser sensor sensitivity , sampling methodology and portability suggest that this approach could significantly improve the diagnosis of VL infected dogs in Brazil with additional potential for effective diagnosis of VL in humans as well as for the diagnosis of other parasitic diseases .
Visceral leishmaniasis ( VL ) is a neglected tropical disease caused by protist parasites belonging to the genus Leishmania . Globally over 350 million people are at risk of infection with an estimated 200–400 thousand cases annually and an estimated 10% fatality rate . Ninety percent of all reported VL cases occur in only six countries including Brazil[1 , 2] . In Brazil , transmission of Leishmania ( Leishmania ) infantum ( Kinetoplastida: Trypanosomatidae ) occurs between domestic dogs Canis familiaris ( Carnivora: Canidae ) ( the reservoir host ) and from dogs to humans when an infected female sand fly vector Lutzomyia longipalpis ( Diptera: Psychodidae ) takes a blood meal . Despite substantial efforts by the Brazilian Ministry of Health ( MoH ) the burden of VL in Brazil more than doubled between 1990 and 2016[3] . The increase is probably due to the spread of the vector into urban areas as a result of human migration into cities[4] and the expansion of the range of the vector into new areas because of environmental degradation[5–7] . Given the spread of the disease and increase in cases it is also likely that current VL control measures are inadequate[8] . The control of VL in Brazil has three main components . Insecticides are applied in houses and animal sheds to lower the vector population density and reduce vector-human contact . Secondly , diagnosis and treatment of human cases to prevent severe forms of the disease and death . Finally , the identification and euthanasia of seropositive canine cases to decrease the sources of infection for the vector[9 , 10] . Modelling predicts that the dog-culling program in Brazil should be effective in areas of low , medium but not high Leishmania transmission[11] . However , the practice is controversial and despite the euthanasia of thousands of canines with suspected and confirmed infection each year the program has been unsuccessful[10 , 12] . There are a number of possible explanations for this situation A ) . Shortage of qualified professionals caused by financial constraints leading to delays in collections , performance of routine diagnostic tests and subsequent removal of seropositive dogs . B ) . Failure to identify and remove the high proportion of asymptomatic animals C ) . Refusal of dog owners to comply with surveillance measures . D ) . The high rate of dog replacement with young immunologically naïve dogs . E ) . Lack of an accurate point-of-care diagnostic test[13] . Identification of dogs infected with canine VL ( CVL ) follows a two stage serodiagnostic protocol recommended by the Brazilian MoH . Initial screening using the Dual-Path Platform ( DPP CVL ) immunochromatography diagnostic test is followed by a laboratory-based ELISA ( EIE CVL ) confirmatory test . Overall the 2-step protocol was reported to have a 73% sensitivity and 98% specificity however the relatively low sensitivity indicates the maintenance of false-negative dogs in endemic areas which represents a public health concern[14] . The DPP CVL test has also been assessed several times since it was introduced and most recently it has been shown that overall it has 86% sensitivity and 94% specificity[14] or 89% sensitivity and 70% specificity[15] . The concept of volatile organic compounds ( VOCs ) as diagnostic aids to signal a disease is well established and since antiquity , many physicians have used odours associated with disease to help diagnose their patients[16] . Modern analytical techniques such as single ion flow tube mass spectrometry ( SIFT-MS ) and chemi-resistive sensors have taken the concept to the point of widespread clinical application . Volatile markers from human breath can be used to identify a variety of disease states e . g . inflammatory bowel disease , chronic liver disease , diabetes , Pseudomonas aeruginosa infection and adenocarcinomas[17 , 18] . A recent study has shown that the use of VOCs is sufficiently robust to discriminate between 14 cancerous and other disease states[19] . Parasite infections of humans and animals also alter the odour of the host animal . The odours of golden hamsters infected with Le . infantum are more attractive to female sand flies than the odours of uninfected hamsters[20 , 21] . The odour obtained from the hair of dogs infected with Le . infantum in Brazil was found to be significantly different to the odour of uninfected dogs . These odour differences which were detected by coupled gas chromatography-mass spectrometry ( GC/MS ) and multivariate statistical analysis indicated the increased presence of a small number of primarily low molecular weight aldehydes ( octanal , nonanal ) , alkanes ( undecane , heptadecane ) and 2-ethylhexyl-salicylate[22 , 23] . More recently , odours were also implicated in children infected with the infectious gametocyte stage of the malaria parasite Plasmodium falciparum were found to be more attractive to the mosquito vector Anopheles gambiae[24] . This phenomenon occurred even when the gametocytemia was very low and was associated with changes in aldehyde concentration of the foot odours of the infected children[25 , 26] . GC/MS analysis is a useful research tool but its use as a widely available diagnostic tool is unrealistic because of significant costs associated with the infrastructure and personnel costs . An alternative means of detecting the odour change associated with parasitaemia is required that would fulfil the majority of the World Health Organisation ASSURED criteria; affordable , sensitive , specific , user-friendly , rapid and robust , equipment free and deliverable to end-users[27] for all new point-of-care diagnostics tools . VOC analysers ( eNoses ) may fulfil WHO criteria , they can detect differences in the odours from sputum of tuberculosis ( TB ) infected and TB uninfected patients with sensitivity , specificity and accuracy of around 70%[28] . The aim of the present study was to determine if the odour of dogs naturally infected with Le . infantum could be detected with high sensitivity and specificity using a commercially available VOC analyser .
Governador Valadares ( 18°51′S , 41°56′W ) ( Minas Gerais State , Brazil ) , located in the valley of the Rio Doce 320 km northeast of Belo Horizonte is a city of approximately 280 , 000 people . The climate is temperate , characterised by dry winters and hot , wet summers[29] . Studies in Governador Valadares in 2013 found that an average of 30% of dogs from 16 , 529 samples taken from 35 urban and rural districts were seropositive for canine visceral leishmaniasis ( CVL ) [30] . From 2008 until 2017 , 194 human VL cases were recorded in Governador Valadares with a fatality rate of 15 . 5%[31] . Dog blood and hair samples were taken from dogs that were also microchipped with the informed consent of their owners . Ethical approval was obtained from the Comissão de Ética no Uso de Animais ( CEUA ) , Instituto Oswaldo Cruz ( licence L-027/2017 ) in Brazil and Lancaster University Animal Welfare and Ethics Review Board ( AWERB ) in the UK . The CEUA approval complies with the provisions of Brazilian Law 11794/08 , which provides for the scientific use of animals , including the principles of Brazilian Society of Science in Laboratory Animals ( SBCAL ) . The AWERB approval complies with the UK Home Office guidelines of the Animals in Science Regulation Unit ( ASRU ) and in compliance with the Animals ( Scientific Procedures ) Act ( ASPA ) 1986 ( amended 2012 ) regulations and was consistent with UK Animal Welfare Act 2006 . A 2 year cohort study in the Altinópolos district of Governador Valadares was initiated in August 2017 by initial recruitment and sampling of 185 dogs . The area was chosen because of the high prevalence of CVL ( average incidence 33 . 8% ) [30] and the large population of household-owned dogs ( ca . 2000 ) ( Centro de Controle de Zoonoses ( CCZ ) survey ) located there . The dogs were microchipped to aid their identification . Inclusion criteria: dogs aged ≥ 3 months , dogs without previous clinical assessment or laboratorial diagnosis for CVL . Exclusion criteria: pregnant/lactating bitches; aggressive dogs; stray dogs . In April 2018 149 dogs were sampled , this number included 133 dogs that were resampled from the 2017 cohort and an additional 16 from CCZ which had been collected in the same area and at the same time as our sample collections . Between 5ml and 10ml of peripheral blood was collected in 10ml K2 EDTA-coated tubes ( BD Vacutainer , UK ) via cephalic or jugular venepuncture by a qualified vet in 2017 and by a CCZ qualified phlebotomist in 2018 . Samples were placed in containers marked with the microchip bar code to aid subsequent tracking and identification . Blood samples were stored in a cool box with a freezer pack before being transferred to a fridge ( 4°C ) prior to processing . Hair samples were obtained by cutting the dorsal hair close to the skin using surgical scissors that had been washed with hexane prior to the collection of each sample by members of the LU research team . A minimum of 2g of hair was collected from each dog . All hair samples were placed in individual foil bags ( 110mm x 185mm; Polypouch UK Ltd , Watford , England ) heat sealed and stored at 4°C prior to analysis . All dogs were assessed for clinical signs of Leishmania infection by veterinarians and CVL control specialists at CCZ . The animals were classified according to the presence of clinical signs which were recorded for each dog . The main signs of CVL considered were onychogryphosis , ophthalmologic abnormalities , adenitis , cachexia , hepatosplenomegaly , alopecia , crusted ulcers and lesions; dogs were classified as asymptomatic ( the absence of clinical signs ) , oligosymptomatic ( the presence of one to three clinical signs ) , or symptomatic ( the presence of more than three clinical signs[32] . Initial VOC analysis was carried out on all ( n = 11 ) of the infected dog hair samples and a sub-set of the uninfected dog hair samples ( n = 44 ) collected in 2017 . The choice of 44 uninfected dog hair samples ( 4 matched uninfected dog hair samples for each infected dog hair sample ) was obtained by a power analysis to optimise the control size . The uninfected dogs were selected from groups of dogs matched by shared characteristics ( age , sex and whether or not treatment for ectoparasites was received ) with infected dogs ( Table 1 ) . Subsequently the VOC analysis was carried out on all the infected dog hair samples ( n = 44 , including 10 CCZ infected dogs ) and all of the uninfected dog hair samples ( n = 105 , including 6 CCZ uninfected dogs ) collected in 2018 . The number of uninfected dog hair did not exceed the “4 uninfected for each infected dog” rule set above , and for this reason all the dogs were used . A VOC analyser ( Model 307 , RoboScientific Ltd , Leeds , UK ) with 11 functioning semi-conducting polymer sensors was used for the analysis . Each sensor has 2 outputs ( positive and negative ) giving a total of 22 responses . Two calibration points were automatically set by the sensor unit; the first was the baseline obtained when carbon-filtered air was passed over the sensor at a flow rate of 200ml min-1 which was automatically adjusted to zero on the Y-axis scale , and the second was a reference point obtained from sampling the head space of 5ml of a liquid water control in a plastic vial . The chemical sensors were thin films of semi-conducting polymers deposited onto interdigitated gold structures on a silicon substrate . We used 12 different sensor types chosen from a group of polymers that included polyaniline , polythiophene and polypyrrole . Each sensor had semi-selectivity to a different group of volatile chemicals; aldehydes , alcohols , amines , organic acids and ketones etc . In this way a digital fingerprint of the VOC mixtures emanating from the samples was generated . Two similar sensor arrays were used in the study , the second array ( used for the 2018 analysis ) was a derivation of the first with 50% of the sensors being identical to the first array . The interaction of the mixtures of VOCs in the samples with the semi-conducting polymer surfaces produced a change in electrical properties ( e . g . voltage and resistance ) over time . This change was measured , recorded and simultaneously displayed on the VOC analyser data logger screen for each sensor . Four parameters were used from each sensor response; the divergence from the baseline ( maximum response ) , the integrated area under each response curve , absorbance and desorbance . Therefore , the total number of VOC measurements produced for each sample were 88 ( 11 sensors x 4 parameters and 2 outputs–positive or negative ) . The sampling profile was set at 2 seconds baseline , 7 seconds of absorption , a 1 second pause , 5 seconds desorption and 12 seconds flush to bring the sensors back to baseline . Water ( DD;10μl ) was injected into each foil bag containing the dog hair samples with a Hamilton syringe and inflated with 140ml of laboratory air using a diaphragm pump . The samples were then incubated at 50°C for 15 minutes in an oven , then allowed to cool to room temperature for 5 minutes prior to head space analysis . For the analysis each foil bag ( containing the dog hair + water ) was sampled by insertion of an 18-gauge needle connected to a PTFE tube through the sidewall of the bag with the tip placed into the head space of each bag . This was connected to the sample port of the VOC analyser and the head space sample was therefore passed over the 12 sensor surfaces . The original flow rate for the sampling was 200 ml min-1 . The headspace of each foil bag was sampled 4 times . The first sample was disregarded as potentially it could contain volatile carryover from the previous sample and thus , we retained the data from the next 3 samples for analysis . The individual dog hair samples in each experiment were tested randomly with each sample used once only . To test the ability of the VOC analyser to differentiate between the odours of infected and uninfected dogs , we employed mclust[37] , a model-based clustering and classification algorithm ( R-CRAN statistical software[38] . This was applied to the known data classes ( infected or uninfected dogs ) . The initial analysis indicated that the model was overfitted , therefore we identified the infected and uninfected dog sub-classes ( unsupervised clustering ) and the analysis was repeated [39] . The robustness of the classification was evaluated by out-of-sample cross validation ( CV ) while the within-group homogeneity of the overfitting models was evaluated by a novel algorithm developed by the authors and termed confounder cross validation ( CCV ) . Finally , the importance of each variable produced by the VOC analyser in discriminating between the infected and uninfected sub-classes was assessed by variable permutation analysis . A more detailed explanation of the rationale for this analysis approach is provided in the S1 Material and a more extensive description of the algorithms is provided in[40] . The VOC analysis dataset contained data from: Three replicate VOC analyser readings were obtained for each dog odour sample . These replicates were considered to be independent , i . e . the three VOC replicates for each dog were considered as coming from three different dogs ( a common procedure for repeated data in clustering analyses ) . The analysis aimed to identify any significant differences in the VOC analyser variables ( used to obtain the means and covariances of the infected and uninfected classes and/or sub-classes ) of infected and uninfected dogs so as to be able to accurately predict the infection state of newly sampled dogs . Initially , the data was evaluated to determine 1 . if infected and uninfected dogs in both 2017 and 2018 could be statistically separated and 2 . if the uninfected dogs in 2017 were statistically separate from uninfected dogs in 2018 . As described above , to take account of overfitting [41] , we reclassified the infected and uninfected classes into sub-classes using the mclust function ( mclust package ) . The optimal inferential method and number of subclasses for infected and uninfected classes was obtained by Bayesian information criterion ( BIC ) ( S1 Material ) , bootstrapping and the likelihood ratio test ( function mclustBootstrapLRT ( mclust package ) .
The best model for the analysis of all infected and uninfected dog classes was EEE apart from the uninfected 2017 dogs which was VVI [40] . The EEE model assumes ellipsoidal covariances and equal shape , volume and orientation for all the classes . The VVI model assumes diagonal covariances with orientation parallel to the coordinate axes with variable shape and volume for all the classes [43] . Between 14 models x 1 to 9 classes were assessed ( i . e . 126 mixture models ) [44] ( S1 Table ( 2017 data ) and S2 Table ( 2018 data ) ) . Clustering analysis of 2017 dogs identified 1 class for uninfected dogs and 3 classes for infected dogs and for the 2018 dogs 2 classes for uninfected dogs and 6 classes for infected dogs were identified . Confusion matrices of the separation obtained from the training set of uninfected vs infected dogs in 2017 and uninfected vs infected dogs in 2018 without sub-classes are given in Table 2A and 2C respectively and with sub-classes in Table 2B and 2D respectively below . These data show that in both years the infected dog odours were significantly different from the uninfected dog odours . In 2017 uninfected dogs were discriminated with 96% specificity and 97% sensitivity , that was improved to 100% for both metrics when the data was divided in sub-classes . The overall training error was reduced ( from 2 . 8% to 0% ) when the 2017 data was divided in sub-classes . In 2018 uninfected and infected dogs were discriminated with 89% specificity and 100% sensitivity and that was improved to 94% specificity and 97% sensitivity when the data was divided in sub-classes . The overall training error was reduced ( from 7 . 6% to 4 . 2% ) when the 2018 data was divided in sub-classes .
The results presented in this study show that by combining VOC ( eNose ) data with robust clustering analysis we can identify dogs infected with Le . infantum by analysis of their odour with very high sensitivity and specificity , regardless of parasite load or the presentation of clinical symptoms . We observed this outcome in two data sets from dog hair samples collected in 2017 ( 99% [0 . 95 , 0 . 99] specificity and 90% [0 . 75 , 0 . 96] sensitivity ) and in 2018 ( 89% [0 . 85 , 0 . 92] specificity and 100% 0 . 97 , 1] sensitivity ) . When the small size of both data sets ( 2017 , 55 dog hair samples: 2018 , 149 dog hair samples ) and consequent potential for overfitting was accounted for by improving the mixture of the models , both sensitivity and specificity increased ( 2017; 100% [0 . 96 , 1] specificity , 100% [0 . 88 , 1] sensitivity: 2018; 94% [0 . 91 , 0 . 96] specificity , 97% [0 . 93 , 0 . 99] sensitivity ) . The robustness of the models was further tested by cross-validation and an novel approach which we have termed confounder cross validation analyses . The model prediction was poor when we used 2 classes ( infected and uninfected ) . However , when we accounted for the heterogeneity within each of these classes and subdivided them into either 4 subclasses ( 2017 data ) or 8 subclasses ( 2018 data ) the sensitivity and specificity and their confidence intervals improved substantially . In both cases the models accurately placed dog odours in the correct infected or uninfected class with a high degree of specificity and sensitivity ( 93% sensitivity and 92% specificity ) . The results suggested that the VOC analyser response was not related to the parasite load in the dog peripheral blood . As the analysis gave sensitivity and specificity responses substantially better than 90% , the effect of parasite load on the VOC analyser response is unlikely to have been significant as the majority of infected animals , regardless of parasite load were detected . However , determining the limits of detection will be important in the future . Previous work has suggested that symptomatic dogs with a greater parasite load produced greater quantities of volatiles than infected asymptomatic dogs[23] . However , asymptomatic dogs can contribute to disease transmission and VL control strategies should target infectious dogs rather than infected dogs per se and in particular the super-spreaders in the population[21 , 45] . In this study we identified Leishmania DNA in circulating blood obtained by cephalic and jugular venepuncture , however the relationship between numbers of circulating parasites in peripheral blood and the infection status of the dog is unclear . In future studies , the skin parasite load , which appears to be more closely related to infectiousness[45] could be correlated with the odour profile . In our study we used molecular techniques , PCR and qPCR , to diagnose and quantify Le . infantum infection in dogs . Although the gold standard diagnosis is considered to be the direct parasitological assessment of lymph node or bone marrow aspirates , in this study we chose to take blood and hair samples from the dogs at their homes . This methodology reduced the possibility of cross-contamination between infected and uninfected dog odour which might have occurred if the dogs had been kept together e . g . at the CCZ facility . This less invasive sampling protocol also reduced stress on the dogs , did not require large facilities ( e . g . for sedation required to obtain bone marrow aspirates ) , reduced the risk of infection to the animal and was more likely to receive owner consent and compliance . A recent study[46] evaluated the accuracy of serological tests , immunochromatographic ( Dual Path Platform: DPP ) and enzyme-linked immunosorbent ( ELISA EIE ) , for CVL in relation to the detection of Leishmania DNA through real-time PCR ) in samples from symptomatic and asymptomatic dogs . The PCR analysis demonstrated greater homogeneity between symptomatic and asymptomatic groups of infected dogs compared with DPP and ELISA . Solcà et al . showed that The diagnosis of CVL through the amplification of kinetoplast DNA presented the highest rates of sensitivity and specificity in comparison with parasitological and serological methods[47] . These authors concluded that molecular methods are required to confirm the infection . Even though serological tests are routinely employed for diagnosing CVL , they have limitations in sensitivity , especially in asymptomatic dogs , and therefore may underestimate Leishmania infection rates[48] . Despite the high specificity , the serological tests present low capacity to detect Leishmania infection in relation to molecular tests [49] . The authors of that study concluded that their study “demonstrated that real-time PCR identified the presence of Leishmania DNA in asymptomatic dogs that had a negative result in serological tests recommended by the official Brazilian protocol for CVL . In addition in a recent study[50] , 34 out of 36 ( 96% ) Leishmania isolates from dogs sampled in GV were found to be Le . infantum , the other 2 isolates were from the Leishmania mexicana complex , Le . ( Le . ) amazonensis Le . ( Le . ) mexicana ) . Therefore , for the purposes of the current study our molecular diagnosis was likely to be representative of the true infection status of the dogs . The prevalence of CVL recorded in our 2017 sample ( 6% ) is low compared to the prevalence recorded in our 2018 sample ( 25 . 6% ) . It is possible that the extensive monitoring carried out by Governador Valadares health authorities in the Altinópolis district of GV , where the study was carried out , immediately prior to our sample collection in 2017 , had an impact on CVL prevalence . However , these values are within the range of prevalence seen previously in studies carried out in the State of Minas Gerais generally e . g . a prevalence of 8 . 1% was observed in Belo Horizonte in dogs surveyed between 2008–2010[42] and 13 . 6% in Divinópolis in 2011[51] . In GV specifically , a study carried out in Altinópolis , between 2008–2011 found 33 . 8% of dogs surveyed to be infected[30] whereas a survey of dogs carried out in 2014–2015 found 22% of dogs to be positive by serology[52] . It has been suggested that change in odour of dogs infected with Le . infantum might be related to the immune response[23] . Changes in odour profile have been observed in other disease states where changes in relatively low molecular weight compounds were expressed as distinct and immediate changes arising from pathophysiological processes occurring and altering the body’s metabolism[19] . However , although the very low parasite loads in some dogs might suggest recent infection , parasite load in the peripheral blood is not indicative[53] and as this study did not determine if the dogs had seroconverted it therefore remains unclear if the odour changes are related to the host immune response or not . Our results also suggest that there was no relationship between clinical state of infection ( symptomatic , oligosymptomatic and asymptomatic ) and detector response . The analyser could accurately detect asymptomatic dogs with low parasite levels as well as symptomatic dogs with high parasite loads . It has been proposed that manipulation of the hosts chemical communication system could enhance the transmission of the parasite to the insect vector and potentially have a significant effect on the epidemiology of the disease[20 , 54 , 55] . Our study examined volatile odours present on the dog hair only , it did not consider the effect of other volatiles , semi-volatiles and non-volatiles from other sources e . g . breath compounds , specialized scent gland secretions , sweat , urine or faeces[56] . The source of the odours that were detected by the VOC analyser is not clear , they could have arisen from the skin , as a result of the metabolic activity of skin microbiota[57] , the immune response or potentially directly from the Le . infantum parasites . Our study did not examine the effect of other infections and the ability of the VOC analyser to differentiate between dogs infected with Le . infantum and other Leishmania spp . or other infections was not determined . In Governador Valadares dogs infected with Le . amazonensis have been found[50] and the sand fly vector Lu . longipalpis infected with multiple Leishmania spp . have been also been found[58] indicating that the epidemiological features require further work . The application of VOC analyser technology is potentially a significant step towards the application of volatile odour analysis in diagnosis of parasitic disease . It raises the possibility that in the future a modified VOC device could provide a rapid , accurate , non-invasive point-of-care diagnostic tool for the specific diagnosis of leishmaniasis in dogs and humans . In our study we found that a small proportion of the sensor variables ( 2 out of 88 in 2017 and 3 out of 88 in 2018 ) contributed significantly to the outcome . Therefore , there is considerable scope for enhancing the sensitivity and specificity of the device through modifications to the sensor chemistry as well as incorporating further improvements to the field collection and analysis of odour . As well as further developments in robustness , portability and simplicity of the device all of which would improve the reliability and utility in the field . A reliable , rapid , accurate , non-invasive additional point-of-care test that identifies Leishmania infection using a different set of disease markers in addition to the DPP CVL test could potentially eliminate the need for the in-laboratory ELISA confirmatory test that currently fails to rapidly diagnose and remove infected dogs from the population . The ability of a VOC analyser , that conforms to the WHO ASSURED criteria , to recognise infection in asymptomatic dogs and dogs with low levels of infection would be of great benefit to VL control programs . The possible integration of a rapid , non-invasive and point-of-care test would be much better accepted by dog owners and could be an important epidemiological tool . In addition to being useful for the selection of infected animals for euthanasia , it could also be useful in the implementation , monitoring and evaluation of leishmania control activities such as insecticide impregnated dog collars that are currently being implemented and sex pheromone-based Lu . longipalpis control programs that are currently being evaluated[59 , 60] . Further work to compare the sensitivity and specificity of a VOC test combined with DPP CVL diagnostics against DPP CVL combined with ELISA is required . The development of a non-invasive POC diagnostic tool based on host odour opens up a myriad range of opportunities to diagnose Leishmania infections in humans and other diseases such as malaria , trypanosomiasis and Chaga’s disease .
|
Visceral leishmaniasis ( VL ) is an insect transmitted , tropical parasitic disease and in Brazil it causes thousands of human deaths every year . Domestic dogs can also be infected , and they are a risk factor for people . The Brazilian Ministry of Health tries to control the disease in 3 ways; first by reducing the population of insects that can carry the disease , second by using therapeutic drugs to treat the disease in humans , and third by identifying and euthanising infected dogs . However , despite these efforts the burden of VL has doubled since 2010 and a significant contributing factor is the lack of a rapid and accurate pathway for diagnosing dogs . In this study we have shown that an eNose can differentiate between the smell of VL infected and uninfected dogs . The analysis was highly sensitive i . e . if the dog was infected , the eNose would detect it in > 97% of the cases and it was highly specific i . e . if the dog was uninfected eNose detect it in >94% of the cases . The outcome was not dependant on the numbers of parasites or the clinical status of the dog . The results suggest that eNose analysis could be used to identify VL infected dogs with improved the speed and accuracy compared to current methods .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"medicine",
"and",
"health",
"sciences",
"body",
"fluids",
"tropical",
"diseases",
"geographical",
"locations",
"vertebrates",
"parasitic",
"diseases",
"dogs",
"parasitic",
"protozoans",
"animals",
"mammals",
"organisms",
"protozoans",
"veterinary",
"diagnostics",
"leishmania",
"neglected",
"tropical",
"diseases",
"molecular",
"biology",
"techniques",
"veterinary",
"science",
"veterinary",
"medicine",
"research",
"and",
"analysis",
"methods",
"infectious",
"diseases",
"zoonoses",
"artificial",
"gene",
"amplification",
"and",
"extension",
"south",
"america",
"protozoan",
"infections",
"molecular",
"biology",
"brazil",
"people",
"and",
"places",
"eukaryota",
"blood",
"anatomy",
"physiology",
"leishmaniasis",
"biology",
"and",
"life",
"sciences",
"amniotes",
"polymerase",
"chain",
"reaction"
] |
2019
|
eNose analysis of volatile chemicals from dogs naturally infected with Leishmania infantum in Brazil
|
Long-chain flavodoxins , ubiquitous electron shuttles containing flavin mononucleotide ( FMN ) as prosthetic group , play an important protective role against reactive oxygen species ( ROS ) in various microorganisms . Pseudomonas aeruginosa is an opportunistic pathogen which frequently has to face ROS toxicity in the environment as well as within the host . We identified a single ORF , hereafter referred to as fldP ( for flavodoxin from P . aeruginosa ) , displaying the highest similarity in length , sequence identity and predicted secondary structure with typical long-chain flavodoxins . The gene was cloned and expressed in Escherichia coli . The recombinant product ( FldP ) could bind FMN and exhibited flavodoxin activity in vitro . Expression of fldP in P . aeruginosa was induced by oxidative stress conditions through an OxyR-independent mechanism , and an fldP-null mutant accumulated higher intracellular ROS levels and exhibited decreased tolerance to H2O2 toxicity compared to wild-type siblings . The mutant phenotype could be complemented by expression of a cyanobacterial flavodoxin . Overexpression of FldP in a mutT-deficient P . aeruginosa strain decreased H2O2-induced cell death and the hypermutability caused by DNA oxidative damage . FldP contributed to the survival of P . aeruginosa within cultured mammalian macrophages and in infected Drosophila melanogaster , which led in turn to accelerated death of the flies . Interestingly , the fldP gene is present in some but not all P . aeruginosa strains , constituting a component of the P . aeruginosa accessory genome . It is located in a genomic island as part of a self-regulated polycistronic operon containing a suite of stress-associated genes . The collected results indicate that the fldP gene encodes a long-chain flavodoxin , which protects the cell from oxidative stress , thereby expanding the capabilities of P . aeruginosa to thrive in hostile environments .
Microorganisms living in aerobic environments are constantly exposed to the harmful effects of reactive oxygen species ( ROS ) , including H2O2 and the superoxide radical , which are generated as unavoidable by-products of oxygen utilization [1] . In addition , commensal and pathogenic bacteria have to face the host oxidative response , such as H2O2 production from phagocytes [1] , [2] . Aerobic organisms have evolved multigenic responses to prevent and/or repair the cellular damage potentially inflicted by these toxic compounds . Whenever defenses are overcome by the amounts of ROS produced , cells are afflicted by a condition called oxidative stress [1] , [2] . Protective mechanisms deployed by stressed organisms include regulation of membrane permeability , antioxidant and repair systems , and replacement of ROS-sensitive targets by resistant isofunctional versions . In microorganisms as distantly related as enterobacteria and cyanobacteria , induction of the mobile electron shuttle flavodoxin ( Fld ) appears to be a common feature of the antioxidant response [3] , [4] . Flds contain flavin mononucleotide ( FMN ) as prosthetic group and are largely isofunctional with the ubiquitous electron carrier ferredoxin ( Fd ) , exchanging reducing equivalents with a promiscuous lot of donors and acceptors . Fld induction is assumed to act as a backup for Fd , which harbors a ROS-sensitive iron-sulfur cluster as redox-active cofactor and whose levels are down-regulated under conditions of environmental stress or iron starvation [5] , [6] . Accordingly , Fld overexpression has been shown to confer augmented tolerance toward various sources of oxidative stress in organisms with very different lifestyles , such as Escherichia coli [7] , rhizobia [8] and plants [9] . Unlike Fds , which are present in all major kingdoms , Flds are restricted to various groups of prokaryotes and some oceanic algae [10] . From sequence alignments and structural considerations , they can be divided into two classes , short-chain and long-chain Flds , which differ by the presence of a 20-amino acid loop of a so far unknown function [11] . Phylogenetic analyses indicate that the two lineages have diverged only once [12] . Pseudomonas aeruginosa is a free-living bacterium commonly found in soil , water , moist locations , and most man-made environments throughout the Earth . P . aeruginosa has a wide metabolic versatility as the reflection of a large and flexible genome with a substantial number of genes , which facilitates its adaptability to thrive in different habitats , and allows a quick response to diverse environmental stimuli and challenges [13] . This remarkable versatility enables P . aeruginosa to infect damaged animal tissues or immunocompromised individuals , where it constitutes an important opportunistic pathogen highly prevalent in nosocomial infections [14] , [15] . Indeed , this bacterium is of particular concern to patients with cystic fibrosis ( CF ) who are highly susceptible to P . aeruginosa and suffer severe and often fatal chronic airway infections [16] . The present research seeks to identify and characterize putative long-chain Flds in P . aeruginosa which could play a protective role under environmental stress conditions . Both P . aeruginosa and its relative Pseudomonas putida contain a single-copy gene encoding a short-chain Fld [17] , [18] , annotated as mioC due to the homology of the product with the homonymous Fld from E . coli , but no long-chain Fld orthologs have been so far reported in pseudomonads . We identified a gene , PA14_22540 ( hereafter referred to as fldP , for flavodoxin from P . aeruginosa ) , which displays low but significant sequence homology to the Anabaena and E . coli Fld genes ( named isiB and fldA , respectively ) , and whose recombinant product was able to display Fld activity in vitro . Expression of the fldP gene in P . aeruginosa was induced by H2O2 treatment via an OxyR-independent pathway , whereas its disruption increased H2O2-induced killing and accumulation of intracellular ROS . The mutant phenotype could be complemented by transformation with the isiB gene from Anabaena . Overexpression of fldP mitigated H2O2-induced cell death in a mutT-deficient P . aeruginosa strain as well as the hypermutability caused by DNA oxidative damage . The presence of a functional fldP gene contributed to P . aeruginosa survival in two model systems of infection: cultured mammalian macrophages and Drosophila melanogaster . Improved bacterial endurance in Drosophila resulted in higher death tolls of the infected flies . In line with its presumptive adaptive role , the fldP gene was found to be part of a self-regulated operon belonging to the P . aeruginosa accessory genome , a collection of strain-specific gene clusters which are acquired en bloc and expand the genomic repertoire to fit the needs for survival in adverse environments . Then , the collected results indicate that the fldP gene encodes a long-chain flavodoxin which is induced when P . aeruginosa is under oxidative stress to exert a protective role against the physiological and mutational damage caused by ROS .
We performed an in silico survey of putative Flds in the genome of P . aeruginosa PA14 and compared the retrieved sequences with those reported for Anabaena ( IsiB ) and E . coli ( FldA ) long-chain Flds , whose protective roles against oxidative stress have been extensively documented [12] . To find orthologs , the IsiB and FldA sequences were compared to the P . aeruginosa PA14 genome using the Domain Enhanced Lookup Time Accelerated Basic Local Alignment Search Tool ( DELTA-BLAST ) , which is sensitive in detecting remote protein homologs [19] . Four unique open reading frames ( ORFs ) displaying both sequence homology and similar domain organization were retrieved . They were a putative oxidoreductase with a covalently-linked Fld-like domain ( PA14_58560 ) , the repressor binding protein WrbA ( PA14_51990 ) , a short-chain Fld similar to E . coli MioC ( PA14_19660 ) , and an ORF ( PA14_22540 , tentatively referred to as FldP ) , which displayed the highest similarity in length and sequence with the long-chain Flds used as baits . Analysis of ORF PA14_22540 indicated that it would encode a 184-amino acid protein with a molecular mass of ∼20 kDa , which is in the range of those typically observed for long-chain Flds ( 170–185 amino acids ) . Multiple-sequence alignment between FldA , IsiB and FldP showed that while FldA and IsiB display 47% identity and 67% similarity , FldP exhibits 23% identity with both flavodoxins , and 50% and 41% similarity with IsiB and FldA , respectively . By using the JPred3 software we next performed a prediction of the secondary structure of FldP and compared it with those of IsiB ( Accession number P0A3E0 ) and FldA ( Accession number P61949 ) . According to this prediction , FldP displays a significant similarity , at the level of the secondary structures , with both FldA and IsiB ( Figure 1 ) . The three proteins fit the common scheme of five β-sheets intercalated with five α-helices , as well as the loosely structured region dividing β5 into β5a and β5b ( between residues 130 and 150 of FldP ) , which is typical of the long-chain class of Flds [11] , [12] . To determine if the product of the fldP gene is a functional Fld , the coding sequence was expressed in E . coli under the control of an inducible promoter ( see Materials and Methods ) . The recombinant protein accumulated to high levels in the bacterial host , but unlike IsiB , which was readily soluble and assembled its prosthetic group in the E . coli cytosol [20] , the P . aeruginosa protein was recovered largely as insoluble inclusion bodies ( Figure S1A ) . Only after the simultaneous expression of a suite of molecular chaperones , a minor but significant amount of the protein could be solubilized and purified by affinity chromatography on Ni-NTA columns ( Figure S1B ) . The purified protein showed a typical flavoprotein spectrum with absorption maxima at 374 and 446 nm , close to those of free FMN in aqueous solution ( Figure 2A ) . The 446-nm peak , which corresponds to transition I of the flavin , is strongly red-shifted in the flavodoxins from Anabaena and E . coli ( Figure 2A ) , indicating that the environment of the prosthetic group is more hydrophobic and/or less solvent-exposed in these flavoproteins . The flavin moiety of IsiB is sandwiched between the aromatic side-chains of Trp58 and Tyr95 , in a coplanar conformation , and the π–π interaction thus established has been regarded as the major cause for the spectral red shift [12] , [21] . The two amino acids are conserved in FldA , but not in FldP , where the tryptophan is replaced by a tyrosine and the tyrosine by a leucine ( Figure 1 ) . The substitutions will prevent aromatic stacking of the isoalloxazine ring system , and introduce conformational changes in its immediate environment , since both residues are located in flexible loops between 3β-3α and 4β-4α ( Figure 1 ) . Then , the absence of red-shifted peaks in the visible absorption spectrum of FldP likely results from a combination of higher solvent exposure and decreased aromatic stacking on the flavin . FldP was assayed in vitro as a substrate of cyanobacterial ferredoxin-NADP ( H ) reductase ( FNR ) . Figure 2B shows that purified FldP was able to mediate FNR-driven cytochrome c reduction in a concentration-dependent manner , with an apparent KM of 1 . 3±0 . 2 µM and a kcat of 22 . 0±1 . 8 min−1 . Under similar conditions , IsiB displayed a kcat of 48 . 5±3 . 8 min−1 ( data not shown ) . The collected results indicate that the product of the fldP gene displayed the structural and functional properties of a bona fide flavodoxin . In order to determine the functional role of FldP in P . aeruginosa , we tested the tolerance exhibited by a fldP-deficient mutant strain to H2O2 toxicity . In the absence of stress , the viabilities of the wt and fldP mutant strains were similar ( Figure S2 ) , but fldP cells were ∼5-fold more sensitive than their wt siblings to the H2O2 treatment ( P = 0 . 024 , Figure 3A ) . Complementation of this mutant with the fldP gene cloned in plasmid p2 ( p2-fldP ) led to an approximately 1 . 4-fold increase in the percentage of surviving cells respect to that observed with the wt strain , although the difference was not statistically significant ( P = 0 . 3929 , Figure 3A ) . Noteworthy , expression of IsiB from the p2-isiB plasmid provided even higher levels of protection against the deleterious effects of H2O2 , with a rise of 2 . 2-fold relative to the wt strain ( P = 0 . 125 , Figure 3A ) . Taking into account these observations , we further tested whether FldP may be involved in the control of intracellular ROS build-up , presumably the cause of H2O2-induced killing . ROS accumulation was quantified in bacterial extracts from the wt , the fldP and the complemented mutant strains after exposure to H2O2 . Cells from the wt strain displayed a small increase of their ROS ( -OOH ) levels as the H2O2 concentration was raised ( Figure 3B ) . Although this increase was not statistically significant , it was consistently observed in a number of experiments . On the other hand , lack of a functional FldP led to ROS build-up in the fldP mutant , whereas expression of FldP from a plasmid in the complemented cells decreased the total -OOH levels to those observed in untreated wt bacteria ( Figure 3B ) . ROS accumulation was also detected in whole P . aeruginosa cells by using the fluorogenic dye 2′ , 7′-dichlorofluorescein diacetate ( DCFDA ) , and visualized by confocal microscopy . Figure 3C shows that the fraction of labelled cells above the detection threshold was significantly higher in the fldP mutant compared to their wt siblings ( 33% vs . 14% ) , but decreased to 2% after complementation with the p2-fldP plasmid , presumably due to the effect of increased genic doses provided by the plasmid . It has been recently reported that P . aeruginosa strains deficient in the 8-oxodeoxiguanine system ( GO ) are particularly vulnerable to oxidative stress [22] , [23] . Specifically , mutT-deficient cells showed to be the most susceptible to oxidants such as H2O2 or methyl viologen . We therefore used this strain to further characterize the protective activity of FldP against ROS . A mutT P . aeruginosa strain was transformed with either p2-fldP or p2-isiB , and the resulting transformants were tested for their susceptibility to H2O2 . Parallel controls were carried out by using the wt and mutT strains harboring an empty p2 plasmid . As shown in Figure 4 , the mutT strain showed a 20-fold decrease in survival after exposure to H2O2 , being significantly more susceptible than the wt strain ( P = 0 . 0119 ) . However , when fldP or isiB were overexpressed in the mutT mutant , cell survival increased dramatically ( 8- and 16-fold , respectively , P = 0 . 0119 ) , relative to mutT transformed with the empty p2 vector ( Figure 4 ) . The results suggest that the antioxidant role of FldP and IsiB can partially compensate the increased susceptibility of mutT-deficient cells to oxidative stress . DNA damage produced by ROS can lead to increased mutation frequencies [24] and emergence of adaptive phenotypes [25] , [26] . Particularly in mutT-deficient mutants , in which the damage produced by oxidative stress cannot be avoided , mutation frequencies can increase 100- to 1000-fold , leading to hypermutator phenotypes [22] . We therefore investigated whether fldP and isiB could display an antimutator effect and alleviate the hypermutability that is typically observed in a mutT-deficient background . Thus , we exposed a mutT strain , which overexpressed fldP or isiB to H2O2 and measured the mutation frequency by determining the emergence of mutants resistant to streptomycin . The mutT strain showed a ∼1200-fold increase in the H2O2-induced mutation frequency ( 3 . 44×10−6 ) , relative to the wt ( 2 . 83×10−9 ) . Expression of either fldP or isiB in the mutT strain diminished the mutation frequencies to 4 . 49×10−7 and 2 . 63×10−7 , which represent a 13% ( P = 0 . 05 ) and 8% ( P = 0 . 0286 ) , respectively , of the mutation frequency showed by mutT cells transformed with p2 ( Figure 5 ) . This last result indicates that overexpression of a functional Fld could avoid the majority ( ∼90% ) , but not all H2O2-induced lesions produced as a consequence of mutT deficiency . The protective effect presumably results from the antioxidant properties of these flavoproteins . Accordingly , the antimutator effect conferred by both fldP and isiB was not observed when the spontaneous mutation frequency was tested ( data not shown ) . To gain further insight into the role played by FldP in oxidative stress tolerance , we studied the transcriptional induction of the fldP gene by H2O2 treatment . The expression of fldP was monitored by semi-quantitative , two-step , reverse transcription-PCR ( RT-PCR ) , using expression of the constitutive housekeeping rpoD gene as control for equal amounts of cDNA in each reaction . Figure 6 shows that expression of fldP in the wt strain was clearly induced ( 3 . 9±0 . 7-fold relative to untreated cultures , P = 0 . 004 ) , after H2O2 treatment . We further investigated whether this increased expression of fldP was dependent on OxyR , the main oxidative stress response regulator of P . aeruginosa [27] . An oxyR deletion mutant strain of P . aeruginosa was constructed and tested for fldP expression in response to H2O2 . Interestingly , no differences were observed in the oxyR strain compared to its isogenic wt , showing a 3 . 0±0 . 7-fold induction in the expression of fldP relative to untreated cultures ( P = 0 . 0045 , Figure 6 ) . Importantly , we also evaluated expression of fldP in a second oxyR strain ( ID54029 ) from the PA14 insertion mutant library [28] , which yielded equivalent results ( data not shown ) . These observations indicate that although fldP is a stress-responsive gene in P . aeruginosa , as in other bacterial species [3] , [4] , this response is triggered by an OxyR-independent mechanism . The use of ROS to kill bacterial pathogens , such as H2O2 production from phagocytes , is a common feature of the innate immune response of eukaryotic organisms . Considering that FldP sheltered P . aeruginosa from oxidative stress under in vitro conditions , we tried to determine if this protective role of FldP could also provide an advantage to cope with the host immune defenses . As a first approach , we evaluated the capacity of the different P . aeruginosa strains to survive in the intracellular milieu of phagocytes by using monolayers of the macrophagic cell line RAW 264 . 7 , which were inoculated with wt P . aeruginosa or its isogenic fldP-deficient strain , complemented or not with p2-fldP . Then , with the addition of antibiotics to kill extracellular bacteria we were able to compare the proportion of bacterial cells of each strain that survived during a 3-h period inside the phagocytes ( see Materials and Methods ) by lysing the cell monolayer and plating the lysates on LB agar . Figure 7A shows that inactivation of fldP produced a moderate but significant decrease of ∼24% in the intracellular survival of P . aeruginosa in phagocytic cells ( P = 0 . 0103 ) . Importantly , this decrease could be reverted by complementation with p2-fldP , even surpassing the wt values . This result indicates that FldP is contributing to the intracellular survival of P . aeruginosa in macrophagic cells , probably by enhancing bacterial resistance to the ROS produced by phagocytes . To evaluate if this protective role of FldP could also be observed in the context of a whole organism infection , which is a more complex system than cultured cells , we used the insect-infection model of D . melanogaster , as previously validated to evaluate virulence traits of P . aeruginosa [29] , [30] . Thus , flies were fed with the same strains of P . aeruginosa mentioned above and left for periods of 1 , 3 or 5 days , after which the ability of each P . aeruginosa strain to survive within the host was scored by counting colony forming units ( CFU ) . Interestingly , the capacity of the fldP mutant to survive within the host dropped to 28 to 36% ( P<0 . 05 ) at the three time-periods assayed , which could recover to wt values after complementation with p2-fldP ( Figure 7B ) . These results suggest that increased bacterial loads could have an impact on host survival . To assess this , D . melanogaster flies were starved for 3 h and then continuously fed with P . aeruginosa . The number of surviving flies was monitored every day until all flies died . The results showed that the flies that had been fed with the wt strain of P . aeruginosa died faster than those fed with the fldP mutant ( Figure 7C ) , which is in agreement with the better capacity of the wt P . aeruginosa strain to survive within the host . Complementation of the mutant bacteria with p2-fldP increased the mortality rate to wt values ( Figure 7C ) . Taken together , these findings suggest that FldP could be playing a role in P . aeruginosa pathogenesis by increasing its resistance to the ROS-dependent clearance carried out by the host immune system , which in turn would result in a more lethal infection due to a higher load of viable bacterial cells in the host . The P . aeruginosa genome is made up by a mosaic of “core” and “accessory and variable” gene clusters [31] , [32] . While the gene composition of the core genome is conserved in almost every strain of P . aeruginosa , genes belonging to the accessory genome are found in discrete patches , referred to as regions of genome plasticity ( RGP ) , which can vary in occurrence and location among strains [33] . By using genome sequence information available online , we evaluated the occurrence of the fldP gene in thirteen P . aeruginosa strains ( namely PAO1 , PA14 , 2192 , 39016 , C3719 , LESB58 , PACS2 , M18 , NCGM2 . S1 , B136-33 , RP73 , DK2 and PA7 ) , whose genomes have been completely sequenced and made available at the Pseudomonas Genome Database ( www . pseudomonas . com ) [34] . The genome of the environmental strain Hex1T [35] , which has been recently sequenced ( Feliziani et al . , unpublished ) , was also included in the analysis . The survey showed that only six of the fourteen strains ( PA14 , 39016 , NCGM2 . S1 , B136-33 , PA7 and Hex1T ) contained the fldP gene , suggesting that it is a component of the P . aeruginosa accessory genome . Indeed , fldP is located in the region of genome plasticity RGP32 [33] . With the exception of the taxonomic outlier PA7 , all other strains showed a conserved synteny of RGP32 , with fldP ( PA14_22540 ) being the fifth gene in a six-gene cluster ( PA14_22500 to PA14_22550 in the PA14 genome ) ( Figure 8 ) . The DNA sequence of RGP32 is highly conserved among strains PA14 , 39016 , NCGM2 . S1 , B136-33 and Hex1T , displaying up to 97% identity . The five ORFs accompanying fldP in RGP32 have been assigned putative functions , based on similarity with known genes ( Table S1 ) . Interestingly , RGP32 is flanked by two palindromic sequences ( inverted sequence repeats ) which could have played a role in the acquisition of this gene cluster . On the other hand , the genome of PA7 only showed orthologs for fldP ( PSPA7_2449 ) with 62% identity , and for the two flanking genes of RGP32 ( PSPA7_2618 and PSPA7_2620 ) , both with 61% identity respect to their PA14 orthologs . Importantly , none of these genes share the genomic location observed for RGP32 in the other five strains . To further investigate the prevalence of fldP , we carried out PCR analysis using conserved primers ( Table S2 ) on a collection of clinical and environmental isolates of P . aeruginosa , which were selected on the basis of being clonally different [36] . While highly divergent fldP versions could have been overlooked by this experimental approach , the high degree of sequence conservation of fldP observed among the characterized P . aeruginosa genomes makes this possibility rather improbable . The analysis showed that ∼23% of clones amplified a fragment corresponding to the fldP gene ( Figure S3 ) . Considering this low prevalence , the origin of RGP32 in P . aeruginosa is better explained as a consequence of DNA acquisition through one or more horizontal gene transfer processes rather than a genetic loss in those isolates which do not harbor RGP32 . We subsequently analyzed the transcriptional organization of RGP32 in order to elucidate whether the fldP gene behaved as a single transcriptional unit or showed co-expression with other genes of the same variable region , thereby constituting an operon . We carried out PCRs using cDNA as template and primers designed to amplify fragments containing regions of two neighboring genes . We measured co-expression of those RGP32 genes which shared the same orientation and were located downstream from fldP ( fldP to PA14_22500 ) . As shown in Figure 9A , we were able to amplify fragments between fldP and PA14_22530 , PA14_22530 and PA14_22520 , and PA14_22520 and PA14_22510 . In contrast , no PCR fragments could be amplified between PA14_22510 and PA14_22500 . Importantly , amplicons were obtained among all adjacent genes , even between PA14_22510 and PA14_22500 , when genomic DNA templates were used as positive controls ( Figure 9A ) . These results depict a transcriptional organization of RGP32 structured in one polycistronic operon containing fldP , PA14_22530 , PA14_22520 and PA14_22510 , and two monocistronic transcriptional units for genes PA14_22500 and PA14_22550 . To strengthen further these observations , we performed RT-PCRs to measure transcripts of PA14_22530 and PA14_22500 in cultures which were treated or not with H2O2 . Interestingly , expression of PA14_22530 exhibited a 3-fold H2O2-dependent induction , similar to that previously observed for fldP , whereas transcripts of PA14_22500 showed only a small increase ( Figure 9B ) . This result is in line with the transcriptional organization of RGP32 described above , and raises the question as to how this fldP-containing operon could be regulated . It has been previously observed that genomic islands often possess their own regulatory elements . In this sense , gene PA14_22550 , which is predicted to encode a LysR family transcriptional regulator , appears as a promising candidate to fulfill this role . We investigated this possibility by using a PA14_22550 null mutant strain ( ID53714 ) [28] , and compared the expression of fldP and its response to H2O2 with those previously observed for the isogenic wt strain . Surprisingly , RT-PCR analyses showed that inactivation of PA14_22550 produced a 3 . 4-fold increase in the expression of fldP ( P = 0 . 0285 ) , even in untreated cells ( Figure 6 ) . In fact , treatment with H2O2 only raised this difference to 4 . 1-fold ( P = 0 . 004 ) . Thus , following the increase observed due to PA14_22550 inactivation , exposure to H2O2 did not produce any further rise of fldP expression ( P = 0 . 4558 ) . Then , the collected results indicate that fldP is part of a polycistronic operon which is repressed by the LysR-like transcriptional regulator present in the same RGP .
Even under optimal growth conditions a low percentage of the electrons involved in cellular redox pathways are diverted to oxygen with concomitant ROS generation . This fraction can be dramatically increased under adverse environmental situations , leading to a condition known as oxidative stress . Moreover , cellular auto-oxidations are not the only source of ROS and oxidative stress; they can also be generated by external redox processes , such as phagocytes and other eukaryotic cells , which douse invading pathogens with H2O2 as a strategy to prevent infection ( reviewed in [1] , [2] ) . A common strategy for coping with ROS damage and oxidative stress is the use of electron shuttles to relieve the excess of reducing equivalents and redox-active compounds . One of the most conspicuous among them is the electron carrier flavoprotein Fld , which has been consistently associated with stress protection in a number of organisms . The long-chain Flds from Anabaena and E . coli ( IsiB and FldA , respectively ) are encoded by oxidant-inducible genes , and confer increased tolerance to environmental and nutritional hardships [3] , [4] . Despite the physiological importance of Fld as adaptive resource , little had been described about these flavoproteins in the versatile opportunistic human pathogen P . aeruginosa . In this work , we report the presence of a long-chain Fld ( FldP ) in P . aeruginosa and describe its role in the defense of the bacterial cell against oxidative conditions . FldP is encoded by the PA14_22540 gene as a 184-amino acid product sharing overall sequence similarity and common structural signatures with well-characterized flavodoxins such as FldA and IsiB ( Figure 1 ) . The fldP gene was cloned , expressed in E . coli and the resulting product purified to homogeneity , displaying the spectral properties and activity of a functional Fld ( Figure 2 ) . We deemed it necessary to confirm that FldP was a functional flavodoxin because short-chain Flds have been already identified in both P . aeruginosa and P . putida [17] , [18] . These genes have been annotated as mioC by comparison with their ortholog in E . coli . Pseudomonas MioC behaves as a functional Fld , mediating FNR-catalyzed electron transfer to cytochrome c with a kcat of about 6 min−1 [18] , comparable to that displayed by FldP ( 22 min−1 , Figure 2B ) , but its physiological role remains yet to be determined . Null mutants in mioC did not display growth phenotypes in E . coli [37] , [38] or P . aeruginosa [17] . In the latter organism , however , mioC mutations exhibited other pleiotropic effects , including altered response to iron stress , increased production of the extracellular pigments pyocyanin and pyoverdine , and modified resistance to antibiotics [17] . In contrast to the enhanced sensitivity of fldP mutants toward H2O2 toxicity ( Figure 3A ) , P . aeruginosa cells deficient in mioC were more tolerant than the wt to oxidative stress caused by methyl viologen or H2O2 [17] . The presence of different non-redundant Flds in a single organism is not uncommon . The E . coli genome , for instance , contains at least four genes predicted to encode flavodoxins: fldA , fldB , mioC , and yqcA [38] . E . coli Flds engage in different cellular pathways , and in general they cannot be functionally exchanged [38] , [39] . Moreover , the fldA gene is specifically induced by H2O2 and redox-cycling oxidants [40] , [41] , whereas mioC is not [37] . Likewise , FldP and MioC , while displaying essentially the same activity in vitro , appear to play different roles in P . aeruginosa , with FldP contributing to the protection against oxidative challenges , and MioC in the response to iron stress [17] . A search of fldP orthologs in the Pseudomonas genomes available in the Pseudomonas Genome Project ( http://www . pseudomonas . com ) [34] and the environmental Hex1T strain ( Feliziani et al . , unpublished ) revealed their presence in just six out of fourteen strains . The P . aeruginosa genome is made up of a conserved core component disrupted by numerous strain-specific RGPs , the sum of which conform the accessory and variable genome [31] , [33] . The fldP gene is part of the region of genome plasticity RGP32 , belonging to the accessory genome ( Figure 8 ) . Synteny and genome location are highly conserved in the PA14 , PA39016 , NCGM2 . S1 , B136-33 and Hex1T strains , with flanking genes being present in PAO1 , suggesting that RGP32 is the result of a common insertion event . In this sense , we identified conserved palindromic inverted repeats flanking RGP32 in all these genomes , which provide circumstantial evidences on the origin and acquisition of this genome block . While the conserved repertoire of the core genome codes for central functions required for survival and reproduction in any habitat [31] , [42] , the variable regions may play customized roles in adaptation to particular environments [33] , [43] . This singular genome architecture , combining conserved and variable components , bestows P . aeruginosa pangenome upon its capability to handle a broad metabolic potential in order to adapt to the widest range of environmental niches . In this context , fldP , as part of the accessory genome , may provide specialized oxidative stress-related functions that benefit survival under stressful conditions , conferring RGP32 a role as adaptive island . Indeed , several lines of evidence indicate that FldP is involved in protection against oxidative stress , including ( i ) the strong induction of the fldP gene in response to H2O2 ( Figure 6 ) , ( ii ) the enhanced ROS build-up and lower survival of fldP null mutants exposed to H2O2 ( Figure 3 ) , and ( iii ) the partial protection conferred by FldP overexpression to P . aeruginosa cells deficient in mutT against the deleterious effects ( Figure 4 ) and increased mutational burden ( Figure 5 ) caused by H2O2 treatment . P . aeruginosa houses a multifaceted antioxidant response , comprising scavenging enzymes such as catalases and superoxide dismutases , thioredoxins and glutaredoxins , as well as small antioxidant molecules such as glutathione and melanine [44] . OxyR can be considered the master P . aeruginosa oxidative stress adaptive response [27] . Furthermore , recent studies have extended the implication of OxyR in other P . aeruginosa important responses , such as quorum sensing regulation , iron homeostasis and oxidative phosphorylation , also identifying a large number of target genes and revealing a far more complex cellular response than previously envisaged [45] . It was therefore tempting to speculate that fldP could be just another effector gene of this transcriptional regulator . Analysis of fldP expression in two oxyR loss-of-function mutants ruled out this possibility ( Figure 6 ) , indicating that induction of fldP in response to oxidative stress proceeds through an OxyR-independent pathway . Inspection of the gene content and organization of RGP32 suggested another attractive possibility . The fldP gene is the first in a row of five ORFs displaying the same transcriptional orientation and extending up to the left end of RGP32 ( Figure 8 ) . This accessory genomic region contains still another ORF ( PA14_22550 ) , which is divergently transcribed and encodes a putative protein related to the LysR family of transcriptional regulators ( Figure 8 ) . LysR is the prototype for the most extended family of transcription factors in the bacterial world . Although originally described as activators of divergently transcribed genes , subsequent research placed LysR proteins as global transcriptional regulators , acting as either activators or repressors of single or operonic genes ( reviewed in [46] ) . We found that four of the five divergent ORFs ( including fldP ) were co-transcribed in response to oxidative stress ( Figure 9A ) , therefore constituting an operon . Moreover , inactivation of PA14_22550 led to full induction of fldP in the absence of oxidants , and to H2O2 insensitivity ( Figure 6 ) , indicating that the LysR homologue acts as a repressor of fldP and presumably the entire operon . The sixth ORF of RGP32 , whose presumptive product is homologous to protein-disulfide isomerases ( Table S1 ) , is not part of the operon . Stress-associated traits are often encoded by loci adjacent to those of other defensive products , especially when they are co-regulated [31] , [42] . In RGP32 , the ORF immediately downstream from fldP encodes a putative glutathione S-transferase ( Table S1 ) . Involvement of this superfamily of conjugative enzymes in the protection against oxidative stress has been extensively documented in all types of organisms ( see , for instance [47] ) . On the other hand , nucleoside-disphosphate-sugar epimerases as that presumably encoded by PA14_22520 have been consistently associated with stress responses in plants [48] , fungi [49] and bacteria [50] . Finally , PA14_22510 encodes a putative H protein from the glycine cleavage system ( Table S1 ) . These are lipoate-containing carrier proteins which provide reducing power ( in the form of thiols ) to the multi-enzymatic complex involved in glycine decarboxylation . The gene encoding this protein has been shown to be strongly up-regulated by H2O2 in streptococci [51] , and lipoic acid is known to participate in the cellular response against oxidative stress in different organisms [52] . Then , all members of the operon have the potential to contribute to the defense against oxidative stress at different levels . We therefore propose that RGP32 represents a stress-inducible , self-regulated genetic element which confers increased tolerance to oxidative challenges . Under normal growth conditions , expression of RGP32 genes should be repressed by the LysR-type regulator encoded by PA14_22550 . Oxidants somehow inactivate this transcription factor allowing induction of fldP and other components of the operon . The mechanism by which oxidants modulate the activity of the LysR-like protein of RGP32 is at present unknown and deserves further investigation . The protective effect displayed in vitro by FldP against oxidative stress prompted us to evaluate whether this tolerance could be advantageous to P . aeruginosa cells exposed to the oxidative assault of cells belonging to the mammalian immune system . Indeed , wt and complemented P . aeruginosa strains expressing FldP from the chromosome or a plasmid exhibited better survival within infected macrophages relative to an isogenic mutant lacking this flavoprotein ( Figure 7A ) . Cifani et al . [53] have recently shown that the oxidative burst produced by P . aeruginosa-infected macrophages plays a key role in the short-term killing of intracellular bacteria following invasion , strongly suggesting that the protective effect of FldP stems from its antioxidant function as it occurs in vitro . We also used the insect model system of D . melanogaster ( in which the PA14 strains has been shown to be particularly aggressive [29] ) , to further investigate the importance of this protective effect on a real infection process . In good agreement with the differential sensitivity observed in macrophages , fldP mutant bacteria accumulated to lower levels in D . melanogaster , as compared to the wt or the complemented strains ( Figure 7B ) . Noteworthy , the mortality kinetics was delayed for ∼24 h in the mutant-infected flies , whereas the death rate of the flies infected with the fldP-deficient bacteria complemented with p2-fldP was equivalent to that of the wt strain ( Figure 7C ) . Thus , the oxidative killing of P . aeruginosa within Drosophila hemolymph may involve mechanisms similar to those utilized by mammalian hosts . While the FldP effect suggests that the electron shuttle could be involved in P . aeruginosa virulence , it is more likely that the different death rates observed in Figure 7C are a consequence of longer persistence of FldP-containing bacteria in the fly due to the adaptive advantages conferred by the flavoprotein . In line with this proposal , mutation of the P . aeruginosa oxyR gene had similar effects on bacterial survival and Drosophila killing as those reported here [54] . Then , our data identify an oxidant-responsive long-chain flavodoxin in P . aeruginosa , which participates in the defense against ROS and contributes to the bacterial tolerance to the oxidative onslaught elicited by the host immune system . We anticipate that studies on this direction will lead to a more comprehensive panorama of the mechanisms allowing this opportunistic pathogen to adapt and persist in stressful and dynamic environments .
P . aeruginosa PA14 and its isogenic strains ID38939 ( PA14_22540 mutant , fldP ) , ID53714 ( PA14_22550 mutant ) and ID54029 ( PA14_70560 mutant , oxyR ) were kindly provided by Dr Eliana Drenkard and Dr Jonathan Urbach from the Massachusetts General Hospital , Boston , USA [28] , whereas P . aeruginosa MPAO1 and its isogenic mutT strain were provided by Dr Michael Jacobs from the University of Washington Genome Center , USA [55] . Insertion of the MAR2xT7 mini-transposon in mutant ID38939 is unlikely to have polar effects on the expression of downstream genes since the consecutive aacC1 promoter is oriented in the same direction . To prepare inocula , bacteria were routinely cultured on Luria-Bertani ( LB ) agar plates from frozen stocks and subcultured overnight in LB liquid medium at 37°C with shaking at 220 r . p . m . Antibiotics were used at the following concentrations: 30 µg ml−1 gentamicin ( Gm ) ; 250 µg ml−1 kanamycin ( Km ) . To find Fld homologs in P . aeruginosa PA14 , the Anabaena IsiB and E . coli FldA sequences were compared against the entire PA14 genome using the DELTA-BLAST Search Tool [19] . Multiple DNA sequence alignments were performed by using ClustalW ( http://www . clustal . org ) . Secondary structures were predicted using the Jpred3 software provided by the Dundee Scotland University ( http://www . compbio . dundee . ac . uk/www-jpred ) . A DNA fragment containing the entire coding region of the fldP gene ( PA14_22540 ) was amplified by PCR from PA14 genomic DNA using oligonucleotides FldP-F and FldP-R ( Table S2 ) , containing BamHI and HindIII restriction sites , respectively . The PCR product was ligated to the pGem-T Easy vector ( Promega ) and subsequently cloned into the broad-host-range plasmid pBBR1MCS2 ( p2 ) , which harbors a Km resistance marker [56] , to generate p2-fldP . A similar strategy was employed to prepare p2-isiB containing the Fld-encoding gene from Anabaena PCC7119 , originally cloned in pEMBL8-isiB [20] , using oligonucleotides IsiB-F and IsiB-R as forward and reverse primers , respectively ( Table S2 ) . The resulting plasmids ( p2-fldP and p2-isiB ) and the empty p2 vector ( as control ) , were introduced in the different P . aeruginosa strains via electroporation [57] . Complete deletion of the oxyR gene was carried out as previously described [57] . All primer sequences are described in Table S2 . Briefly , a first round of three PCR reactions was performed in which the 5′ and 3′ flanking regions of oxyR , as well as a Gm resistance cassette were amplified from plasmid pPS856 [58] using four gene-specific primers ( Oxy-UpF-GWL , Oxy-UpR-Gm , Oxy-DnF-Gm and Oxy-DnR-GWR ) and the common Gm-specific primers ( Gm-F and Gm-R ) . This generated three fragments with partial overlaps either to each other or the attB1 and attB2 recombination sites . The purified fragments were then assembled in vitro by overlap extension during the second round PCR using the common primers GW-attB1 and GW-attB2 . This resulted in an oxyR-deletion-mutant PCR fragment which was subsequently cloned into pDONR221 ( Invitrogen ) via the BP clonase reaction to create pDONR221-oxyR::Gm . This construct served as the substrate for LR clonase-mediated recombination into the destination vector pEX18ApGW . The resulting suicide vector pEX18ApGW-oxyR::Gm was then transferred to P . aeruginosa and the plasmid-borne oxyR-deletion mutation was exchanged with the chromosome via homologous recombination to generate the chromosomal deletion mutant . For expression in E . coli , a 568-bp fragment encoding the complete sequence of the fldP gene was obtained by PCR amplification , using p2-fldP as template , and primers Rec-FldP-F and Rec-FldP-R , which contain restriction sites for NdeI and HindIII , respectively ( Table S2 ) . The amplified fragment was digested with the corresponding enzymes , cloned into compatible sites of pET-TEV ( Novagen ) under the control of the T7 promoter , and fused in-frame to an N-terminal His-tag . To improve solubility , BL21 E . coli cells were co-transformed with this vector and plasmid pG-Tf2 ( Takara Bio Inc ) expressing E . coli molecular chaperones ( GroEL , GroES and Trigger Factor ) . After induction with 0 . 2 mM isopropyl-β-D-thiogalactoside ( IPTG ) , the soluble flavoprotein was purified from cleared lysates in a Ni-NTA column by elution with 500 mM imidazole . Expression and purification of IsiB were carried out according to Fillat et al . [20] . UV-visible spectra of the purified recombinant proteins and FMN were recorded in 50 mM Tris-HCl pH 8 . 5 . The ability of both electron shuttles , FldP and IsiB , to mediate the cytochrome c reductase activity of Anabaena FNR was assayed according to Shin [59] . The reaction mixture contained 3 mM glucose 6-phosphate , 0 . 3 mM NADP+ and 1 unit ml−1 glucose 6-phosphate dehydrogenase ( G6PDH , to generate NADPH ) , 0 . 5 µM FNR , 50 µM equine heart cytochrome c and various amounts of Fld in 50 mM Tris-HCl pH 8 . 5 . Cytochrome c reduction was followed at 30°C by the increase in absorbance at 550 nm ( ε550 = 19 mM−1 cm−1 ) . A modified version of the FOX II assay [60] was used to quantify the presence of peroxides in bacterial extracts . Cultures of the parental PA14 and the fldP mutant strains transformed with either p2 or p2-fldP were grown aerobically in LB broth at 37°C for 5 h with the appropriate antibiotics . Then , H2O2 was added to final concentrations of 0 , 25 and 50 mM , and bacterial suspensions were incubated for 30 min with vigorous shaking . Cultures were split into two equal portions; one of them was used to measure protein concentration , while the other was centrifuged , washed with 0 . 9% ( w/v ) NaCl and finally resuspended in 1 ml of an 80∶20 ethanol/water solution containing 0 . 01% ( w/v ) butylated hydroxytoluene ( BHT ) . Samples were disrupted by sonic oscillation ( 10 times for 10 sec each , 30% amplitude ) , centrifuged at 10 , 000 g for 10 min , and 250 µl of the supernatants were combined with 250 µl of 10 mM Tris-phenyl phosphine in methanol ( TPP , a -OOH reducing agent ) , or with 250 µl of methanol , to measure total oxidants . Mixtures were incubated for 30 min to allow complete -OOH reduction by TPP . Five hundred µl of FOX reagent ( 100 µM xylenol orange , 4 mM BHT , 250 µM ferrous ammonium sulphate and 25 mM H2SO4 in 90% ( v/v ) methanol ) were then added to each sample , and the absorbance at 560 nm was recorded 10 min after reagent addition . The absorbance differences between equivalent samples with and without TPP indicate the amounts of -OOH , which were calculated using a 0–20 µM H2O2 standard curve . Protein concentrations were estimated in cleared lysates in 50 mM Tris-HCl pH 8 . 0 by a dye binding method [61] , using bovine serum albumin as standard . For observation at the confocal microscope , DCFDA was introduced into P . aeruginosa cells by electroporation . Briefly , 10-ml overnight cultures of the various strains were collected by centrifugation at 12 , 000 g for 2 min , washed three times with 1 ml of 0 . 3 M sucrose and finally resuspended in 100 µl of the same solution . DCFDA ( in dimethyl sulfoxide ) was added to a final concentration of 500 µM and the suspension transferred to electroporation cuvettes of 1-mm width . Cells were electroporated in an Electro Cell Manipulator 600 , BTX electroporation System at 25 µF , 2 . 5 kV and 200 Ω for 5 ms , diluted with 300 µl of 0 . 3 M sucrose and divided into two equal fractions . One of them was incubated with 25 mM H2O2 at 37°C for 15 min and the other kept under the same conditions without the oxidant . Seven µl of the suspensions were mixed with 3 µl of the membrane-staining dye FM 4-64 ( 0 . 1 mg ml−1 ) on the surface of a glass plate and the lid was sealed with colorless nail paint . Cells were observed at excitation/emission maxima of ∼515/640 nm in a Nikon TE-2000-E2 confocal microscope . The number of fluorescent spots was determined for each dye by using the Image J 1 . 46 software . For H2O2 susceptibility assays , cells were grown to mid-exponential phase ( OD600∼0 . 8 ) , collected by centrifugation and washed with 0 . 9% ( w/v ) NaCl solution . Then , bacteria were normalized to ∼108 cells and treated with 50 mM H2O2 for 30 min at 37°C . Cells were centrifuged again , washed twice with 0 . 9% ( w/v ) NaCl and finally resuspended in fresh LB medium . Serial dilutions were spotted on LB agar plates and incubated overnight at 37°C to determine viability . Non-treated controls were included . Bacterial susceptibility was expressed as the ratio between the survivals of treated to untreated cells . For estimation of spontaneous-mutation frequencies , five bacterial colonies of the different P . aeruginosa strains were cultured overnight in 10 ml LB medium at 37°C with shaking at 220 r . p . m . Appropriate dilutions of the cultures were plated on LB agar to determine the total number of viable cells , or on LB agar supplemented with 500 µg ml−1 streptomycin to count the number of streptomycin-resistant cells , after overnight incubation at 37°C . Spontaneous-mutation frequency was determined as the ratio between the number of streptomycin–resistant cells and the number of viable cells . Determination of H2O2-induced mutant frequency was carried out according to previously described protocols [62] , with some modifications . Briefly , five independent bacterial cultures for each strain were grown in LB to mid-exponential phase , collected by centrifugation and washed with 0 . 9% ( w/v ) NaCl . Cells were normalized to ∼108 and subsequently treated with 50 mM H2O2 for 30 min at 37°C , centrifuged again and washed twice with 1 ml of 0 . 9% ( w/v ) NaCl . An aliquot of treated cells ( 0 . 5 ml ) was diluted 10-fold in fresh LB broth and cultured overnight at 37°C with shaking at 220 r . p . m . Then , 100 µl of each overnight culture were plated onto LB agar supplemented with 500 µg ml−1 streptomycin , whereas aliquots from appropriate dilutions were plated on LB agar without antibiotic to measure cell viability . The H2O2-induced mutation frequency was calculated as above . P . aeruginosa PA14 mid-exponential phase cultures ( OD600∼0 . 8 ) were treated with 50 mM H2O2 for 15 min , centrifuged and subsequently washed twice with 0 . 9% ( w/v ) NaCl . Treated and untreated cultures were used to extract total RNA using the RNA Purification Kit ( Fermentas ) . RNA was quantified by UV spectrophotometry , and its integrity was checked by electrophoresis in 1 . 5% ( w/v ) agarose gels . Then , 1 µg of total RNA was reverse-transcribed using the QuantiTect Reverse Transcription Kit ( QIAGEN ) . PCR primers were manually designed with the assistance of the Netprimer software ( PREMIER Biosoft International , Palo Alto , CA ) and evaluated for their specificity with the BLAST program at the NCBI Web site . Specific transcripts were semi-quantitatively measured by RT-PCR using primers FldP-RT-F and FldP-RT-R ( for fldP ) , 22530-RT-F and 22530-RT-R ( for PA14_22530 ) , and 22500-RT-F and 22500-RT-R ( for PA14_22500 ) . Transcripts of the rpoD gene were amplified with primers RpoD-RT-F and RpoD-RT-R and served as housekeeping controls . All primer sequences are described in Table S2 . The optimal number of cycles was determined in advance to evaluate expression in the exponential phase of amplification . Final cycling conditions included a hot start at 95°C for 10 min , followed by 32 cycles at 94°C for 30 sec , 60°C for 30 sec and 72°C for 30 sec , and a final extension at 72°C for 5 min . Specificity was verified by agarose gel electrophoresis . Fold change in gene expression was calculated by measuring band intensities with the Gel-Pro Analyzer Software . No amplification was observed in PCR reactions containing water or non-reverse transcribed RNA as template . Primers were designed in order to amplify PCR products containing regions of two neighboring genes for those genes of RGP32 which share the same orientation and are located downstream from fldP ( Figure 9A ) . The cDNA to be used as a template for PCR was obtained by reverse transcription of purified total RNA as described above . Thus , primers FldP-F and FldP-30-R were employed to determine the co-expression of genes fldP and PA14_22530 ( Table S2 ) . In the same way , co-expression of downstream contiguous genes was analyzed with the following primers: a ) 30-20-F and 30-20-R to analyze PA14_22530 and PA14_22520 co-expression; b ) 20-10-F and 20-10-R for PA14_22520 and PA14_22510; and c ) 10-00-F and 10-00-R for PA14_22510 and PA14_22500 ( Table S2 ) . Genomic DNA was used as template for positive controls of the PCR reactions . No amplification was observed in PCR reactions containing water or non-reverse transcribed RNA as template . Cells of the macrophagic-derived line RAW 264 . 7 were grown on 12-mm-diameter wells until they reached 95–100% confluence . To prepare inocula , overnight cultures of P . aeruginosa were washed with phosphate buffered saline ( PBS ) and suspended in Dulbecco's Modified Eagle's Medium ( DMEM ) at a final concentration of 107 cells ml−1 . Bacterial antibiotic protection assays were conducted as previously described with few modifications [63] . Briefly , macrophagic cells were inoculated with the various P . aeruginosa strains by the addition of 500 µl of the bacterial suspension , corresponding to a multiplicity of infection ( MOI ) of about 20 , followed by centrifugation for 10 min at 1000 r . p . m . to facilitate cell contact . After a 2-h incubation period at 37°C , the supernatant was removed and the cells were washed three times with PBS to remove non-associated bacteria . To count intracellular P . aeruginosa , fresh DMEM with streptomycin ( 1 mg ml−1 ) and carbenicillin ( 400 µg ml−1 ) was added to the cell monolayer and incubated for 2 h to kill extracellular bacteria . Antibiotic bactericidal activity was confirmed by plating 100-µl aliquots from the wells directly on LB-agar . Cells were again washed to remove antibiotics , then lysed by incubation with 0 . 1% ( v/v ) Triton X-100 in PBS . At this initial stage ( Ti ) intracellular bacteria were enumerated by plating serial dilutions of cell lysates on LB agar and counting CFU . To estimate P . aeruginosa intracellular survival , an equivalent set of wells were incubated with antibiotics for an additional 3-h period ( Tf ) , cells were lysed and surviving bacteria enumerated as described above . Intracellular survival was calculated as the ratio between the final and the initial ( Tf/Ti ) CFU counts . Three wells of cells were used for each strain in each experiment , and all experiments were repeated three times . To rule out the pre-existence of resistant bacteria in the overnight cultures , 10× inocula ( 5×107 cells ) were plated on LB agar containing streptomycin ( 1 mg ml−1 ) and carbenicillin ( 400 µg ml−1 ) , with not a single CFU observed even after 72 h of incubation . Phagocytes viability was monitored at every time point of the experiment by measuring release of LDH ( CellTiter 96 Aqueous Non-Radioactive Cell Proliferation Assay , Promega ) , which showed an initial decrease of ∼50% in the first 2 h , but remained unaltered after the antibiotic treatment . P . aeruginosa cells of the various strains were grown overnight , washed and diluted in PBS , 10% ( w/v ) sucrose , 250 µg ml−1 Km to OD600 = 0 . 025 . D . melanogaster strain W1118 flies ( 4–5 days old ) were starved for 3 h and then fed continuously at 25°C on cottons plugs , which had been previously embedded in the bacterial solution . For each time point , 3 groups of 5 flies were pestle homogenized in 200 µl of PBS and the homogenate was serially diluted in PBS and plated on LB agar containing 250 µg ml−1 Km , to quantify the number of bacterial colonies . Overnight cultures of each P . aeruginosa strain were diluted in fresh LB to an OD600 = 0 . 25 . This solution was then diluted 10-fold in PBS , 10% ( w/v ) sucrose , 250 µg ml−1 Km . Groups of 15 D . melanogaster W1118 male flies were starved during 3 h and then placed in vials with sterile cotton plugs , which had been previously embedded in 3 ml of the bacterial suspension . Flies were kept at 25°C and survival was monitored daily . Three groups of 15 flies were used for each condition in three independent experiments . Statistical analyses were performed using one-tailed Mann-Whitney test appropriate for nonparametric adjustment . When appropriate , one- or two-tailed Student's t test was applied . In all cases , P values less than or equal to 0 . 05 were considered statistically significant . The collection of P . aeruginosa clinical isolates used in this work has already been published by Feliziani et al . [36] . In this previous study , the informed consent as well as the approval from an Institutional Research Committee were appropriately evaluated and fulfilled the PLOS ethical standards .
|
Coping with toxic reactive oxygen species ( ROS ) generated as by-products of aerobic metabolism is a major challenge for O2-thriving organisms , which deploy multilevel responses to prevent ROS-triggered damage , including membrane modifications , induction of antioxidant and repair systems and/or replacement of ROS-sensitive targets by resistant isofunctional versions , among others . The opportunistic pathogen Pseudomonas aeruginosa is frequently exposed to ROS in the environment as well as within the host , and we describe herein a new response by which this microorganism can deal with oxidative stress . This pathway depends on a previously uncharacterized gene that we named fldP ( for flavodoxin from P . aeruginosa ) , which encodes a flavoprotein that belongs to the family of long-chain flavodoxins . FldP exhibited a protective role against ROS-dependent physiological and mutational damage , and contributed to the survival of P . aeruginosa during in vivo infection of flies as well as within mammalian macrophagic cells . Thus , fldP increases the adaptive repertoire of P . aeruginosa to face oxidative stress .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"genetics",
"microbiology",
"biology",
"bacterial",
"pathogens",
"gene",
"function"
] |
2014
|
A Long-Chain Flavodoxin Protects Pseudomonas aeruginosa from Oxidative Stress and Host Bacterial Clearance
|
Despite the high degree of HIV-1 protease and reverse transcriptase ( RT ) mutation in the setting of antiretroviral therapy , the spectrum of possible virus variants appears to be limited by patterns of amino acid covariation . We analyzed patterns of amino acid covariation in protease and RT sequences from more than 7 , 000 persons infected with HIV-1 subtype B viruses obtained from the Stanford HIV Drug Resistance Database ( http://hivdb . stanford . edu ) . In addition , we examined the relationship between conditional probabilities associated with a pair of mutations and the order in which those mutations developed in viruses for which longitudinal sequence data were available . Patterns of RT covariation were dominated by the distinct clustering of Type I and Type II thymidine analog mutations and the Q151M-associated mutations . Patterns of protease covariation were dominated by the clustering of nelfinavir-associated mutations ( D30N and N88D ) , two main groups of protease inhibitor ( PI ) –resistance mutations associated either with V82A or L90M , and a tight cluster of mutations associated with decreased susceptibility to amprenavir and the most recently approved PI darunavir . Different patterns of covariation were frequently observed for different mutations at the same position including the RT mutations T69D versus T69N , L74V versus L74I , V75I versus V75M , T215F versus T215Y , and K219Q/E versus K219N/R , and the protease mutations M46I versus M46L , I54V versus I54M/L , and N88D versus N88S . Sequence data from persons with correlated mutations in whom earlier sequences were available confirmed that the conditional probabilities associated with correlated mutation pairs could be used to predict the order in which the mutations were likely to have developed . Whereas accessory nucleoside RT inhibitor–resistance mutations nearly always follow primary nucleoside RT inhibitor–resistance mutations , accessory PI-resistance mutations often preceded primary PI-resistance mutations .
HIV-1 is a highly mutable pathogen . In the decades since it entered human populations , it has accumulated extensive sequence variation leading to the development of different subtypes and recombinant forms [1] . Although the enzymatic targets of therapy are among the most conserved parts of the HIV-1 genome , these too can develop marked variation , particularly in the setting of selective antiretroviral drug pressure . Indeed , it is not uncommon for drug therapy to select for protease and reverse transcriptase ( RT ) variants containing substitutions at more than 10% of their amino acids [2] . However , despite this high degree of mutation , the spectrum of possible virus variants appears to be limited by patterns of amino acid covariation . In 2003 , we published two studies that examined the extent of covariation among RT and protease residues in the presence and absence of antiretroviral therapy [3 , 4] . Despite the relatively large size of the datasets in these studies—2 , 244 protease sequences and 1 , 210 RT sequences—there were insufficient data to examine patterns of covariation of different mutations at the same position . As more sequence data have become available , we are now analyzing covariation among mutations ( rather than positions ) in protease and RT . This expanded analysis uses a highly specific measure of covariation , the Jaccard similarity coefficient , and a multidimensional scaling based on this coefficient . In addition , we examine the relationship between conditional probabilities associated with a mutation pair and the order in which those mutations develop in viruses for which longitudinal sequence data are available .
Protease sequences from 3 , 982 protease inhibitor ( PI ) –naive individuals and from 3 , 475 PI-experienced individuals were available for analysis . The PI-experienced individuals had received a median of 1 PI ( interquartile range , 1–3 ) . Jaccard similarity coefficients and their standardized Z scores were calculated for all pairs of mutations at different positions present three or more times among the sequences from PI-naive and PI-experienced individuals . Among 19 , 203 pairs of mutations from the PI-experienced individuals , 161 pairs were significantly associated after adjusting for multiple comparisons by controlling the family-wise error rate at <0 . 01 . Of these 161 pairs , 92 ( 57% ) were positively associated ( Z > 5 . 1 , unadjusted p < 4 . 4 × 10−7 ) and 69 ( 43% ) were negatively associated ( Z < −5 . 0 , unadjusted p < 4 . 8 × 10−7 ) . Table 1 shows the Jaccard similarity coefficients and conditional probabilities of the 40 strongest positively associated protease mutation pairs and the ten strongest negatively associated protease mutation pairs . Table S1 shows the complete list of 161 statistically significant mutation pairs . For the positively associated mutation pairs , Table 1 also contains two columns with data on the temporal order in which correlated mutations occurred in sequences with both mutations from persons in which an earlier sequence was available that contained only one of the two mutations . For example , the first row shows that among persons with both I54V and V82A in whom an earlier sequence contained only one of these two mutations was available , I54V occurred first in nine ( 26% ) of 34 people , and V82A occurred first in 25 ( 74% ) of 34 people ( p < 0 . 01 ) . In contrast , the fourth row shows that among persons with both A71V and L90M , each of the mutations was as likely to occur first ( 26 of 51 versus 25 of 51; p = NS ) . Figure S1 plots the relationship between the log of the ratio of the conditional probability of two mutations versus the log of the ratio in which two mutations develop , indicating that the conditional dependence between mutations is highly correlated with the order in which the mutations develop when they occur together ( r2 = 0 . 56 , p < 0 . 001 ) . Among the 18 positively associated pairs in Table 1 containing a major and an accessory PI-resistance mutation ( as defined in Methods ) , the accessory mutation appeared first more often in 12 of the 18 pairs . There were several striking patterns of temporal association among these 18 pairs of correlated major and accessory mutations . The major mutation L90M preceded the accessory mutation G73S in 31 of 34 persons for whom temporal data were available . In contrast , the accessory mutation L63P preceded L90M in 160 of 172 persons , and the accessory mutations L10I and A71V preceded the major mutation I84V in 51 of 59 and 35 of 38 persons , respectively . The Jaccard dissimilarity coefficients associated with 595 pairs of 35 mutations were used for a multidimensional scaling . The mutations included in this analysis were the 22 positively associated mutations in Table 1 and 13 additional clinically relevant PI-resistance mutations ( L10F , V32I , L33F , I47V , I50V/L , F53L , I54L/M , Q58E , L76V , V82T , and N88S ) . Figure 1 plots the mutations along axes representing the first two principal components . The first principal component accounted for 10% of the total inertia and separates the nelfinavir-resistance mutations D30N and N88D from the main group of PI-resistance mutations . The second principal component accounted for 7% of the total inertia and separates V82A-associated mutations ( I54V , L24I , and M46L ) from L90M-associated mutations ( M46I , G73S , and I84V ) . Finally , the lower-left part of the figure contains a cluster with seven of the 11 mutations recently reported to be associated with phenotypic and clinical resistance to the newest PI , darunavir ( V32I , L33F , I47V , I50V , I54L/M , and L76V ) . At several positions , there was sufficient data to contrast covariation patterns for different mutations . For example , M46I/L were each significantly associated with L10I , L24I , V32I , L33F , I54V , V82A , and L90M . However , M46I was uniquely associated with F53L , G73S/T , V82F/T , I84V , and N88S . I54V was significantly associated with L10F , L24I , L33F , M46I/L , G48V , F53L , V82A/F/T , I84V , and L90M . In contrast , I54L/M were significantly associated only with L33F , M46I , I47V , I84V , and L90M . N88D was positively associated with D30N and negatively associated with M46I , whereas N88S was negatively associated with D30N and positively associated with M46I . Of note , the divergent associations of different mutations at positions 46 and 88 have previously been reported by Hoffman and coworkers [5] . Among 7 , 131 pairs of mutations in sequences from PI-naive persons , 65 pairs were significantly associated ( family-wise error rate < 0 . 01; Table S2 ) . All but three of the positive associations among PI-naive persons were weaker ( i . e . , had a lower Z score ) than the positive associations among treated persons in Table 1 . RT sequences from 2 , 601 RT inhibitor–naive and from 5 , 188 RT inhibitor–experienced individuals were available for analysis . The RT inhibitor experienced individuals had received a median of three nucleoside RT inhibitors ( NRTIs; interquartile range , 2–4 ) and zero nonnucleoside RT inhibitors ( NNRTIs; interquartile range , 0–1 ) . Jaccard similarity coefficients and their standardized Z scores were calculated for all pairs of RT mutations at different positions present three or more times among the sequences from RT inhibitor–experienced and –naive persons . Among 65 , 624 pairs of mutations from the RT inhibitor–experienced persons , 327 pairs were significantly associated after adjusting for multiple comparisons by controlling the family-wise error rate at <0 . 01 . Of these 327 pairs , 213 ( 65% ) were positively associated ( Z > 5 . 2 , unadjusted p < 2 × 10−7 ) and 114 ( 35% ) were negatively associated ( Z < −5 . 0 , unadjusted p < 5 × 10−7 ) . Table 2 shows the Jaccard similarity coefficients and conditional probabilities of the 40 strongest positively associated RT mutation pairs and the ten strongest negatively associated RT mutation pairs . Table S3 shows the complete list of 327 statistically significant RT mutation pairs . Positively associated mutation pairs consisted primarily of Type I or II thymidine analog mutations ( TAMs; as defined in Methods ) ; accessory NRTI mutations that occurred in combination with Type I or II TAMs ( K43E , E44D , V118I , H208Y , D218E ) ; and Q151M-associated mutations ( V75I , F77L , F116Y ) . Among the top 40 associated mutation pairs , there were only three positive associations between Type I and II TAMs ( M41L , L210W , and T215Y with D67N ) . The strongest significant association between an NRTI and an NNRTI mutation was between L74V and Y181C ( J = 0 . 17 , Z = 8 . 9 , unadjusted p < 1 × 10−11 ) . Of note , the associations between the five accessory mutations listed above and Type I and II TAMs have also previously recently been described by Svicher and coworkers [6] and Cozzi-Lepri and coworkers in independent datasets [7] . The conditional probabilities and the temporal data columns show that each of the accessory NRTI mutations consistently follows the Type I or II TAMs . Among 12 pairs with a TAM and an accessory mutation , the TAM occurred first more often in all 12 pairs and was preceded by the accessory mutation in only 6% of pairs . In addition to the five accessory mutations in Table 2 ( K43E , E44D , V118I , H208Y , and D218E ) , other NRTI mutations that consistently followed TAMs included the known treatment-selected mutations T69D and T69N . Figure S2 plots the relationship between the log of the ratio of the conditional probability of two mutations versus the log of the ratio in which two mutations develop , indicating that the conditional dependence between mutations is highly correlated with the order in which the mutations develop when they occur together ( r2 = 0 . 81 , p < 0 . 001 ) . The Jaccard dissimilarity coefficients associated with the 561 pairs of 34 mutations were used for a multidimensional scaling . The mutations included in this analysis were the 23 positively associated mutations in Table 2 and 11 additional clinically relevant NRTI-resistance mutations ( K65R , A62V , T69ins , L74I/V , V75M , Y115F , M184V , and K219R/E/N ) . Figure 2 plots the mutations along axes representing the first two principal components . The first principal component accounts for 13% of the total inertia and separates the TAMs from the Q151M-associated mutations , whereas the second principal component accounts for 9% of the total inertia and separates the Type I and Type II TAMs . A62V , K65R , and Y115F are mutations that cluster with Q151M but may also occur with Type II ( but not Type I ) TAMs . D67N is a Type II TAM that can also occur with Type I TAMs , and it therefore occurs between Type I TAMs and Type II TAMs in terms of the second principal component . The non-TAM mutations , M184V and L74V , demonstrated no clustering with other NRTI-associated mutations . At several positions , there was sufficient data to contrast covariation patterns for different mutations ( Table 2 , Figure 2 , and Table S3 ) . The Type I TAM , T215Y , clustered with other Type I TAMs , whereas the Type II TAM , T215F , clustered with other Type II TAMs . K219Q/E were Type II TAMs that cluster with other Type II TAMs . In contrast , two less common mutations at this position ( K219N/R ) were positively associated with Type I TAMs . T69D was associated with both Type I and Type II TAMs , whereas T69N was associated only with Type II TAMs . L74V was associated with the NNRTI-resistance mutations L100I , K103N , and Y181C , whereas L74I was associated with M41L . V75I was associated with Q151M-associated mutations , whereas V75M was positively associated with the Type I TAMs . Among 19 , 431 pairs of mutations in sequences from RT inhibitor–naive persons , 41 pairs were significantly associated ( family-wise error rate <0 . 01; Table S4 ) . However , all of the positive associations among RT inhibitor–naive persons were weaker ( i . e . , had a lower Z score ) than the positive associations among treated persons in Table 2 .
In this analysis of amino acid covariation in protease and RT sequences from more than 7 , 000 persons infected with HIV-1 subtype B viruses , we confirmed several previously reported patterns of amino acid covariation and identified many new patterns of covariation . Multidimensional scaling further organized many of the correlations into clusters of co-occurring mutations . RT covariation was dominated by the distinct clustering of the TAMs and Q151M-associated mutations , and by the separation of the Type I and Type II TAMs . Protease covariation was dominated by the clustering of nelfinavir-associated mutations ( D30N and N88D ) , two main groups of PI-resistance mutations associated either with V82A or L90M , and a newly identified cluster of the mutations V32I , L33F , I47V , I50V , I54L/M , and L76V . This new cluster of mutations is associated with decreased susceptibility to all PIs , including the salvage therapy PIs amprenavir and lopinavir and the recently approved PI darunavir . Although none of the sequences in this study were from patients who received darunavir , this drug is highly similar to amprenavir and is affected by the same PI-resistance mutations . Previous studies of HIV-1 covariation have used either the Pearson correlation for binomial random variables or mutual information [3–6 , 8–10] . The correlation coefficient is overly sensitive to rare pairs of mutations because its statistical significance is based on a departure from equality between the diagonal and off-diagonal products of a 2 × 2 contingency table . In contrast , mutual information is insensitive to rare pairs of mutations , approaching a high level only for commonly occurring pairs of mutations . We therefore used the Jaccard similarity coefficient , which uses only those sequences in which at least one of a pair of mutations is present , and we assessed the significance of this coefficient using a distribution based on the underlying data . We also used a conservative correction for multiple comparisons ( Holm's method ) because our analysis was not designed to identify all covarying mutations but only those with the strongest association . Without a correction for multiple comparisons , 753 pairs of protease mutations from PI-experienced persons and 2 , 061 pairs of RTI mutations from RTI-experienced persons had a significant Jaccard similarity coefficient at a p-value of 0 . 01 but with the Holm's correction , only 161 pairs of protease mutations and 327 pairs of RTI mutations were significantly associated using a family-wise error rate of 0 . 01 . Covariation between two mutations may result from the shared inheritance of the mutations from a founder virus , from a shared evolutionary pressure ( e . g . , an antiretroviral drug ) that independently selects for each mutation , or from a functional dependency between the mutations . In our analysis , covariation was unlikely to result from shared inheritance because the most strongly covarying mutations occurred solely among treated HIV-1 isolates , consistent with the repeated selection of the correlated mutations in many different isolates as a result of selective drug pressure rather than the inheritance of the correlated mutations from a small number of ancestral viruses . However , the possibility that many of the covarying residues resulted from similar selective pressures rather than from functional dependency cannot be excluded . For example , it is possible that some pairs of covarying protease amino acids result from the selective pressure of the same PI or possibly pair of PIs . Shared selective pressure is a possible explanation for why covarying mutations are not necessarily close to one another in tertiary structures ( Figure S3 ) [4] . An analysis of covariation that controls for treatment history would be better able to distinguish functional dependency from shared selective pressure . However , for most PIs and NRTI combinations , insufficient data are available for such an analysis . Identifying similar patterns of covariation in one or more independent lineages ( e . g . , other non-B subtypes ) would also provide additional independent evidence for functional dependency . Our examination of conditional dependency between mutation pairs , the temporal order in which mutations occur , and the relationship between these two types of data provided new insights into the evolution of protease and RT in persons receiving antiretroviral therapy . A strong positive relationship between the conditional dependency ratio of two mutations and the order in which the mutations occur would represent the most parsimonious mechanism for HIV-1 to develop multiple mutations ( i . e . , the mutation that occurs more often in a pair of mutations would be on average more likely to occur first ) . Nonetheless , we found that the positive relationship between conditional dependency and the order of mutation occurrence was stronger for covarying RT ( r2 = 0 . 81 ) compared with protease ( r2 = 0 . 56 ) mutation pairs . This suggests that the number of mutational steps required to develop multiple PI-resistance mutations may be greater on average than that required for developing the same number of multiple NRTI-resistance mutations . We also found that accessory NRTI-resistance mutations nearly always followed primary NRTI-resistance mutations ( particularly the TAMs ) . In contrast , the commonly recognized accessory PI-resistance mutations were as likely to precede as to follow major PI-resistance mutations . This frequent precedence of accessory PI-resistance mutations results in part from the fact that many of the accessory PI-resistance mutations are polymorphic and therefore present prior to the start of therapy . However , this alone does not explain the marked dependency of some major mutations on polymorphic accessory PI-resistance mutations that occur only at low levels in untreated persons . The strong positive relationship between conditional probabilities and temporal data that we describe support the validity of previous research , which used cross-sectional data to infer mutational pathways [11] and causality [12 , 13] . Our results also suggest that there is a complex process underlying the order in which major and accessory PI-resistance mutations develop during PI therapy , and that the designation of major PI-resistance mutations as primary and accessory PI-resistance mutations as secondary often refers only to their roles in causing resistance and not to the order in which they develop .
Sequences included HIV-1 subtype B RT and protease sequences from published studies in the Stanford HIV Drug Resistance Database ( http://hivdb . stanford . edu ) [14] . For patients with more than one sequence , only the latest sequence obtained while receiving treatment was analyzed . For each gene , separate analyses were done for the sequences from treatment-experienced and treatment-naive individuals . RT positions 1–240 and protease positions 1–99 were analyzed . Mutations were defined as differences from the consensus wild-type subtype B amino acid reference sequence ( http://hivdb . stanford . edu/pages/asi/releaseNotes/index . html ) . For each pair of mutations ( X , Y ) , the numbers of sequences containing both mutations ( X and Y ) , only one mutation ( X or Y ) , or neither mutation ( not X , not Y ) were counted and used to populate a contingency table . Sequences containing mixtures at either of the two positions were excluded from analysis of that pair of positions . Antiretroviral treatment–selected mutations were defined based on the results of a previous study , as mutations that were significantly more common in treated than untreated persons after adjusting for multiple comparisons [15] . PI-selected mutations included L10I/V/F/R , V11I , K20R/M/I/T , L23I , L24I , D30N , V32I , L33F/I , E34Q , E35G , M36I/V , K43T , M46I/L/V , G48V/M , I50V/L , F53L , I54V/M/L/T/A/S , K55R , Q58E , L63P , I66F , C67F , A71V/T/I , V72L , G73S/T/C/A , T74A/P/S , L76V , V77I , V82A/T/F/S/L/M , I84V/A/C , I85V , N88D/S/T/G , L89V , L90M , T91S , Q92R/K , I93L , and C95F . Several PI-resistance mutations—particularly those that occur in the substrate cleft or that have a major impact on drug susceptibility—are considered major PI-resistance mutations [2 , 16] . For the purposes of this study , we defined mutations at positions 24 , 30 , 32 , 46 , 47 , 48 , 50 , 53 , 54 , 76 , 82 , 84 , 88 , and 90 as being major PI-resistance mutations . Several PI-resistance mutations—including several that are polymorphic in untreated persons—are commonly considered accessory drug resistance mutations that either compensate for the decreased replication associated with many of the major mutations or that reduce drug susceptibility further when present with a major mutation . Mutations at positions 10 , 20 , 33 , 36 , 58 , 63 , 71 , 73 , 74 , 77 , and 93 are usually considered to be accessory mutations . Little attention has been given to the remaining PI-selected mutations , and for the purposes of this paper , we leave them unclassified with respect to the designations major and accessory . NRTI-selected mutations included T39A , M41L , K43E/Q/N , E44D/A , A62V , K65R , D67N/G/E , T69D/N/S/insertion , K70R , L74V/I , V75I/M/T/A , F77L , V90I , K104N , Y115F , F116Y , V118I , Q151M , M184V/I , E203K , H208Y , L210W , T215Y/F/D/C/E/S/I/V , D218E , K219Q/E/N/R , H221Y , K223Q , and L228H/R . These mutations included the Type I TAMs M41L , L210W , and T215Y , and the Type II TAMs D67N , K70R , T215F , and K219Q/E [7] . Recently described accessory NRTI mutations included T39A , K43E/Q/N , E44D/A , V118I , E203K , H208 , D218E , H221Y , K223Q , and L228H/R [3 , 6 , 17] . Q151M-associated mutations included A62V , V75I , F77L , F116Y , and Q151M [18 , 19] . NNRTI-selected mutations included A98G , L100I , K101E/P/N/H , K103N/S , V106A/M , V108I , V179D/E , Y181C/I/V , Y188L/C/H , G190A/S/E/Q , P225H , F227L , M230L , P236L , and K238T . We used the Jaccard similarity coefficient ( J ) to assess covariation among protease and RT mutations . For a given pair of mutations X and Y , the Jaccard similarity coefficient is calculated as J = NXY / ( NXY + NX0 + N0Y ) where NXY represents the number of sequences containing X and Y , NX0 represents the number of sequences containing X but not Y , and N0Y represents the number of sequences containing Y but not X . This coefficient represents the probability of both mutations occurring together when either mutation occurs and , therefore , does not inflate the correlation between two mutations that may appear correlated by other measures when both mutations are nearly always absent . To test whether observed Jaccard similarity coefficients were statistically significant , the expected value of the Jaccard similarity coefficients ( JRAND ) and its standard error ( JSE ) assuming two mutations ( X and Y ) occur independently were calculated for each pair of mutations . JRAND was calculated as the mean Jaccard similarity coefficient after 2 , 000 random rearrangements of the X or Y vector ( containing 0 or 1 for presence or absence of a mutation , respectively ) . JSE was calculated using a jackknifed procedure , which removed one sequence at a time , repeatedly for each sequence . The standardized score Z , Z = ( J − JRAND ) / JSE , indicates a significant positive association ( Z > 2 . 56 ) or a significant negative association ( Z < −2 . 56 ) at an unadjusted p < 0 . 01 . Holm's method was used to control the family-wise error rate for multiple hypothesis testing [20] . The p-values of observed Jaccard similarity coefficients for all pairs of mutations were ranked in descending order . Starting from the smallest p ( rank r = n , where n is the number of pairs ) , we compared each p of rank r with a significance cutoff of 0 . 01 / r as long as pr ≤ 0 . 01 / r . All p-values from pr…pn were considered to be statistically significant . To deal with contingency tables containing 0 for NXY ( potentially leading to Z scores of −∞ ) , we generated a conservative nonzero approximation of JSE using the following procedure . Given a dataset of n sequences , x with mutation X and y with mutation Y , we computed the probability of both mutations ( PXY ) , mutation X but not Y ( PX0 ) , mutation Y but not X ( P0Y ) , and neither mutation ( P00 ) under the null hypothesis of independence by PXY = ( x / n ) × ( y / n ) , PX0 = ( x / n ) × ( y / n ) / n , P0Y = ( n − x ) / n × ( y / n ) and P00 = 1 − PXY − PX0 − P0Y . These probabilities were used to create 200 two-by-two contingency tables with cells containing randomly distributed numbers adding up to 20 , 000 based on the null hypothesis probabilities of independence . Given the matrix of dissimilarity coefficients ( 1 − Jaccard similarity coefficient ) for a list of mutations ( X1 , X2 , . . . , Xn ) , multidimensional scaling was used to construct points in 2-D space such that the Euclidean distances between these points approximate the entries in the dissimilarity matrix [21] . For a given k , it computes points X1 , X2 , … , Xn in 2-D space such that S = is minimized where dist ( Xi , Xj ) is the Euclidean distance between Xi and Xj and dij is the dissimilarity between Xi and Xj in the matrix D . This was performed using the R function cmdscale ( classical multidimensional scaling ) . Multidimensional scaling captures the inertia in a dataset in terms of a set of variables ( or principal components ) that define a projection that encapsulates the maximum amount of inertia in a dataset and is orthogonal ( and therefore uncorrelated ) to the previous principal component . Using the first and second principal components , we summarized the relationship among mutations in a graphical model , placing pairs of mutations with low Jaccard dissimilarity coefficients close together and mutations with high Jaccard dissimilarity coefficients far apart .
The 11 , 355 GenBank ( http://www . ncbi . nlm . nih . gov/Genbank ) accession numbers of the sequences used in this study are provided in Text S1 .
|
The identification of which mutations in a protein covary has played a major role in both structural and evolutionary biology . Covariation analysis has been used to help predict unsolved protein structures and to better understand the functions of proteins with known structures . The large number of published genetic sequences of the targets of HIV-1 therapy has provided an unprecedented opportunity to identify dependencies among mutations in these proteins that can be exploited to design inhibitors that have high genetic barriers to resistance . In our analysis , we identified many pairs of covarying drug-resistance mutations in HIV-1 protease and reverse transcriptase and organized them into clusters of mutations that often develop in a predictable order . Inhibitors that are active against early drug-resistant mutants are likely to be less prone to the development of resistance , whereas inhibitors that are active against fully evolved clusters of mutations may be useful drugs for salvage therapy .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods",
"Supporting",
"Information"
] |
[
"mathematics",
"viruses",
"infectious",
"diseases",
"computational",
"biology"
] |
2007
|
HIV-1 Subtype B Protease and Reverse Transcriptase Amino Acid Covariation
|
In human B cells infected with Epstein-Barr virus ( EBV ) , latency-associated virus gene products inhibit expression of the pro-apoptotic Bcl-2-family member Bim and enhance cell survival . This involves the activities of the EBV nuclear proteins EBNA3A and EBNA3C and appears to be predominantly directed at regulating Bim mRNA synthesis , although post-transcriptional regulation of Bim has been reported . Here we show that protein and RNA stability make little or no contribution to the EBV-associated repression of Bim in latently infected B cells . However , treatment of cells with inhibitors of histone deacetylase ( HDAC ) and DNA methyltransferase ( DNMT ) enzymes indicated that epigenetic mechanisms are involved in the down-regulation of Bim . This was initially confirmed by chromatin immunoprecipitation analysis of histone acetylation levels on the Bim promoter . Consistent with this , methylation-specific PCR ( MSP ) and bisulphite sequencing of regions within the large CpG island located at the 5′ end of Bim revealed significant methylation of CpG dinucleotides in all EBV-positive , but not EBV-negative B cells examined . Genomic DNA samples exhibiting methylation of the Bim promoter included extracts from a series of explanted EBV-positive Burkitt's lymphoma ( BL ) biopsies . Subsequent analyses of the histone modification H3K27-Me3 ( trimethylation of histone H3 lysine 27 ) and CpG methylation at loci throughout the Bim promoter suggest that in EBV-positive B cells repression of Bim is initially associated with this repressive epigenetic histone mark gradually followed by DNA methylation at CpG dinucleotides . We conclude that latent EBV initiates a chain of events that leads to epigenetic repression of the tumour suppressor gene Bim in infected B cells and their progeny . This reprogramming of B cells could have important implications for our understanding of EBV persistence and the pathogenesis of EBV-associated disease , in particular BL .
EBV is a B lymphotropic gammaherpes virus that asymptomatically and persistently infects >90% of humans; it occasionally causes infectious mononucleosis in adolescents and in rare instances is associated with the development of several different types of B cell lymphoma and various epithelial tumours [1] . EBV can also produce B cell lymphomas in non-human primates and induce the continuous proliferation ( ‘immortalisation’ ) of primary human B cells in vitro . The lymphoblastoid cell lines ( LCLs ) that are produced in culture carry the viral genome as extra-chromosomal episomes and express only nine ‘latent’ EBV proteins . There are six nuclear antigens ( EBNAs 1 , 2 , 3A , 3B , 3C & LP ) and three membrane associated proteins ( LMP1 , LMP2A & 2B ) ; together with several RNA species , these factors activate the quiescent B cells from G0 into the cell cycle , initiate and sustain proliferation and maintain the viral episome in its extra-chromosomal state . Using recombinant viruses it has been shown that of all the transcripts expressed in LCLs only EBNA1 , 2 , 3A , 3C , -LP and LMP1 are essential for the efficient transformation of primary B cells into LCLs ( reviewed in [2] ) . Current data on the normal asymptomatic persistence of EBV in humans are consistent with the viral genome residing long-term in a resting memory B cell population . However , in order to establish persistence , EBV infects non-dividing naïve B cells and drives these to proliferate as activated B blasts that express all the viral proteins found in LCLs . The transient expansion of an infected B blast population is accompanied by their differentiation , probably in germinal centres , to become centroblasts and centrocytes and finally resting memory B cells . The precise series of events that the EBV-positive B cells undergo to reach the memory compartment is not yet known . However , it appears to involve the regulated shut-down and silencing of latent EBV gene expression from an initial state called latency III or the growth programme ( as found in LCLs ) , via latency II ( also known as the default programme ) , until in quiescent memory B cells no EBV proteins can be detected in a state called latency 0 ( or the latency programme ) ( reviewed in [3] ) . Using recombinant viruses established with a bacterial artificial chromosome ( BAC ) system , we recently showed that EBNA3A and EBNA3C are both necessary for repression of the cellular gene Bim [4] . The Bim protein ( Bcl-2 interacting mediator of cell death ) is a pro-apoptotic BH3-only , Bcl-2-family member that appears to be a uniquely important tumour suppressor in the development of B and T lymphocytes . It regulates apoptosis during lymphocyte development by binding and inactivating pro-survival members of the Bcl-2 family and binding and activating the pro-apoptotic family member Bax ( reviewed in [5] , [6] ) . The role of Bim in lymphomagenesis came sharply into focus when it was discovered that in Eμ-Myc transgenic mice constitutively expressing Myc in B cells , loss of even a single Bim allele significantly accelerated lymphoma development and revealed Bim as a haploinsufficient tumor suppressor [7] . Deregulation of Myc through reciprocal chromosome translocations that put the gene under the influence of immunoglobulin locus control elements is a hallmark of all BLs ( reviewed in [8] , [9] ) . The importance of Bim in a cell carrying a deregulated Myc became apparent when it was discovered that under these conditions combined activation of both the ARF/p53 pathway and Bim leads to apoptosis [10] . However , when Myc is mutated or either the activation of ARF/p53 or Bim is impaired , the result is B lymphomagenesis [10] , [11] . The clear implication is that if EBV inhibits an increase in Bim expression when wild-type Myc is deregulated by translocation , this could be a mechanism through which EBV directly contributes to the development of BL . Since during latency III , EBNA2 constitutively activates Myc , this repression of Bim expression is probably critical for the immortalisation of B cells by EBV , persistence in vivo and perhaps the development of the endemic EBV-positive form of BL [4] , [12] – it is therefore central to EBV biology . However the details of how Bim levels are modulated by EBV is a controversial subject since it has been reported that EBV can alter both Bim gene expression and Bim protein stability [4] , [13] . Here the molecular mechanism by which EBV regulates the amount of Bim has been explored further and this has revealed that heritable , epigenetic modifications in the 5′ regulatory region of Bim play a major role in determining the level of Bim protein expressed in EBV infected B cells .
This study was conducted according to the principles expressed in the Declaration of Helsinki . The samples for this study were obtained from the archives of the VU University medical centre . These were collected during 1996–2007 as part of collaborative studies in Malawi and Uganda on the diagnosis of Epstein-Barr virus associated malignancies . Written , informed consent was obtained from the guardians of study participants at the time of collection . All B cell lines were cultured in RPMI-1640 medium ( Invitrogen ) supplemented with 10% fetal calf serum , penicillin , streptomycin , 1 mM sodium pyruvate ( Sigma ) and 50 µM α-thioglycerol ( Sigma ) . 100 µg/ml hygromycin B ( Roche ) was added to cultures of BL containing recombinant hygromycin- resistant EBV ( BL31 WT , BL31 E2KO ) , as described in [4] . All other cell lines used in this study were described previously in [14] , [15] , [16] . 24 hours before any experimental treatment , cells were seeded at a density of 2 . 5×105 cells/ml . MG-132 ( Calbiochem ) was used at a final concentration of 5 µM , 10 µM or 15 µM . Actinomycin D ( Sigma ) was used at a final concentration of 5 µg/ml . Trichostatin A ( TSA ) , 5-azacytidine ( AZA ) and sodium butyrate were all purchased from Sigma and were used at concentrations of 500 nM , 5 µM and 2 . 5 mM respectively . Western immunoblotting was performed as described previously [4] , [15] , [17] . Primary antibodies used for western blot probing were: rabbit polyclonal anti-Bim/BOD ( Stressgen , AAP-330 ) , mouse monoclonal anti-p21WAF1 ( SX118 , kind gift from Prof . Lu Xin , Ludwig Institute , Oxford ) , mouse monoclonal anti-γ-tubulin ( Sigma , T6557 ) , mouse monoclonal anti-BZLF1 ( kind gift from Prof . Paul Farrell , Imperial College London ) , sheep polyclonal anti-EBNA3A ( Exalpha , USA ) , mouse monoclonal anti-EBNA3C ( A10 , kind gift from Prof . Martin Rowe , University of Birmingham ) . For Q RT-PCR , RNA was extracted from approximately 5×106 cells for each cell line using the RNeasy mini kit from Qiagen and following the manufacturer's instructions . To quantify BimEL mRNA after HDAC and DNMT inhibitors treatment , 25 ng of RNA were used for each reaction with one-step QuantiTect SYBR green RT-PCR kit from Qiagen and ABI PRISM 7700 Sequence Detector , according to the manufacturer's instructions . BimEL primers were GCTGTCTCGATCCTCCAGTG and GTTAAACTCGTCTCCAATACG as described previously [4] . GAPDH was used to normalize BimEL levels . GAPDH primers were purchased from Qiagen ( Hs_GAPDH_2_SG; QT01192646 ) . The PCR conditions were 30 min at 50°C , 15 min at 95°C and then 30 sec at 95°C , 30 sec at 58°C and 30 sec at 72°C for 40 cycles . The standard curve method was used . Standards were a mix of all RNAs used and five 10-fold serial dilutions of this . For quantification of mRNA after actinomycin D treatment , 5 ng of RNA were used for each reaction with one-step QuantiTect SYBR green RT-PCR kit from Qiagen and ABI PRISM 7900 Sequence Detector , according to the manufacturers' instructions . The same primers as above were used for BimEL . For Myc the primers were AGCTGCTTAGACGCTGGATTT and GAGGTCATAGTTCCTGTTGGTGAA and for β-Actin they were GATGGAGTTGAAGGTAGTTTCGTG and GCGGGAAATCGTGCGTGACATT [18] . Reaction conditions were the same as above . The calculated errors in the graphs are the standard deviations from three replicate Q RT-PCR reactions for each mRNA . Chromatin Immunoprecipitation Assay Kit from Upstate ( 17-295 ) was used , according to the manufacturer's protocol . To obtain sheared chromatin with DNA of 200 bp–1000 bp in length , extracted chromatin from 1×106 cells per ChIP was sonicated in 200 µl lysis buffer for four 20 sec sonication rounds , using a Heat Systems Sonicator Ultrasonic Processor XL at 10% intensity . The antibodies for the acetylated histones IPs were rabbit polyclonal IgG anti-acetyl-Histone H3 ( Upstate; 06-599 ) , rabbit polyclonal IgG anti-acetyl-Histone H4 ( Upstate; 06-866 ) and as a negative control Rabbit IgG serum ( Upstate; PP64 ) . Precipitated DNA was assayed by quantitative PCR ( Q-PCR ) using Qiagen's QuantiTect SYBR green PCR kit . 2% of input was compared to the IP sample and the values from the IgG negative control were subtracted as background . Primers for Bim were CTGGTCTGCAGTTTGTTGGA and GGTGGCTGCAAGAATCAAGT . β-Actin was used as an independent control . β-Actin primers were TGCACTGTGCGGCGAAGC and TCGAGCCATAAAAGGCAA [19] . The PCR conditions were 15 mim at 95°C and then 20 sec at 95°C , 30 sec at 55°C , 30 sec at 72°C for 40 cycles , using ABI PRISM 7700 Sequence Detector . ChIP for histone H3 trimethylated at K27 was performed as described above , using anti-H3K27-Me3 antibody from Upstate ( 17-625 ) and as a negative control Rabbit IgG serum ( Upstate; PP64 ) . The primers used to quantify precipitated DNA were: pair A: TTTAGAAAGAATCTTGGCAGTCAACTCCTC and CAATGGCTGGTGAAAAGGAGGGTTT , pair B: GAAGGACCAGGGAGGAAGGACCAAG and TGACACCTAGCCCAGTGGAAACCCC , pair C: CGAGCGGGAAAAAAGGTTTGGTTCA and TAGGCTCCCACTTCCTTCTCCCAGT , pair D: AAGAGCAAAGTTCGTCCGCGGTAGG and TATTTCGCTGCAAGAGGGAAAAGGCAC , pair E: GACCCTCAGAGGGAGGAGAGCTCAAA and GCCCTGAGTTTCTAAGCCGCTCTGG , pair F: CGCCAGCAGGCAGAGTTAC and CAGGCTCGGACAGGTAAAGG . The cycling conditions were 2 min at 50°C , 10 min at 95°C and then 15 sec at 95°C , 1 min at 60°C for 40 cycles , using ABI 7900 384-well Real-Time PCR machine . The calculated errors in all graphs presenting ChIP data are the standard deviations from three replicate Q-PCR reactions for precipitated chromatin , input chromatin and background ( chromatin precipitated with non-specific rabbit IgG ) . For the methylated DNA control , in vitro methylated Jurkat DNA was used ( Active Motif , 102163 ) . B cells were isolated from a buffy coat residue purchased from the blood transfusion service , using the AutoMACS 3 sorting system from Millitenyi Biotec . CD19 positive selection , with the CD19 MicroBeads from the same manufacturer ( 130-050-301 ) , was performed according to their instructions . PBMC genomic DNA was extracted from blood of a healthy donor . Genomic DNA was extracted with the Qiagen DNeasy Kit for all the cells described . 200–500 ng of genomic DNA was converted by bisulphite using EZ DNA methylation kit by Zymo Research ( D5002 ) . For MSP the converted DNA was amplified using 5 sets of primers , specific for 5 different regions of the Bim promoter . The sequences were – For Set I: U2F: GTTTTGGGGTTTTGTAAGGAAAT; U2R: AAATAATAAAAATATTTTCCACACC; M2F: TTTTGGGGTTTTGTAAGGAAAC; M2R: TCAAATAATAAAAATATTTTCCGCG . For Set II: U3F: GGTTTAGTTTTGGGAGGATTTTTT; U3R: ACCAAAAATAACAAATTCATACATC; M3F: GTTTAGTTTCGGGAGGATTTTTC; M3R: ACCAAAAATAACGAATTCATACGTC . For Set III: U4F: TTTTTGAGTTTATTTAGTTGTGTTTATGT; U4R: AAAACCTTAAATTCCCAAAATCACT; M4F: CGAGTTTATTTAGTCGTGTTTACGT; M4R: CTTAAATTCCCGAAATCGCT . For Set IV: U1F: TGTTTAAAATTTATTTGAAAATGTGT; U1R: ACAAATAAAAAAACATTATCCCACC; M1F: TCGTTTAAAATTTATTCGAAAATGC; M1R: AACAAATAAAAAAACGTTATCCCG . For Set V: U5F: GTAGATTTAGAGGATTGGAGAGGTG; U5R: ACCAATAAAAACAAAACAACTAAATTCA; M5F: GTAGATTTAGAGGATTGGAGAGGC; M5R: CAATAAAAACAAAACAACTAAATTCGA . For bisulphite sequencing primers AAAACCCCCAAAATTTACTAAACTC and GGGAATTTAAGGTTTTTTTTATTT were used to amplify a region of the Bim promoter ( see below ) , in order to sequence and study the methylation state of 36 CpG dinucleotides in this region . The amplified PCR products were cloned using the Invitrogen TA cloning kit . Plasmids containing the cloned sequences were amplified by rolling circle amplification ( Templiphi; GE healthcare ) and sequenced . Between 8 and 24 clones were sequenced for each cell line . Frozen tissue was collected in parallel with routine biopsy specimens from suspected BL-cases at the Uganda Cancer Institute by Dr . J . Orem ( Makarere University , Kampala , Uganda ) and were retrieved from the archives of Dept . Pathology , VU University medical centre in Amsterdam , The Netherlands . All biopsies were classified by routine cytology on paraffin-embedded formalin-fixed tissue and EBER-RISH using PNA probes ( DAKO , Glostrup , Denmark ) [20] . After isolating genomic DNA as described above , 2 µg of DNA was sonicated in 200 µl of H2O for five 20 sec sonication rounds , using a Heat Systems Sonicator Ultrasonic Processor XL at 10% intensity , to produce sheared DNA with length of 200–1000 bp . Methylated DNA was precipitated from 1 µg of sheared DNA using the MethylCollector kit ( Active Motif , 55002 ) according to the manufacturer's instructions . Precipitated DNA was quantified in exactly the same way as described for the ChIP for H3K27-Me3 , with the same primer pairs ( A–F ) , the same quantitative PCR conditions and the same method of analysis . As background control , precipitations with just the magnetic beads were performed , without addition of His-MBD2b . As additional controls , precipitations were performed using DNA from PBMCs ( unmethylated ) and Jurkat DNA ( fully methylated in vitro ) . The calculated errors in all graphs presenting methylated DNA precipitation data are the standard deviations from three replicate Q-PCR reactions for precipitated DNA and input DNA .
It has been reported that EBV reduces Bim expression by targeting it to the proteasome system for degradation [13] . Since EBV induces the phosphorylation of extracellular signal regulated kinase ( ERK ) and this activated form of ERK can phosphorylate Bim and mark it for ubiquitinylation , it was suggested that latent EBV enhances Bim turnover by proteasomes . However , we recently showed that inhibition of ERK phosphorylation in EBV-positive BL cells does not necessarily result in significantly higher levels of Bim , but that in EBV-positive cells Bim mRNA was always substantially reduced [4] . Nevertheless , this does not exclude the possibility that Bim protein is degraded by the proteasome system via another signalling pathway . In order to investigate this in more detail , EBV-negative BL31 cells , BL31 cells infected with a wild type ( B95 . 8 strain ) EBV-BAC ( BL31 WT ) and BL31 infected with an EBV-BAC with EBNA2 deleted ( BL31 E2KO ) were treated with the inhibitor of proteasome function , MG-132 . Western blots in Figure 1A show that the addition of increasing amounts of MG-132 had no significant effect on the expression of the major Bim isoform BimEL after eight hours . The same was true after twenty-four hours of treatment , however at this time there was considerable cell death ( data not shown ) . Under the same conditions p21WAF1 – a protein with a relatively short half-life because it is rapidly degraded by proteasomes [21] , [22] – accumulated , thus demonstrating that the MG-132 successfully inhibited proteosome-mediated proteolysis . It is unlikely therefore that enhanced protein degradation is the main cause of the lower levels of Bim seen in B cells containing latent EBV . Although latent EBV gene expression can alter ERK signalling and may therefore affect Bim by phosphorylation and ubiquitinylation , our data are consistent with EBV primarily inhibiting the de novo synthesis of Bim , rather than enhancing its turnover . Recently it was shown that Bim levels are also regulated by modulation of Bim mRNA stability [23] . It is therefore conceivable that EBV reduces the stability of Bim mRNA , leading to lower steady-state levels of Bim transcripts and protein in infected cells . To examine this possibility , the same BL31 cell lines were treated with the inhibitor of transcription actinomycin D for up to eight hours to block de novo synthesis of RNA . BimEL mRNA was then analysed by quantitative ( Q ) RT-PCR to assess whether in EBV infected cells BimEL mRNA was degraded more rapidly ( Figure 1B ) . Although , as expected , the starting levels of Bim mRNA were different in each cell line ( see histogram in Figure 1B ) , in each cell line the mRNA levels dropped , showing that transcription was successfully inhibited . Moreover , the rate of BimEL mRNA degradation was very similar in EBV infected and uninfected cells indicating that the presence of latent EBV does not significantly influence BimEL mRNA stability . In parallel , the stability of an mRNA known to have a short half-life ( Myc ) and one known to be very stable ( β-Actin ) were assessed ( Figure S1 ) . The data from these experiments using inhibitors of either proteosome enzymes or mRNA synthesis are therefore most consistent with the regulation of Bim expression occurring principally at the level of transcription as we previously suggested [4] . In the absence of evidence supporting an alternative mechanism for Bim regulation by EBV , we investigated whether latent EBV could influence transcription of Bim via the modification of local chromatin . There are several reports of Bim expression being regulated by modulation of transcription ( for examples see [24] , [25] , [26] ) and a striking feature of the gene is an unusually large CpG island of more than 6000 bp extending either side of the transcription initiation site ( http://genome . ucsc . edu/cgi-bin/hgGateway; Figure 2 ) ; this region could be subject to control by DNA methylation . Since both histone deacetylation and DNA methylation are associated with repressed , inactive chromatin , these were investigated initially using chemical inhibitors ( reviewed in [27] , [28] , [29] ) . To get an indication of whether the regulation of Bim transcription might involve local chromatin or DNA modifications BL31 , BL31 WT , and LCL-CH cells were treated with the HDAC inhibitor trichostatin A ( TSA ) , the DNMT inhibitor 5′ azacytidine ( AZA ) or with sodium butyrate – which is reported to suppress HDACs and might affect DNA methylation [30] , [31] . Cells were treated with one of the inhibitors and samples were taken after 24 and 48 hours . The levels of BimEL protein following these treatments can be seen in western blots ( Figure 3A ) . Exposure of EBV-positive BL31 WT cells to TSA resulted in a significant up-regulation of BimEL levels , but in uninfected BL31 cells the already high level remained unchanged . In the LCL cells up-regulation was more pronounced – even after 24 hours of treatment . AZA had more moderate effects on BimEL protein levels , with little change in the LCL , but significant up-regulation in BL31 WT . Exposure of both BL31 WT and the LCL to sodium butyrate produced the most substantial increase of BimEL . Two other EBV negative/positive “pairs” of BL cell lines , BL41/BL41 B95 . 8 and Ramos/Ramos-AW together with two BL-derived cell lines Eli and Akata 6 , were treated in a similar way and these showed the same trend of BimEL up-regulation as described above ( Figure S2A and B ) . Since both TSA and AZA have been shown to induce the lytic cycle in EBV infected cells [32] , [33] , western blots were performed to determine whether the EBV lytic switch protein BZLF1 was induced . No correlation between lytic cycle and Bim expression was detected in these cells ( Figure S2C ) . This is consistent with our previous observations [16] . Subsequently the amount of BimEL mRNA in cells harvested after similar treatments was assessed by real-time Q RT-PCR ( Figure 3B ) . This analysis was performed on samples taken 48 hours after the addition of the inhibitor . BimEL sequences were amplified using primers described previously [4] . Values for BimEL mRNA were normalized to values obtained for GAPDH primers . Quantification used standard curves for each pair of primers , always using the same standards . Similar trends of BimEL up-regulation were seen at the mRNA level ( Figure 3B ) as at the protein level ( Figure 3A ) , indicating that latent EBV is associated with repression of Bim transcription via mechanisms involving both histone deacetylation and DNA methylation . An exception was LCL-CH in which the up-regulation of Bim mRNA by 5-azacytidine was significantly greater than the amount of protein induced . We do not know the reason for this discrepancy , but it should be noted that all three inhibitors were very toxic in the cells used here , inducing a great deal of cell death and so making these experiments rather unsatisfactory . Nevertheless the results were each largely consistent with the hypothesis that EBV repression of Bim involves local chromatin modifications . Since Bim levels can be increased in EBV infected cells by treatment with HDAC inhibitors , it seems likely that EBV regulates Bim , at least in part , through the modulation of histone acetylation . If the effect is direct , this means that in the presence of EBV , histones associated with the Bim promoter should be less acetylated relative to uninfected cells . To directly assess occupancy of the Bim promoter by acetylated histones , chromatin immunoprecipitations ( ChIPs ) were performed using antibodies directed against acetylated histone H3 ( K9Ac , K14Ac ) and acetylated histone H4 ( K4Ac , K7Ac , K11Ac , K15Ac ) . The results are shown in Figure 4 . Real-time Q PCR was used to measure the abundance of a DNA fragment from the Bim promoter that was associated with immunoprecipitated histones . The primers used correspond to a region just downstream of the transcriptional start site ( shown in Figure 2 ) . Acetylated histone occupancy of the β-Actin promoter was measured as a control . Values for each immunoprecipitation were normalized to 2% of input DNA , and in each case the values obtained from the control sample incubated with rabbit pre-immune IgG were subtracted as background . Two EBV-negative BL lines were used , BL31 and BL41 , and their EBV-positive counterparts , BL31 WT and BL41 B95 . 8 . It is clear that EBV significantly decreases acetylated histone ( H3 and H4 ) occupancy in both EBV positive lines , relative to the uninfected ones . Histone acetylation on the β-Actin promoter was largely unaffected by EBV , indicating that the effect on Bim is specific . The data are consistent with EBV inhibiting Bim transcription by the direct modulation of biochemical marks on chromatin of the Bim promoter . DNA methylation is associated with inactive chromatin and the repression of transcription [27] , [28] , [29] , [34] . The observation that treatment of some EBV-infected B cells with the DNMT inhibitor 5′ azacytidine led to the de-repression of Bim , and the presence of a remarkably large CpG island spanning the 5′ regulatory region of Bim , led us to investigate the DNA methylation status of the Bim promoter region . In order to discover whether EBV directly influenced DNA methylation patterns on the promoter we used methylation specific PCR ( MSP ) and bisulphite sequencing directed at various sites upstream of the transcription initiation site ( see schematic in Figure 2 ) . For the initial analysis , twenty-six cell lines ( plus primary cells ) were investigated by MSP and of these , twenty were studied further using bisulphite sequencing to determine the DNA methylation state of Bim relative to their EBV status ( ie EBV-positive or -negative ) . The EBV-negative cells used were either primary cells [peripheral blood mononuclear cells ( PBMCs ) or purified CD19+ primary B cells] or EBV-negative BL cell lines . The EBV-positive cell lines were LCLs ( early passage or late passage ) , in vitro BL ‘converts’ ( that is BL lines established from EBV-negative tumours and then infected in vitro with EBV ) , or BL lines established from EBV-positive tumours . For the MSP analysis five sites were tested with primer sets ( I–V ) of two primer pairs specific for either methylated or unmethylated DNA . The regions of the Bim promoter amplified by the MSP primers are indicated in Figure 2 . The results are summarized in the matrix shown in Figure 5 . Cell lines that gave a band with any of the methylation-specific primers were scored as methylated ( M ) . Cell lines that only gave a band with primers for unmethylated DNA , and none with primers for methylated DNA were scored as unmethylated ( U ) . These analyses were generally performed at least three times and an example of representative primary data for twenty cell lines plus primary cells using primer set III can be seen in Figure S3 . In PBMCs and purified primary B cells Bim was always completely unmethylated in these assays – as would usually be expected of a CpG-island associated with an active gene and protected by transcription factors from DNA methylation [34] . Similarly , all the established EBV-negative BL lines ( that had all been cultured on and off for decades ) were found to be unmethylated at these sites , with the exception of Ramos , which gave a faint band for methylation only with primer set III . In contrast , all late passage LCLs that have also been cultured for many years , showed methylation at nearly all these sites . One of the early passage LCLs ( LCL-CH , recently established in our lab by the infection of CD19+ B cells with B95 . 8 strain of EBV [15] and only grown continuously in culture for 2–3 months before being frozen ) was found to be unmethylated . EBV-positive in vitro BL converts gave a mixed picture , with two being methylated and one unmethylated . In contrast , all the BL lines established from EBV-positive tumours were methylated at two or more sites on the Bim promoter . This group included four BL cell lines exhibiting a latency I pattern of EBV gene expression . These latency I cells do not express EBNA3A or EBNA3C , the viral proteins that appear to be necessary for the initial down-regulation of Bim . This strongly suggests that DNA methylation of the Bim promoter is a secondary epigenetic effect of latent EBV infection and may be triggered in BL progenitor cells ( for a more detailed and considered discussion of this point see below ) . Bisulphite sequencing was used for a more comprehensive investigation of the CpG methylation pattern on the Bim promoter of twenty-one cell lines plus the PBMCs . After bisulphite modification , a region of 565 bp was subjected to DNA sequence analysis and the 36 CpG dinucleotides were scored for their methylation state . An average of 12 and a minimum of 8 clones were sequenced for each cell type . The results presented in Figure 6 give a detailed “snapshot” of the DNA methylation state of a significant region of the Bim promoter and CpG-island . The data are largely consistent with and extend the MSP data that are summarized in Figure 5 . PBMCs and all the EBV-negative BL lines are essentially unmethylated across this region of the Bim promoter . In contrast to these EBV-negative cells , all the EBV-positive cell lines show a significantly higher frequency of methylated CpGs . Nevertheless , again the BLs infected with EBV in vitro and early passage LCLs are generally less methylated than late passage LCLs and all the EBV-positive BL lines derived from EBV-positive tumours . These and the MSP data are consistent with epigenetic repression via chromatin modification preceding DNA methylation as has been described previously for other genes [35] , [36] . Included in the series that were subjected to bisulphite sequencing was a sub-clone of Akata BL ( Akata 31 [37] ) that has lost the EBV episome in culture and therefore expresses no EBV factors . The Bim promoter in Akata 31 appears to be almost as methylated on CpG dinucleotides as the EBV-positive clone Akata 6 , indicating that – once it is established – DNA methylation of this locus does not require EBV for its maintenance . The same MSP primer sets used for all the B cell lines shown above were used to test the methylation state of the Bim promoter in DNA extracted from 14 randomly selected African BL biopsy samples ( summarized in Figure 7 ) . All these samples were isolated from patients with suspected BL and confirmed as true BL by routine pathological examination ( monomorphic lymphoma with starry-sky macrophages ) and a positive reaction by EBER-RISH , reflecting EBV presence in the tumour cells ( see Figure 8A for two representative examples ) . DNAs isolated from purified CD19+ primary B cells and PBMCs were used as negative controls . As might be expected from biopsy-derived material , the results are rather heterogenous . Nevertheless the vast majority of BL ( 13/14 ) showed some evidence of Bim promoter DNA methylation by MSP . The only biopsy scoring completely negative was one that histopathology identified as including a large epithelial field , so the amount of BL-derived DNA from this sample is uncertain . Again the primary B cells , PBMC and in addition two EBV-negative lymphoid samples were totally negative for DNA methylation . DNA from two samples ( 9 and 27 ) was used for bisulphite sequencing and the results – together with EBER staining and MSP analysis – are shown in Figure 8 . The level of methylation across the 565 bp stretch of sequenced DNA appears to be non-random , and at several loci 70–100% of the clones sequenced were positive for CpG methylation . From all the biopsy samples , amplification products were also apparent with primers specific for unmethylated DNA ( eg Figure 8B and Figure S4 ) . This may be because in some of these rapidly proliferating tumours , at the time of harvesting , cells had just replicated their DNA and methylation was not established on the newly formed DNA strand . Alternatively , the unmethylated DNA could be the result of non-BL cells in the samples resulting from infiltration into the tumour or contamination during isolation of the biopsy samples . Nevertheless , the presence of amplification products with primers specific for the methylated state in most EBV-positive BL but not EBV-negative cells is consistent with the hypothesis that CpG methylation of Bim can occur during the pathogenesis of EBV-positive BL and is directed by EBV in some unidentified way . The results of the DNA methylation studies on EBV-positive and -negative B cells suggested that CpG methylation was probably not the primary event in the suppression of Bim transcription by EBV . Specifically , early passage LCLs and EBV-negative BL cells newly converted by EBV infection in vitro have a reduced level of Bim ( protein and transcripts ) but show only very modest amounts of DNA methylation . In contrast late passage LCLs and BL cell lines derived from EBV-positive tumours have relatively high levels of DNA methylation that is non-randomly distributed on the Bim promoter . Since in cell transformation and oncogenesis , methylation of DNA is often preceded by trimethylation of lysine 27 on histone H3 ( H3K27-Me3 , [38] , [39] , [40] , [41] , reviewed [42] ) , we asked whether there were differences in this epigenetic chromatin mark on the Bim promoter in EBV-positive compared to EBV-negative B cells . Using the same two pairs of cell lines used for the histone acetylation ChIPs shown in Figure 4 and an essentially similar ChIP assay , but using mAbs directed against H3K27-Me3 , the distribution of this mark across the Bim promoter was determined . The results shown in Figure 9 are consistent with the hypothesis that EBV targets Bim for repression via the methylation of H3K27 . Using six sets of primer pairs ( indicated as A–F in Figure 2 ) for amplification and real-time Q PCR analysis of various loci in the Bim promoter/CpG-island , it is clear that latent infection of both BL31 and BL41 with EBV is accompanied by a significant increase in H3K27-Me3 throughout most of the region . For reasons that are unknown the only primers sets that did not produce this differential effect were E and F when used on BL41 cells . In order to obtain greater insight into the relationship between H3K27-Me3 and CpG methylation at specific sites across the Bim promoter , we utilized a quantitative assay that depends on the precipitation of methylated DNA fragments by a histidine-tagged methyl-CpG-binding domain ( MBD ) protein ( MBD2b ) . Using this technology it was possible to quantify the overall level of CpG methylation on the DNA fragments amplified by primers A–F ( Figure 2 and 9 ) and in parallel measure the overall level of H3K27-Me3 at the same loci ( see schematic in Figure 10A ) . These analyses were performed on BL41 and its EBV-positive equivalent , three cell lines derived from EBV-positive BLs ( Namalwa , Eli and Akata 6 ) and two early passage LCLs ( CH and Otis ) that were compared with two late passage LCLs ( IB4 and X50-7 ) . The results are shown in Figure 10B . There was a trend in all the cell lines for the distribution of these epigenetic marks to be non-random . In general , the maximum amounts of both H3K27-Me3 and CpG methylation were associated with the loci amplified by primer sets C and D . Consistent with all the preceding data , cells expressing the EBV latency III pattern exhibited a significantly higher level of H3K27-Me3 than the EBV-negative and type I latency cells . The only anomaly was Eli , that was confirmed by western blotting to have retained its latency I phenotype ( Figure S2D ) , but shows relatively high levels of H3K27-Me3 on the Bim promoter . One of the most striking observations from this series of experiments was that – as we saw using two other assays for CpG methylation ( MSP and bisulphite sequencing ) – in early passage LCLs ( CH and Otis ) the amount of DNA methylation on the Bim promoter was low . In contrast analysis of both late passage LCLs ( IB4 and X50-7 ) showed Bim has become heavily methylated . Since Bim expression is similarly low in both early and late LCL cells , the repression of Bim transcription in LCL-CH and LCL-Otis is almost certainly associated with the very high levels of H3K27-Me3 distributed across the promoter region . The data are again consistent with the hypothesis that H3K27-Me3 on Bim precedes , and may then be replaced by the more stable epigenetic mark , DNA methylation .
EBV represses the expression of Bim in human B cells [4] , [13] , [16] and this study has revealed a remarkable correlation between latent infection of B cells with this gammaherpesvirus and repressive epigenetic marks on the chromatin of the Bim gene promoter . Bim is a highly regulated inducer of apoptosis , the level of which is a critical rate-limiting factor in B cell survival [5] . It is often deleted in mantle cell lymphoma [43] and in an Eμ-myc transgenic mouse model it is a haploinsufficient tumour suppressor [7] indicating that very modest changes in the amount of Bim in a B cell can have profound effects on its survival and the development of neoplasia . Bim levels are therefore finely regulated by modulation of transcription and modulation of both mRNA and protein stability ( see Introduction ) . Recently Bim mRNA has also been identified as a target of the potentially oncogenic miR-32 and miR-17-92 microRNA clusters [44] , [45] , [46] , [47] . Although it has been appreciated for several years that latent infection with EBV can produce a dramatic reduction in the amount of Bim expressed in B cells , at the onset of this study it remained unclear at precisely what level and by what molecular mechanism EBV regulates Bim expression . By demonstrating that EBV infection leads to a significant down-regulation of Bim mRNA levels and that the reduction in Bim protein does not appear to depend on ERK signalling [4] , proteasome function or changes in Bim mRNA turnover ( this study ) , we have focused attention on Bim transcription as the primary target of EBV . Consistent with these observations , EBV-mediated repression of Bim was shown to be associated with reduced acetylation on histones H3 and H4 . Moreover Bim expression in EBV-infected cells was sensitive to agents that inhibit the maintenance of epigenetic marks on the histone and DNA components of chromatin . Taken together these results indicate that EBV primarily regulates the expression of Bim by reducing the rate of transcription through epigenetic mechanisms that may be initiated and/or maintained by the functional interaction of EBNA3A and EBNA3C [4] . Epigenetic modifications change chromatin structure without altering DNA sequence and they modify gene expression in a heritable manner . There are two closely linked components of the epigenome – methylation of cytosine in CpG dinucleotides and covalent modifications to the N-terminal tails of histones that alter chromatin organisation and function . Combined , these modifiers of gene expression can facilitate the long-term repression or silencing of genes [28] , [34] . Changes in epigenetic marks – particularly regional gains in CpG methylation associated with the silencing of tumour suppressor genes – are common in most cancers [27] . Recently there has been speculation that infectious agents may induce epigenetic modifications that could contribute to the development of human tumours or other chronic conditions ( [48] and see below ) . It is generally concluded that epigenetic repression is progressive and is probably initiated by preventing activation and removing the acetylation marks on histones that ensure chromatin has an ‘open’ active configuration . This is followed by covalent attachment of repressive marks on histones and subsequently promoter DNA methylation [28] , [34] , [35] . In the B cells we have investigated here , it is probable that EBV has inhibited the transcription of Bim – via a mechanism involving EBNA3A and 3C and repressive marks on local chromatin – and prevented the efficient assembly of transcription complexes at or around the transcription initiation site . This then made the unusually large CpG-island available for DNA methylation . The demonstration that DNA methylation is probably a relatively late event after infection with EBV prompted us to investigate a histone modification associated with transcriptional repression . Trimethylation on lysine 27 of histone H3 has been shown in cancer cells to be a precursor to CpG methylation on DNA ( [38] , [39] , [40] , [41] , reviewed in [42] ) . H3K27-Me3 of Bim occurs in low passage LCLs and in vitro ‘converts’ , and this modification then creates an ideal substrate for the natural selection of the more stable epigenetic mark DNA methylation . These data provide compelling support for a model for EBV-associated lymphomagenesis that involves virus induced epigenetic reprogramming . Currently we do not know precisely where or how H3K27-Me and CpG methylation are ‘seeded’ or the rate or mechanism of spread or even whether we have focused on the functionally important nucleotides in the CpG island . Nor do we understand the roles of EBNA3A and EBNA3C in the process . Nevertheless , since presently the only known cellular histone methyltransferase ( HMT ) capable of methylating H3K27 is a polycomb group ( PcG ) protein called EZH2 [42] then it is tempting to speculate that EBV may be utilizing the PcG family of chromatin regulators to suppress expression of Bim . Whatever the mechanisms , we suggest that spread of DNA methylation and stable repression would be driven , at least in BL , by the strong selection pressure to prevent Bim-mediated apoptosis induced by deregulated Myc [10] , [11] . In summary , we propose that when EBV infects naive B cells and type III latency is established , that Bim is down-regulated through the actions of EBNA3A and EBNA3C and that this is necessary because EBNA2 activates Myc , which would otherwise induce an increased amount of Bim and apoptosis ( see [4] , [49] ) . This involves epigenetic changes on the Bim promoter that may prevent its activation by deregulated Myc and reduces the expression of Bim below a critical threshold . Thus Myc activation by EBNA2 – or a translocation event as is found in all BL – can be tolerated in an EBV-infected B cell . The initial epigenetic marks on Bim make the CpG-island available for DNA methylation capable of reinforcing the inhibition of transcription . By definition epigenetic changes will be inherited by progeny cells and during the pathogenesis of EBV-positive BL there will be continuous selection pressure to ensure that the suppression of Bim is retained . Moreover EZH2 , the histone methyltransferase responsible for the H3K27-Me3 mark , can specifically recruit DNA methyltransferases ( DNMTs ) and so accelerate local CpG methylation [38] . This will continue to be the case after the proteins that initiated the reduction of transcription are no longer expressed in the EBNA1-only type I latency characteristic of BL [3] . To our knowledge EBV encodes no DNMTs but we cannot rule out interactions between EBV factors and cellular DNMTs ( similar to those reported for KSHV LANA1 protein and the papillomavirus nuclear oncoproteins ) being involved in the targeted repression of Bim ( see [50] , [51] ) . The recent discovery that KSHV , via the LANA1 protein , can down-regulate expression of the TGFβ-type II receptor and initiate methylation of its promoter [52] , suggests that epigenetic modification of the host genome by viruses capable of a latent infection could be a common feature during disease pathogenesis . At this stage direct interactions between EBV factors and PcG proteins should not be excluded . Finally it should be noted that screens for promoter methylation in a variety of other cancers ( including breast and ovarian ) have revealed no significant or consistent methylation of the Bim promoter using the same MSP primer sets as those used in this study ( PS – unpublished data ) . Furthermore we have seen no consistent down-regulation of any other BH3-only pro-apoptotic proteins ( including Bid , Noxa or Puma ) in EBV-positive relative to EBV-negative B cells ( our unpublished data ) . It is interesting to note that CpG methylation of Puma was recently demonstrated in Myc-driven B cell lymphomas – including some BL – but this was not dependent on infection with EBV and may even be absent in cells expressing the latency III pattern of EBV genes [53] .
|
Bim is a cellular inducer of programmed cell death ( pcd ) , so the level of Bim is a critical regulator of lymphocyte survival and reduced expression enhances lymphomagenesis in mice and humans . Regulation of Bim is uniquely important in the pathogenesis of Burkitt's lymphoma ( BL ) , since in this human childhood cancer the Myc gene is deregulated by chromosomal translocation and Myc can induce pcd via Bim . Latent EBV represses Bim expression , and here we have discovered that this involves mechanisms that reprogramme B cells and their progeny . EBV does not significantly alter Bim protein or RNA stability , but relief of EBV-mediated repression by specific inhibitors suggested it involves modifications to chromatin . Consistent with this , reduced histone acetylation and increased levels of DNA methylation on the Bim promoter were found after latent EBV infection . Further analysis suggested that the DNA methylation is preceded by repression mediated via a polycomb protein repressive complex targeting the Bim gene . By initiating the heritable suppression of Bim , EBV increases the likelihood of B lymphomagenesis in general and BL in particular . This reprogramming of B cells by EBV may also play a role in the development of other chronic disorders such as autoimmune disease and suggests a general mechanism that could contribute to the pathogenesis associated with other microorganisms .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"Methods",
"Results",
"Discussion"
] |
[
"molecular",
"biology/histone",
"modification",
"virology/persistence",
"and",
"latency",
"cell",
"biology/leukocyte",
"signaling",
"and",
"gene",
"expression",
"cell",
"biology",
"virology",
"infectious",
"diseases/viral",
"infections",
"molecular",
"biology/dna",
"methylation",
"virology/viruses",
"and",
"cancer",
"virology/effects",
"of",
"virus",
"infection",
"on",
"host",
"gene",
"expression",
"molecular",
"biology"
] |
2009
|
Epstein-Barr Virus Latency in B Cells Leads to Epigenetic Repression and CpG Methylation of the Tumour Suppressor Gene Bim
|
Species are linked to each other by a myriad of positive and negative interactions . This complex spectrum of interactions constitutes a network of links that mediates ecological communities’ response to perturbations , such as exploitation and climate change . In the last decades , there have been great advances in the study of intricate ecological networks . We have , nonetheless , lacked both the data and the tools to more rigorously understand the patterning of multiple interaction types between species ( i . e . , “multiplex networks” ) , as well as their consequences for community dynamics . Using network statistical modeling applied to a comprehensive ecological network , which includes trophic and diverse non-trophic links , we provide a first glimpse at what the full “entangled bank” of species looks like . The community exhibits clear multidimensional structure , which is taxonomically coherent and broadly predictable from species traits . Moreover , dynamic simulations suggest that this non-random patterning of how diverse non-trophic interactions map onto the food web could allow for higher species persistence and higher total biomass than expected by chance and tends to promote a higher robustness to extinctions .
In his description of the “entangled bank” of species , Darwin illustrated the principle that species must manage complex interdependencies to successfully coexist in natural communities [1–7] . In this context , evolutionary constraints set a landscape of trade-offs over which species must solve their basic needs within the context of other species ( e . g . , competition for refuges among herbivores forced by the common need to avoid predators ) and stringent environmental conditions . To some extent , each species has found unique solutions—in how they manage interactions with other species—that have shaped their distinctive niches . However , beyond species identity , common sets of trade-offs may lead to similarities in the way species are involved in different interaction types . In other words , the apparently endless solutions discovered by species to simultaneously satisfy multiple requirements and deal with multiple stresses might actually be much more limited and structured than we anticipated . Yet we do not know what the full “entangled bank” of species looks like or if there are structural patterns at the community level that reflect common solutions in the way species manage being involved in different interaction types . Indeed , the analysis tools from network science are only recently addressing the “multiplex” nature of most natural networks , i . e . , the fact that they include different interaction types between a given set of species ( e . g . , [8–11] ) . As the first datasets including several interaction types between a given set of species are now emerging in ecology [5 , 12–16] , we have a unique opportunity to disentangle the bank of species interactions . Until now , layers in such ecological networks have been analyzed separately from each other; i . e . , the structure of trophic webs has been analyzed independently of the structure of competition or mutualistic webs ( [13–15 , 17–19] , but see [5] ) . However , the way network layers are intertwined with each other matters for community dynamics and resilience [1 , 2 , 20] . Thus , it is critical to move beyond unidimensional analyses of ecological networks . In this paper , we explore a comprehensive ecological network in which the species of a local community are linked by trophic and widely diverse positive and negative non-trophic interactions [14 , 21] . The network , hereafter referred to as the Chilean web , includes three layers of interactions among 106 co-occurring species in the marine rocky intertidal community of the central coast of Chile: a trophic layer ( i . e . , a food web; 1 , 362 trophic links ) , a negative non-trophic layer ( e . g . , interference , competition for space; 3 , 089 links ) , and a positive non-trophic layer ( e . g . , habitat/refuge provisioning by sessile species that create structure for others; 172 links ) , making it a three-dimensional multiplex network [9 , 11] . We first quantified the three-dimensional structure of this multiplex network using a probabilistic clustering method . We then used dynamical modeling to investigate how the identified structure modulates the multi-species dynamics and the resilience of the ecological community to perturbations . Overall , our results suggest that the enormous ecological complexity of this community can be simplified into surprisingly clear patterns of organization that are taxonomically coherent , can be broadly predicted from simple species traits , and are functionally important for dynamics and resilience . These blocks might represent ecological and evolutionary constraints acting on the multiple requirements and impacts that allow species to persist in complex systems . Our results , therefore , pave the way for a new generation of research untangling complex networks with multiple link types .
Looking at the way pairs of species are three-dimensionally connected in the Chilean web shows that 2 , 891 of these pairwise links are interaction-specific ( Table 1; S1 Fig ) . In other words , pairs of species tend to engage in only one type of interaction: trophic , positive non-trophic , or negative non-trophic interactions . We compared these occurrences to those observed in random multiplex networks with the same expected degree sequence as in the Chilean web ( see Materials and Methods ) . Note that these random networks are very constrained and are , as a consequence , very similar to the Chilean web ( S9 and S10 Figs ) . We found that the interaction-specific links ( i . e . , the cases in which a pair of species is linked by only one interaction type ) are significantly more frequent in the Chilean web than expected in the random counterparts ( p-value < 10−4; Table 1 ) . In contrast , 125 pairs involve two interaction types simultaneously , which is far less than expected ( p-value < 10−4; Table 1 ) . Notably , six pairs of species are linked at the same time by the three interaction types in this interaction web , which is more than expected ( p-value < 10−2; Table 1 ) . These patterns suggest a fine-scale , species-level constraint on how pairs of species interact in webs with several interaction types; i . e . , multiplex pairwise interactions are remarkably rare . It does not mean that species are not involved in multiple interaction types; they usually are , but with different partners . This lack of multiplex pairwise interactions may reflect evolutionary constraints in developing adaptations simultaneously for different interaction types with the same species . For example , in the Chilean web , it is relatively rare for a species to facilitate its prey ( there are only two pairs of species simultaneously linked by a trophic and a facilitation link ) . One exception is the scurrinid limpet Scurria variabilis , which lives on top of the shell of another limpet , the keyhole limpet Fissurela limbata , which , in turn , can eat the juveniles of S . variabilis [22] . The positive effect on S . variabilis is quite strong , since they can spend their entire benthic life grazing on the Fissurella shells [22 , 23] . However , it is likely that the trophic link is weak , because the species are primarily herbivores [24–26] , which would reinforce the notion that such combination of interaction types is rare . There are , however , more examples in the Chilean web of species that compete with their prey or with their predator ( e . g . , anemones eat mussels and compete for space with them ) , of species facilitating their competitor ( e . g . , algae facilitate mussel recruitment but compete for space once mussels are established ) [27] , and , interestingly , of prey facilitating their own predators ( e . g . , mussels facilitate settlement of their predatory crabs ) [4] . While these types of examples tend to dominate our intuitive perception of insurmountable ecological complexity , the data suggests that they are the exception , not the rule . When we take into account all three types of interactions , as well as the identity of the participants , do groups of species have similar interaction profiles ? To address that question , we used a probabilistic algorithm to detect groups of species ( hereafter referred to as “multiplex clusters” ) that resemble each other in the way they interact with others in their combined trophic and non-trophic interactions ( i . e . , the way they interact in three dimensions ) . Our work hereby builds on previous efforts aimed at detecting compartments [28 , 29] or structural patterns [30] in food webs but extends those approaches to networks with several interaction types . In particular , previous studies have used similar approaches to characterize the trophic niche of species by identifying “trophic species” , i . e . , groups of species that are similar in terms of their predators and prey . Here , our approach applied to the Chilean web allows , for the first time , to our knowledge , the visualization of the multidimensional ecological niche of species [31] . When applied to the Chilean web , and associated with a model selection procedure , the probabilistic algorithm identified 14 multiplex clusters , i . e . , much less than the number of species ( Figs 1 and S2 ) . Those clusters differ from each other in the types of links they are involved in , the pattern of incoming and outgoing links ( Fig 2 ) , and the identity of the species they interact with ( S4 and S5 Figs ) . We note that the definition of the clusters requires taking into account the three layers of interactions simultaneously , because none of the layers contains by itself enough information to recover these multiplex clusters ( S6 Fig , S1 Table and S1 Text ) . Clusters 2 , 5 , and 8 are the cornerstone of that organization , both because of the high frequency of interactions engaged in with others and because of the variety of their interaction partners ( Figs 1 and 2 ) . Cluster 5 is an overall hub of interactions , with both a high frequency and a wide variety of interactions with others ( Figs 1 and 2 ) . Clusters 6 and 10 are two groups of species involved in similar interaction types and partners but that do not have a single interaction with each other ( S4 and S5 Figs ) ; indeed , the two groups of species are spatially segregated across the tidal gradient , with one group typically found in the lower shore ( cluster 6 ) and the other found at the uppermost level ( cluster 10 ) . Most of the remaining clusters contain more species ( 7 to 23 species ) that are , from a connectivity point of view , redundant and exchangeable . These clusters differ from one another by the identity of the species they interact with ( e . g . , clusters 9 and 7 are more generalist consumers than cluster 14 ) , but also by the way they interact with the species of clusters 2 , 5 , and 8 ( e . g . , cluster 11 is facilitated while 12 competes with cluster 5; S4 and S5 Figs ) . In particular , cluster 4 comprises peripheral species that share a low interacting frequency with the other clusters . The cluster number and their species composition was largely conserved after removal of up to 30% of the species in the Chilean web ( S3 Fig and S1 Text ) . This shows that the probabilistic algorithm is robust against perturbations due to species removal but also that the retrieved organization is significant . This is , however , not unexpected since , in essence , the multiplex clusters gather species that share similar interaction patterns and are therefore largely substitutable in terms of their multiplex connectivity . Do the specific combinations of trophic and non-trophic links characterizing the clusters have functional consequences ? We examined the relationship between the multiplex connectivity pattern identified in the Chilean web and the dynamic behavior of this network . To this end , we used a bio-energetic consumer-resource model ( as in [32] ) in which we incorporated the broad categories of non-trophic interactions found in the Chilean web . Because of species redundancy in the interaction patterns within a cluster , in this initial investigation , we used the clusters as the simulation units of the model . Later refinements should relax this assumption and look into the coherence of species dynamics within clusters . We compared the dynamics of ( i ) the webs of the 14 clusters identified in the Chilean web to ( ii ) equivalent random webs in which all non-trophic links were randomized throughout the web ( see Materials and Methods ) . Simulation results suggest that the way non-trophic interactions are mapped onto the trophic ones in the Chilean web tends to increase species persistence and the total biomass realized ( Fig 3 left ) , as compared to a random allocation of non-trophic interactions . This occurs for a broad range of trophic and non-trophic parameter values ( S8 Fig and S1 Text ) . Moreover , the mapping of the non-trophic interactions in the Chilean web tends to decrease secondary extinctions ( Fig 3 right ) . The different clusters had very different effects on web dynamics . For instance , biomass loss was observed after the removal of the cornerstone clusters ( clusters 2 , 5 , and 8 ) and at a higher level than expected ( cluster 5 , p-value = 0 . 056; clusters 2+8 jointly , p-value = 0 . 06; see S7 Fig ) . If we go one step further and disregard the identity of the species , can we identify deeper cores of multiplex organization ? By analyzing the interaction parameters estimated in the probabilistic model for the different clusters , we were able to identify groups of clusters whose species are involved ( or not involved ) in similar combinations of interactions , i . e . , “multiplex functional groups” ( Figs 4A and S11 ) . The Chilean web thereby further collapses into a set of only five multiplex functional groups ( Figs 4A and S11 ) . Those multiplex functional groups can broadly be characterized as groups dominated by consumers ( 1 , 4 , 7 , 9 , 14 ) , one composed mostly of competitors ( 3 , 11 , 12 ) , another dominated by facilitators/competitors ( 6 , 10 , 13 ) , a more heterogeneous group composed of consumers/competitors ( 2 , 8 ) , and , finally , one overall hub of species interacting with many other species in many different ways ( 5 ) . We find that the species composition of the functional groups is coherent with broad taxonomic classifications , considered as a coarse proxy for phylogenetic relatedness ( Fig 4C ) . Each functional group has indeed a tendency to gather closely related species ( p-value < 10−4 ) . But exceptions exist . For instance , the group of facilitators/competitors ( made of clusters 6 , 10 , 13 ) is composed of very different species corresponding to different phyla ( mainly algae and barnacles; p-value > 0 . 1 ) , but they share the fact that they are sessile species that create biotic structure for others . Interestingly , the multiplex functional groups are not only characterized by similar multidimensional interaction pattern ( by definition; Figs 4A and S11 ) , but they are also very well predicted by simple species attributes ( Figs 4B and S12 ) , in particular trophic level category ( autotroph , herbivore , intermediate , top ) , mobility ( mobile versus sessile ) , and shore height ( ordinal ) . The analysis first splits the data among autotroph species ( mainly the competitors’ group and a few of the facilitators/competitors’ group ) and the rest of the species . The second split separates mobile ( the consumers’ group ) from sessile species , which are then divided between carnivores ( the consumers/competitors’ group ) and herbivores , themselves split among species from lower ( the multiplex hub and a few consumers ) and those from higher shore ( the facilitators/competitors’ group ) . Higher on the shore is more environmentally stressful because of increased exposure to air and desiccation [33 , 34] . It might , therefore , be more likely for sessile species at mid-high shore to facilitate mobile species that need shelter from environmental stress [35 , 36] , while species lower on the shore are perhaps more likely to provide refuge from predation . Shore height could thereby mediate the frequency of facilitation of mobile by sessile species in this dataset . In sum , the five multiplex functional groups gather species that engage in roughly similar ecological interactions ( Fig 4 ) : ( 1 ) A group of mobile consumers ( clusters 1 , 4 , 7 , 9 , 14 ) , mostly carnivores , composed of crabs , sea snails , chitons , starfishes , and birds , most of which consume prey species and often find themselves in competition with others . ( 2 ) A small group of sessile , inedible consumers ( anemones; clusters 2 and 8 ) that eat dead or detached animals or their fragments are the source and target of many competitive links with other sessile species and are key players in the resilience of the community . Their classification into a separate group likely reflects their peculiar life habits ( sessile scavengers ) . ( 3 ) An overall hub of sessile , edible consumers that also facilitate others and are key in the resilience of the community ( cluster 5 ) . This group contains two common mussel species that differentiate themselves from the other groups by their involvement in all interaction types and particularly in positive interactions ( both incoming and outgoing; Figs 2 , S4 and S5 ) , supporting many ecological studies that highlight their role as foundational or engineering species [4 , 37 , 38] . They indeed provide habitat and substrate for many other invertebrate species seeking shelter . ( 4 ) A group of sessile primary producers ( algae; clusters 3 , 11 , 12 ) that compete for space and usually find themselves in competitive loops while being frequently consumed . ( 5 ) Finally , a group of sessile species ( clusters 6 , 10 , 13 ) that is a mix of algae and barnacles that compete for space with other sessile species while facilitating mobile consumers by creating biotic structure that provides refuges and habitat for other species ( for instance , the kelp Lessonia nigrescens facilitates recruitment and provides critical shelter or habitat to diverse species ) .
The wave-exposed Chilean marine intertidal ecosystem of 106 species includes over 4 , 600 interactions that span predation , competition , and facilitation . Despite the wide range of possible combinations of interactions among species , our data suggests that the combinations of interactions that are actually realized in this intertidal community are constrained to be far fewer than those “possible . ” Our analysis of the Chilean web further reveals a clear organization of species into a small subset of multiplex clusters , which themselves collapse into multiplex functional groups . The identification of this organization into clusters and , therefore , into functional groups requires taking into account the three layers of interactions and would not be possible with a monolayer , unidimensional niche approach of this ecological network . The functional groups identified are taxonomically coherent , with each group gathering closely related species , suggesting some level of conservatism of the three-dimensional interaction niche space . The functional groups are also well-predicted by simple traits , such as trophic level , mobility , and shore height . Previous work on different single-interaction-type networks ( food webs , bipartite mutualistic , and bipartite antagonistic ) showed that only a limited number of traits is required to explain all species interactions in a given ecological network , meaning that ecological networks are structured by a few dimensions ( or trait-axes ) [31] . Our analysis of the Chilean web suggests that this result may hold when considering multiplex ecological networks . Together , the small sets of interaction types in which species engage with each other and the astonishingly limited set of multiplex functional groups seems to reflect predictable evolutionary and ecological constraints operating in this entangled bank of species . This opens up a pathway toward simplifying ecosystem complexity into basic building blocks . Previous theoretical studies have suggested that the incorporation of non-trophic interactions in food webs can have important consequences for species diversity [1 , 5 , 7] , overall productivity [1] , frequency of functional extinctions [39] , stability [6 , 20 , 40–42] , and the complexity–stability relationship [6 , 40 , 43] . May’s pioneering work in the early 1970s already included several interaction types [44] . Combining trophic and competitive interactions and using community matrices derived from real food webs , Yodzis [42] showed that a certain level of intraspecific interference contributed to the local stability of ecological communities , whereas interspecific competition tended to be destabilizing . In recent extensions of May’s work , Allesina and Tang [40] showed that matrices including mixtures of competition and mutualism were less likely to be locally stable than predator–prey matrices . Using a similar approach , Mougi and Kondoh [6] found that introducing a small proportion of mutualistic links could destabilize an otherwise stable food web , but that stability reached a peak at a moderate mixture of both interaction types ( but see [45] ) . Studies on bipartite networks have suggested that the way different bipartite networks ( e . g . , mutualistic and antagonistic networks ) are connected to each other could affect their stability [5] . Our study extends these results to show that the specific three-dimensional signature of the clusters and , in particular , the non-randomness of non-trophic interactions , can promote higher species persistence , higher total biomass , and higher robustness to extinctions than random networks in which the multidimensional connectivity pattern is lost . A long history of theoretical and empirical work on food webs highlighted the importance not only of the structure of food webs ( i . e . , the repartition of the links in the web ) [42 , 46–48] but also of the specific pattern of interaction strength for the stability of ecological communities [18 , 19 , 49] . Here , with the exception of a few common links , we lack information about interaction strengths for the entire Chilean web and especially about the strength of the non-trophic links . Getting information about those interaction strengths , their structure , the way they should be modeled , and their functional relevance remains an important empirical but also theoretical challenge . To what extent the connectivity patterns identified in the Chilean web are unique to this intertidal community or general to all marine organisms or even to all ecosystems must be evaluated by comparison to those other ecosystems as more data on multiplex ecological networks becomes available [13 , 14 , 50] . The five functional groups identified could very well correspond to sets of strategies largely generalizable to other ecosystems . For example , a cluster of mobile consumers ( top predators ) might generally emerge . In the same vein , a group of sessile edible species competing for space is probably identifiable in many ecosystems . In terrestrial ecosystems , such a group would mostly be composed of basal primary producers , whereas in marine systems it could include sessile animals and exclude some primary producers that are not sessile ( e . g . , phytoplankton ) . Groups of sessile species that create biotic structure and habitat for others—notably , mobile consumers—while also competing for space are likely to be common across many ecosystems . Finally , identifying “multiplex hubs” in other ecosystems—such as mussels in the Chilean web , which create structure while also being an important food source—may help target a small subset of species that play a disproportionately important role for the community resilience . Conversely , some groups may be unique to marine benthic systems , such as the group of sessile , inedible scavengers formed here by anemones . It is noteworthy that a number of key groups of species are absent from the current version of the dataset ( Materials and Methods ) . In particular , parasites are not included in the web . Studies have shown that food webs that also take parasites into account have increased connectivity and longer food chains , and the parasite–host links dominate numerically over predator–prey links [12] . Detritus ( and thereby decomposers ) are known to play an important role for the dynamics and structure of many communities and may also affect their stability [49 , 51 , 52] . It is unclear , however , what the significance is of local nutrient recycling in benthic marine and stream communities . In any case , adding missing species into the Chilean web could , depending on the connectivity of the newly introduced species , lead to either the emergence of new functional groups or the splitting of some of the current functional groups into additional groups [15] . The spatial and temporal variations of the patterns identified in the Chilean web remain to be investigated . This variability in space and time has been suggested to be essential to the stability and function of ecosystems [53] . The role of space may be particularly relevant in intertidal communities where mobile species ( mainly consumers ) could connect distant communities along the shore , with possible important consequences for the stability of these communities [48 , 53] . In addition , the incorporation of several interaction types in spatial ecology frameworks has been shown to have important consequences for community dynamics . For instance , Lurgi et al . [41] showed , using a spatially explicit individual-based model , that an increasing proportion of mutualistic links in a food web positively affected the dynamic stability of model communities . How the topological structure of multiplex ecological networks modulates the multi-species dynamics and the resilience of ecosystems to perturbations , such as climate change , must be further investigated through other datasets [15] , further dynamical modelling [1 , 5 , 20 , 41 , 45] , and other approaches incorporating link weighting [3] . Until then , our results will help us guide future empirical studies and move toward a more general theory of how to leverage the full diversity of species interactions for understanding and predicting the dynamics and resilience of complex ecological systems .
The dataset [14] includes all the species that were found to co-occur during community structure surveys carried out at several rocky intertidal sites with similar wave exposure spread along 700 km of the central Chilean coast ( [27] , see [54 , 55] for sampling details and species list ) . Construction of the network was based on expert knowledge [14] . An interaction was included in the network if one species plausibly had a direct measurable effect on the growth , survival , or feeding rates of another species over an ecologically relevant time period ( 1–2 y ) [14] . The dataset does not include parasites , endo-symbionts , or decomposers , because such data was unavailable for that community . The network was split into three separate matrices for trophic , positive non-trophic , and negative non-trophic interactions ( in each matrix , interactions are coded as 0 or 1 ) [14] . As a live and continuously improving network , some changes have been made to the network since first published [14] . These are mostly taxonomic changes and the inclusion of porcellanid crabs as part of the wave-exposed network . Furthermore , the biofilm taxa and plankton ( zooplankton and phytoplankton ) were each considered as a single node in the Chilean web due to lack of information . The main assumptions made to build this network as well as possible related bias are discussed in Appendix A of [14] . In particular , we acknowledge that there may be “a bias in favor of negative non-trophic interactions at lower trophic levels , ” because “measuring the relative importance of interference competition among rare species under natural conditions is particularly challenging” [14] . “When local experimental information was lacking for a pair of sessile species , we probably had a greater tendency in assigning ( i . e . , benefit of doubt ) the interaction to competition for space than when dealing with pairs of mobile species at higher trophic levels . This would create a bias in favor of negative non-trophic interactions at lower trophic levels . However , the sheer number of species at bottom versus high trophic levels would make it difficult to alter the general pattern” [14] . Data deposited in the Dryad repository: http://dx . doi . org/10 . 5061/dryad . b4vg0 [21] . The pairwise multiplex interactions observed in the Chilean web were compared to those observed in random multiplex networks simulated layer by layer . For each layer , we imposed that the expected in- and out-degree sequences were equal to the degree sequences in the original layer of the Chilean web . To do so , we used the procedure explained in the “random network” paragraph hereafter . We calculated the statistical significance of any observed number of links by computing the empirical distribution of the number of links in the 104 random multiplex networks . How can we tell what a multiplex network looks like ? How can we summarize its structure ? To answer these questions , classical approaches consist of pooling nodes that have similar connectivity patterns into clusters to extract the high-level structure of a complex network . Most of these approaches rely on finding modules or communities ( clusters of nodes that are more connected inside than outside their cluster [56] ) . But , in ecological networks , could there be relevant structural patterns that we do not find because we have not thought to search beyond the modular structure ? To circumvent this problem , we used a probabilistic clustering approach based on Stochastic block models [57–59] . Here , the cluster identification does not rely on any a priori hypothesis about the connectivity patterns to be found but aims precisely at identifying significant hidden connectivity patterns ( e . g . , modularity , centrality , hierarchy ) or combinations of these patterns . Stochastic block models have been widely used for networks with one layer ( see [30 , 60] for ecological networks ) , but not for multiplex networks as proposed in this paper . We followed the notations and the estimation procedure previously described in [60 , 61] and extended the model to multiplex networks with 3D-interactions using an appropriate probability distribution . The use of a probability distribution allows us to account for the randomness and the variability of the network and ensures a significant robustness to potential errors ( spurious or missing links , for instance ) . We consider n = 106 interacting species , with Yij standing for the observed measure of these 3D interactions and Y = ( Yij ) . Yij is a 3-dimensional vector such that Yij = ( Yij1 , Yij2 , Yij3 ) , where Yij1 = 1 if there is a trophic interaction from i to j and 0 otherwise , Yij2 for a positive interaction , and Yij3 for a negative interaction . We now introduce the vectors ( Z1 , … , Zn ) , where for each species i Ziq are the component of vector Zi such that Ziq = 1 if i belongs to cluster q and 0 otherwise . We assume that there are Q clusters with proportions a = ( a1 , … , aQ ) and that the number of clusters Q is fixed ( Q will be estimated afterward; see below ) . In a Stochastic block model , the distribution of Y is specified conditionally to the cluster membership: Zi~Multinomial ( 1 , a ) , Zj~Multinomial ( 1 , a ) ( 1 ) Yij|ZiqZjl=1~f ( . , θql ) , ( 2 ) where the distribution f ( . , θql ) is an appropriate distribution for the Yij of parameters θql . The novelty here is to use a 3D-Bernoulli distribution [62] that models the intermingling connectivity in the three layers—trophic , positive non-trophic , and negative non-trophic interactions . The objective is to estimate the model parameters and to recover the clusters using a variational expectation–maximization ( EM ) algorithm [60 , 63] . It is well known that an EM algorithm’s efficiency is governed by the quality of the initialization point . We propose to use the clustering partition obtained with the following heuristical procedure . We first perform a k-means clustering on the distance matrix obtained by calculating the Rogers and Tanimoto distance ( R package ade4 ) between all the 3D interaction vectors Vi = ( Yi . , Y . i ) associated to each species i . Second , we randomly perturb the k-means clusters by switching between 5 and 15 species membership . We repeat the procedure 1 , 000 times and select the estimation results for which the model likelihood is maximum . Lastly , the number of groups Q is chosen using a model selection strategy based on the integrated classification likelihood ( ICL ) ( see S2 Fig ) [61] . The algorithm eventually provides the optimal number of clusters , the cluster membership ( i . e . , which species belong to which cluster ) , and the estimated interaction parameters between the clusters ( i . e . , the probability of any 3D interaction between a species from a given cluster and another species from another or the same cluster ) . Source code ( R/C++ ) is available upon request for people interested in using the method . See S1 Text for a discussion about the choice of this approach . We use the bioenergetic consumer-resource model found in [32 , 64] , parameterized in the same way as in previous studies [28 , 32 , 64–66] , to simulate species dynamics . The changes in the biomass density Bi of species i over time is described by: dBidt=ri ( 1−BiKi ) Bi+eiBi∑jFijTR ( i , j ) −∑kFkiBkTR ( k , i ) −xiBi ( 3 ) where ri is the intrinsic growth rate ( ri >0 for primary producers only ) , Ki is the carrying capacity ( the population size of species i that the system can support ) , e is the conversion efficiency ( fraction of biomass of species j consumed that is actually metabolized ) , Fij is a functional response ( see Eq 4 ) , TR is a n*n matrix with n the number of nodes in the network and whose i , j element is positive if species i consumes species j , and xi is the metabolic rate . The functional response of species i consuming species j is defined as multi-prey Holling-type functional response [67]: Fij=wibijBj1+q1+wihi∑kTR ( i , k ) bikBk1+q ( 4 ) where wi is the relative consumption rate , which accounts for the fact that a consumer has to split its consumption between its different resources; it is defined as 1/ ( number of resources of species i ) , bij is the attack rate of predator i on prey j , hi is the handling time of predator i , 1+q is the Hill-exponent with q the Hill coefficient ( q = 0 yields a type II functional response , q = 1 yields a type III functional response ) . To test the significance of the assemblage of the different interaction types in the Chilean web , we simulated multiplex networks for which the most important topological properties ( number of edges , in/out-degrees , degree correlation between layers ) are identical to those in the Chilean web . For each layer ( trophic , positive and negative non-trophic ) , we imposed that the expected in- and out-degree sequences ( i . e . , the list of species degrees ) were equal to the degree sequences in the original layer of the Chilean web ( S9 and S10 Figs and S1 Text ) . The consequence of these strong constraints is that ( 1 ) any species observed individually has the same 3-dimentional connectivity properties in the random networks , but is likely to have different partners than in the original Chilean web; and ( 2 ) the random networks are ecologically meaningful , because properties such as the trophic levels are conserved . Technically , we extrapolated the procedure in [70] and drew directed edges between species i and j with probability pij = ( diout * djin ) /m , where m , diout , and djin are the number of edges , the out-degree of i , and the in-degree of j in the given layer of the Chilean web . To avoid size effect biases , we only kept the simulated networks for which the number of edges is 100+/-2 . 5% the number of edges in the original Chilean web . For the pairwise analysis ( Table 1 ) , the three layers were randomized . For dynamical modeling , because we wanted to assess the role of the structure of the non-trophic interactions relative to the trophic one , the trophic layer was kept fixed and only the positive and negative non-trophic interaction layers were randomized .
|
Within an ecosystem , species interact with each other in many different ways , including predation , competition , and facilitation , and this can be modelled as a network of multiple interaction types . The variety of interaction types that link species to each other has long been recognized but has rarely been synthesized for entire multi-species ecosystems . Here , we leverage a unique marine ecological network that integrates thousands of trophic and non-trophic interactions . We show that , despite its multidimensional complexity , this ecological network collapses into a small set of “functional groups , ” i . e . , groups of species that resemble each other in the way they interact with others in their combined trophic and non-trophic interactions . These groups are taxonomically coherent and predictable by species attributes . Moreover , dynamic simulations suggest that the way the different interaction types relate to each other allows for higher species persistence and higher total biomass than is expected by chance alone , and that this tends to promote a higher robustness to extinctions . Our results will help to guide future empirical studies and to develop a more general theory of the dynamics of complex ecological systems .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"community",
"ecology",
"conservation",
"biology",
"conservation",
"science",
"ecology",
"and",
"environmental",
"sciences",
"species",
"extinction",
"ecology",
"species",
"delimitation",
"food",
"web",
"structure",
"predation",
"trophic",
"interactions",
"speciation",
"biology",
"and",
"life",
"sciences",
"theoretical",
"ecology",
"species",
"interactions",
"evolutionary",
"biology",
"evolutionary",
"processes"
] |
2016
|
How Structured Is the Entangled Bank? The Surprisingly Simple Organization of Multiplex Ecological Networks Leads to Increased Persistence and Resilience
|
Common genetic variation could alter the risk for developing bladder cancer . We conducted a large-scale evaluation of single nucleotide polymorphisms ( SNPs ) in candidate genes for cancer to identify common variants that influence bladder cancer risk . An Illumina GoldenGate assay was used to genotype 1 , 433 SNPs within or near 386 genes in 1 , 086 cases and 1 , 033 controls in Spain . The most significant finding was in the 5′ UTR of VEGF ( rs25648 , p for likelihood ratio test , 2 degrees of freedom = 1 × 10−5 ) . To further investigate the region , we analyzed 29 additional SNPs in VEGF , selected to saturate the promoter and 5′ UTR and to tag common genetic variation in this gene . Three additional SNPs in the promoter region ( rs833052 , rs1109324 , and rs1547651 ) were associated with increased risk for bladder cancer: odds ratio ( 95% confidence interval ) : 2 . 52 ( 1 . 06–5 . 97 ) , 2 . 74 ( 1 . 26–5 . 98 ) , and 3 . 02 ( 1 . 36–6 . 63 ) , respectively; and a polymorphism in intron 2 ( rs3024994 ) was associated with reduced risk: 0 . 65 ( 0 . 46–0 . 91 ) . Two of the promoter SNPs and the intron 2 SNP showed linkage disequilibrium with rs25648 . Haplotype analyses revealed three blocks of linkage disequilibrium with significant associations for two blocks including the promoter and 5′ UTR ( global p = 0 . 02 and 0 . 009 , respectively ) . These findings are biologically plausible since VEGF is critical in angiogenesis , which is important for tumor growth , its elevated expression in bladder tumors correlates with tumor progression , and specific 5′ UTR haplotypes have been shown to influence promoter activity . Associations between bladder cancer risk and other genes in this report were not robust based on false discovery rate calculations . In conclusion , this large-scale evaluation of candidate cancer genes has identified common genetic variants in the regulatory regions of VEGF that could be associated with bladder cancer risk .
Bladder cancer is primarily a sporadic disease , and environmental factors such as tobacco smoking and occupational exposure to aromatic amines have been established as strong determinants of risk [1] . A moderate familial component has been demonstrated for bladder cancer , but so far no high-penetrance mutations have been described [1] . However , there is strong evidence for the influence of common genetic variants on bladder cancer risk . Most notably , large studies have demonstrated associations with each of the NAT2 and GSTM1 genotypes and a probable interaction between smoking and NAT2 genotype [2] . Specifically , the GSTM1 null genotype increases the overall risk of bladder cancer; while the NAT2 slow acetylator genotype appears to increase risk particularly among cigarette smokers [2] . In this context , we hypothesized that a large-scale effort to screen common variants in candidate cancer genes could identify additional bladder cancer susceptibility genes . The recent development of highly multiplexed single nucleotide polymorphism ( SNP ) genotyping assays has resulted in an opportunity to screen candidate genetic variants in an affordable , high-throughput manner in epidemiological studies . We used a GoldenGate assay by Illumina targeted to analyze over 1 , 500 SNPs in selected candidate cancer genes in order to identify bladder cancer susceptibility genes using samples collected in a large case-control study of bladder cancer in Spain . Because this was one of the first epidemiological studies using this highly multiplexed technology , we performed a detailed analysis of data quality . All SNPs chosen for this platform were drawn from the SNP500Cancer public database ( http://snp500cancer . nci . nih . gov ) , which includes genes or specific genetic variants that could be important in cancer and have been re-sequenced in 102 individuals [3] .
We obtained high-quality genotype calls from 1 , 433 SNP assays in or near 386 genes involved in cancer-related pathways , with a median of two SNPs per gene ( range: 1–37 SNPs per gene ) . About half ( 51% ) of these SNPs were located in introns , 32% in exons , 12% in promoter regions , and 5% in 3′ of stop codon ( STP ) . For SNPs located in exons , 5% were in 5′ UTRs , 68% in coding regions ( approximately half were synonymous and half non-synonymous changes ) and 27% in 3′ UTRs . The median ( range ) minor allele frequency ( MAF ) among controls was 0 . 24 ( 0 . 02–0 . 50 ) . The global genotype completion for study samples was ≥99% . Genotype concordance in 69 blinded duplicate blood DNA pairs was ≥99% . Of the 1 , 433 SNPs in the GoldenGate assay , 72 SNPs had been previously typed using other genotyping platforms on 2 , 256 blood DNA samples from study participants , and 31 of these SNPs had been previously genotyped on 50 buccal DNA samples from the study . Genotype concordance between the GoldenGate and other platforms ( primarily TaqMan ) was ≥98% . About 5% ( 79/1433 ) of genotype assays had a significant ( p < 0 . 05 ) departure from Hardy-Weinberg equilibrium , which is consistent with what would be expected by chance . Gene-based ( Table 1 ) and SNP-based ( Table S1 ) analyses showed promising associations ( i . e . , p-value for trend or likelihood ratio test [LRT]; 2 degrees of freedom [df] <0 . 01 ) with bladder cancer risk for 19 genes: VEGF , STK11 , CYP1B1 , ZNF350 , PTH , GHR , CASP9 , PLA2G6 , GSTA4 , ROS1 , RB1CC1 , TERT , XRCC4 , FZD7 , CETP , CYP24A1 , LIPC , ESR1 , and HSD17B4 . In addition , ARHGDIB , SHBG , GPX4 , and STAT1 showed significant associations with risk according to the LRT ( 2 df ) ; however , estimates for heterozygous and homozygous variants showed associations in opposite directions ( Table S2 ) . The most significant association with bladder cancer risk according to gene-based analyses was observed for the vascular endothelial growth factor ( VEGF ) gene ( Table 1 ) . We evaluated these findings using the false discovery rate ( FDR ) approach described in Materials and Methods . FDR values for the VEGF association with bladder cancer risk , taking into account all 386 genes evaluated in this report , was 0 . 032 based on the global p-values from LRTs ( 2 df ) performed for each gene . The next lowest FDR value was 0 . 56 for STK11 , indicating that the associations for other genes were not robust . We also calculated FDR values for the 386 trend tests performed for each gene , and the lowest FDR value was 0 . 51 for GHR , with a value of 0 . 64 for VEGF . Individual SNP analyses showed the strongest association for a variant allele in the 5′ UTR of VEGF ( rs25648 , RefSNP accession number assigned by dbSNP , http://www . ncbi . nlm . nih . gov/projects/SNP ) . The MAF for this SNP among the control population was 0 . 14 , and the odds ratio ( OR ) ( 95% confidence interval [CI] ) for heterozygote and homozygote variant genotypes compared to the common homozygote genotype was 1 . 12 ( 0 . 91–1 . 37 ) and 5 . 11 ( 2 . 33–11 . 20 ) , respectively; p-values for LRT ( 2 df ) = 1 × 10−5 and for trend test = 0 . 002 ( Table 2 ) . The observed frequency for the homozygote variant genotype was lower than expected under Hardy-Weinberg equilibrium in the control population ( 0 . 8% observed versus 1 . 9% expected , p-value = 0 . 002 ) , while genotype completion and concordance for this SNP were measured at 100% . To explore the impact of the observed departure on estimates of relative risk , we re-estimated ORs ( 95% CIs ) assuming Hardy-Weinberg equilibrium [4] . Estimates for heterozygote and homozygote variant genotypes assuming Hardy-Weinberg equilibrium were 1 . 23 ( 1 . 02–1 . 48 ) p-value = 0 . 03 and 2 . 20 ( 1 . 48–3 . 28 ) p-value = 0 . 00009 , respectively . We performed additional genotyping for 29 SNPs in VEGF ( including the three SNPs previously genotyped ) in an effort to dissect the locus to follow-up findings from our exploratory analysis described above . Two of these SNPs showed low genotypic variation in this population ( no variants were observed for rs3024989 , and three controls and no cases were heterozygote for rs9367173 ) . The concordance for the three VEGF SNPs previously genotyped in the GoldenGate assay ( rs1005230 , rs25648 , and rs3025039 ) was 100% . Genotype completion and concordance rates for all VEGF SNPs exceeded 99% , and all but rs25648 ( p = 0 . 002 ) were in Hardy-Weinberg equilibrium in controls . Analyses showed significant associations with three additional SNPs located in the promoter region of VEGF ( rs833052 , rs1109324 , and rs1547651 ) and one SNP in intron 2 ( rs3024994 ) ( Table 2 ) . However , the association for the rs833052 promoter SNP was only borderline significant . Two of these SNPs ( rs1109324 and rs1547651 ) were in strong linkage disequilibrium ( LD ) with the 5′ UTR SNP ( D′ ≥ 0 . 94 and r2 ≥ 0 . 84 ) . None of the other SNPs showed significant associations with bladder cancer risk ( see Tables 2 , S3 , and S4 for more details ) . We evaluated interactions between the VEGF SNPs significantly associated with bladder cancer risk and other determinants of risk ( i . e . , age , gender , smoking status , family history of cancer in at least one first-degree relative , and NAT2 and GSTM1 genotypes ) . Analyses suggested stronger associations for the two correlated SNPs in the VEGF promoter ( rs1109324 and rs1547651 ) among subjects with a family history of cancer ( p-value for heterogeneity = 0 . 035 and 0 . 036 , respectively; Table S5 ) . We also observed a stronger association for the 5′ UTR SNP ( rs25648 ) among subjects with the GSTM1 null genotype ( p-value for heterogeneity = 0 . 031; Table S5 ) . However , these findings need to be interpreted with caution given the number of interactions evaluated . Haplotype analyses were based on 27 VEGF SNPs ( of the 29 SNPs determined , two were excluded because of low genotypic variation ) . Of the three large blocks defined by LD in our control population ( Figure 1 ) , we observed significant associations between haplotypes and risk for bladder cancer in two LD blocks , including the promoter and 5′ UTR ( global p = 0 . 023 and 0 . 043 for blocks 1 and 2 , respectively; Figure 2 ) . Consistent with individual SNP analyses , the AT haplotype in block 1 carrying the variant allele for rs833052 was associated with increased bladder cancer risk; however , the CT haplotype was related to decreased bladder cancer risk , which was not predicted by individual SNP analyses . Both the individual SNP and haplotype associations were only of borderline significance and thus could be due to chance . Of the nine observed haplotypes in block 2 , only one ( GAGCCGTGCTGGTCCCT ) carried the variant in intron 2 ( rs3024994 ) that was individually associated with reduced risk ( Figure 2 ) . Consistent with SNP analyses , this haplotype was also associated with a reduction in risk . Two other haplotypes carried at least one variant for three correlated SNPs individually associated with risk ( rs1109324 , rs1547651 , and rs25648 ) . Both haplotypes were associated with increases in risk , although the association for the haplotye carrying only the rs25648 variant ( GAGATGCGTCGGCCCCC ) was not significant , possibly due to its low frequency in the population ( 1 . 0% of controls ) ( Figure 2 ) . Therefore , haplotype analyses cannot help distinguish which of the three correlated SNPs is most important in determining risk .
An exploratory analysis of 1 , 433 SNPs in or near 386 genes involved in cancer-related pathways using the GoldenGate assay led to the identification of novel associations for several promising genes , the most notable finding , a 5′ UTR SNP in VEGF . Subsequent analyses that captured nearly all common variants in this gene showed additional associations with SNPs in the promoter and intron 2 , providing further evidence for the importance of variation in regulatory elements of VEGF and bladder cancer risk . The association of common genetic variation in VEGF with bladder cancer risk is biologically plausible for several reasons: ( 1 ) VEGF has been identified as a critical factor in angiogenesis required for tumor growth , ( 2 ) VEGF expression in bladder tumors has been related to tumor progression [5] , and ( 3 ) in vitro studies have suggested that common haplotypes in the 5′ region of VEGF alter gene expression [6] . A large block of LD that extended from the promoter to intron 5 included the 5′ UTR SNP ( rs25648 ) that demonstrated the strongest association with bladder cancer risk in our initial screen . This SNP showed a departure from Hardy-Weinberg equilibrium in controls ( p-value = 0 . 002 ) , which was not observed for two correlated SNPs in the promoter region ( rs1109324 and rs1547651; r2 > 0 . 80 between the 5′ UTR and promoter SNPs ) also associated with risk . We did not observe more deviations from Hardy-Weinberg equilibrium than expected by chance in our control population , the deviation for rs25648 was not observed in other Caucasian populations [7] ( http://snp500cancer . nci . nih . gov ) , and quality control samples did not show evidence for genotype errors; therefore , this departure is likely to have occurred by chance . The observed magnitude of the association for rs25648 was larger than for the two promoter SNPs; however , this difference was less apparent after re-estimation of ORs assuming Hardy-Weinberg equilibrium , suggesting that it might have occurred by chance . Carriers of the 5′ UTR SNP ( rs25648 ) have been found to have increased VEGF mRNA levels in adenocarcinoma tissues of patients with colorectal adenocarcinomas [6] . Although there are no functional studies of the two promoter SNPs associated with bladder cancer risk in our study population , previous studies have shown that variant genotypes or haplotypes falling in the same block of LD ( rs699947 , rs1570360 , rs833061 , and rs2010963 ) were associated with ( 1 ) higher induced gene expression from hypoxia in transient transfection assays [8]; ( 2 ) higher VEGF production in peripheral blood mononuclear cells [9 , 10]; ( 3 ) increased promoter activity and responsiveness to phorbol esters in breast cancer cell lines [11]; and ( 4 ) tumor aggressiveness in breast cancer patients [12] . Finally , the SNP in intron 2 ( rs3024994 ) associated with reduced risk fell in the same block of LD but showed low correlation with other SNPs , and there are no published studies on its functional significance . Only one SNP in VEGF ( rs699947 ) has been previously evaluated in relation to bladder cancer risk in a small study of 153 bladder cancer patients and 153 controls in South Korea [13] . Consistent with our results , this study found no association between this SNP and bladder cancer risk . A SNP in 3′ of STP ( rs3025039 ) has been associated with decreased plasma levels of VEGF and decreased breast cancer risk [14]; however , neither this nor other SNPs in LD were associated with bladder cancer risk in our study . We also observed promising associations with bladder cancer risk for other genes involved in carcinogenesis pathways . However , based on FDR calculations , taking into account all genes evaluated in this report , the additional associations were not robust and thus should be pursued in additional study populations . To the best of our knowledge , this is the first large-scale evaluation of candidate genes in bladder cancer using highly multiplexed technologies . Our data demonstrate that these technologies provide high quality data and that they can be useful in identifying genetic susceptibly factors . In particular , we provide reasonable evidence for an association between common variants in the promoter and 5′ UTR of VEGF and bladder cancer risk . Further work is required to replicate the findings in other populations and to identify the potential causal variant by more detailed genetic mapping , including sequencing and functional characterization of variants .
The study population has been previously described [2] . Briefly , cases were patients participating in the Spanish Bladder Cancer Study diagnosed with histologically confirmed bladder carcinoma in 1998–2001 , aged 21–80 y ( mean [sd] = 66 [10] y ) , of which 87% were males . Controls were selected from patients admitted to participating hospitals for diagnoses believed to be unrelated to the exposures of interest , individually matched to the cases by age at interview within 5-y categories , gender , ethnicity , and region . Demographic and risk factor information was collected at the hospitals using computer-assisted personal interviews . A total of 1 , 219 cases ( 84% of eligible cases ) and 1 , 271 controls ( 88% of eligible controls ) agreed to participate in the study and were interviewed . Of these , 1 , 188 ( 97% ) cases and 1 , 173 ( 92% ) controls provided a blood or buccal cell sample for DNA extraction . Adequate amounts of DNA for genotyping were available from 1 , 116 cases ( including eight from buccal cells ) and 1 , 043 controls ( including 36 from buccal cells ) . Further exclusions were made to reduce heterogeneity ( cases with nontransitional histology and nonwhite subjects ) , or because of DNA contamination or lack of information on smoking status . After exclusions , the available samples for genotype analysis were 1 , 086 cases and 1 , 033 controls . We obtained informed consent from potential participants in accordance with the National Cancer Institute and local institutional review boards . A GoldenGate assay ( Illumina , http://www . illumina . com ) was developed using SNPs in the SNP500Cancer project ( http://snp500cancer . nci . nih . gov ) with previous re-sequence analysis and plausible evidence that the gene is related to carcinogenic processes [3] . SNP selection favored nonsynonymous SNPs , those previously evaluated in relation to cancer risk , or those with evidence for functional significance . The GoldenGate assay was designed to examine 1 , 536 SNPs based on an initial screen of 3 , 072 SNPs drawn from the SNP500Cancer database ( November 2004 ) and subsequently analyzed in the unrelated HapMap Centre d'Etude du Polymorphisme Humain ( CEPH ) Utah samples . Of the 1 , 536 assays chosen for this study , 103 were dropped from the analysis because of low MAF or assay problems . Thus , we obtained data on 1 , 433 SNPs in or near 386 genes ( Table S1 ) . DNA samples from cases and controls were randomly sorted , including 69 duplicated DNA samples for genotyping quality control . Based on our primary analysis that showed the strongest association of bladder cancer with an SNP in VEGF , we performed a comprehensive evaluation of common variation in this gene . We initially selected 31 SNPs spanning 20 kb 5′ of the start of transcription to 10 kb 3′ of the end of exon 8 of the VEGF gene using the following methods: ( 1 ) 15 tag SNPs were chosen based on the aggressive tagging algorithm [15] ( r2 ≥ 0 . 80 , MAF ≥ 0 . 05 ) using genotype data from the unrelated HapMap CEPH Utah individuals; ( 2 ) 16 SNPs from the Single Nucleotide Polymorphism database were added as “fill-in” to ensure the inclusion of an SNP every 2–5 kb across the region , particularly in the 5′ region . iPLEX ( Sequenom , http://www . sequenom . com ) assays were designed and optimized with the SNP500Cancer set of 102 individuals . Two SNPs were dropped because of design and performance problems . Out of 29 , 28 assays were optimized on iPLEX and one SNP ( rs699947 ) that could not be included was analyzed using TaqMan ( Applied Biosystems , http://www . appliedbiosystems . com ) . Because of restricted amounts of DNA available and poor assay performance for a small subset of samples and exclusions for data analyses described earlier ( cases with nontransitional histology , nonwhite subjects , and lack of information on smoking status ) , a total of 926 cases and 900 controls were included in the analyses . For each individual SNP , we estimated OR and 95% CI using logistic regression models adjusting for gender , age at interview in 5-y categories , region , and smoking status ( never , occasional , former , and current; see [2] for details on the definition of these variables ) . The association between individual SNPs and bladder cancer risk was tested using a 2-df LRT and a linear trend test assuming a dose response with increasing number of variant alleles . ORs and 95% CIs “per variant allele” were estimated under the latter assumption ( i . e . , coding genotypes as 0 , 1 , and 2 depending on the number of variant alleles ) . Heterogeneity of genotype ORs among groups of subjects defined by age , gender , smoking status , family history of cancer in at least one first-degree relative , and NAT2 and GSTM1 genotypes ( see [2] for details on the definition of these variables ) were evaluated by introducing interaction terms in logistic regression models . The 1 , 433 individual SNPs evaluated were located within or near 386 candidate genes . We performed two gene-based tests for association: ( 1 ) an LRT for each gene comparing models with and without terms for heterozygous and homozygous variant genotypes for each SNP in a given gene ( df = 2 × number of SNPs per gene ) ; ( 2 ) an LRT for each gene comparing models with and without terms for each SNP ( genotypes coded as 0 , 1 , and 2 ) in a given gene ( df = number of SNPs per gene ) . For highly correlated SNPs ( r2 > 0 . 90 ) within a gene , only one of the SNPs was included in the model to avoid collinearity problems ( Table S1 ) . Haplotype frequencies , ORs , and 95% CIs for genes showing blocks of LD were estimated using HaploStats ( http://mayoresearch . mayo . edu/mayo/research/biostat/schaid . cfm ) . This program reconstructs haplotypes and estimates ORs simultaneously based on a suitable Expectation-Maximization algorithm [16 , 17] . We used the method described by J . Chen and N . Chatterjee to obtain estimates and p-values for genotype associations assuming Hardy-Weinberg equilibrium in the control population [4] . Phylogenetic trees ( neighbor joining [18] ) were constructed using MEGA 3 . 1 [19] ( http://www . megasoftware . net ) to assess nucleotide similarity of different haplotypes . We evaluated the robustness of our results using the FDR . FDR is the expected ratio of erroneous rejections of the null hypothesis to the total number of rejected hypothesis among all the genes or SNPs analyzed in this report . Rather than using an arbitrary threshold FDR value , we report the values for the most significant associations to allow the reader to evaluate the robustness of our findings . The Benjamini and Hochberg method [20] was used to calculate FDR values using “multtest” package in the R project for statistical analyses ( http://www . r-project . org ) . Unless otherwise specified , statistical analyses were performed with STATA Version 8 . 2 , Special Edition ( STATA Corporation , http://www . stata . com ) .
The accession numbers for the Entrez Gene ( http://www . ncbi . nlm . nih . gov/entrez ) genes discussed in this paper are ARHGDIB ( 397 ) , CASP9 ( 842 ) , CETP ( 1071 ) , CYP1B1 ( 1545 ) , CYP24A1 ( 1591 ) , ESR1 ( 2099 ) , FZD7 ( 8324 ) , GHR ( 2690 ) , GPX4 ( 2879 ) , GSTA4 ( 2941 ) , HSD174 ( 3295 ) , LIPC ( 3990 ) , PLA2G6 ( 8398 ) , PTH ( 5741 ) , RB1CC1 ( 9821 ) , ROS1 ( 6098 ) , SHBG ( 6462 ) , STAT1 ( 6772 ) , STK11 ( 6794 ) , TERT ( 7015 ) , VEGF ( 7422 ) , XRCC4 ( 7518 ) , and ZNF350 ( 59348 ) .
|
This article reports findings from a large-scale evaluation of common variation in candidate genes for cancer to identify variants that influence bladder cancer risk . We first evaluated 1 , 433 common variants within or near 386 genes in a large case-control study in Spain . The most significant finding was the gene coding for the vascular endothelial growth factor ( VEGF ) . To further investigate this finding , we identified markers that captured most common variation in the whole gene . Analyses indicated that variants in regulatory regions of VEGF could modify the risk for developing bladder cancer . This association is biologically plausible since VEGF is critical for the growth of new blood vessels , which is important for tumor development , and its elevated expression in bladder tumors correlates with tumor progression . Future studies are required to confirm these findings , as well as to investigate the mechanisms for the observed associations .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"homo",
"(human)",
"genetics",
"and",
"genomics",
"public",
"health",
"and",
"epidemiology"
] |
2007
|
Large-Scale Evaluation of Candidate Genes Identifies Associations between VEGF Polymorphisms and Bladder Cancer Risk
|
Visceral leishmaniasis has emerged as an important opportunistic disease among patients infected with HIV-1 . Both HIV-1 and the protozoan parasite Leishmania can productively infect cells of the macrophage-dendritic cell lineage . Here we demonstrate that Leishmania infantum amastigotes increase HIV-1 production when human primary dendritic cells ( DCs ) are cocultured together with autologous CD4+ T cells . Interestingly , the promastigote form of the parasite does not modulate virus replication . Moreover , we report that amastigotes promote virus replication in both cell types . Our results indicate that this process is due to secretion of parasite-induced soluble factors by DCs . Luminex micro-beads array system analyses indicate that Leishmania infantum amastigotes induce a higher secretion of several cytokines ( i . e . IL-1α , IL-2 , IL-6 , IL-10 and TNF-α ) and chemokines ( i . e . MIP-1α , MIP-1β and RANTES ) in these cells . Studies conducted with pentoxifylline and neutralizing antibodies revealed that the Leishmania-dependent augmentation in HIV-1 replication is due to a higher secretion of IL-6 and TNF-α . Altogether these findings suggest that the presence of Leishmania within DC/T-cell conjugates leads to an enhancement of virus production and demonstrate that HIV-1 and Leishmania can establish complex interactions in such a cellular microenvironment .
Leishmaniasis is today recognized as one of the world's most important parasitic diseases , threatening 350 million people in 88 countries [1] . At least 20 Leishmania species can induce various clinical manifestations better known as leishmaniasis . Infection by these trypanosomatid protozoan parasites may lead to a variety of symptoms , ranging from simple , self-healing skin ulcers to a severe , life-threatening visceral disease caused by the Leishmania donovani complex , including Leishmania infantum . One of the most neglected diseases , visceral leishmaniasis ( VL ) is the most severe form of leishmaniasis , characterized by fevers , suppressed immunity and hepatosplenomegaly . Of particular interest , VL has also emerged as an important opportunistic infection in individuals also infected with human immunodeficiency virus type-1 ( HIV-1 ) . Indeed , the outbreak of the HIV-1/AIDS pandemic during the past two decades has dramatically modified the natural history of infection by Leishmania in co-infected patients . According to the World Health Organization ( WHO ) , HIV-1/Leishmania co-infection has been reported in 34 countries in several regions of the globe such as Africa , Asia , Europe , and South America . The increase in the number of cases of co-infection arises from the overlap between regions of active HIV-1 transmission , mostly cities and urbanized areas , and the regions in which Leishmania is endemic . VL is prevalent in HIV-1/Leishmania co-infected individuals and is known to promote not only the development of AIDS-defining illness conditions but also its clinical progression , therefore diminishing the life expectancy of HIV-1-infected subjects . For example , in dually infected patients from southern Europe , all of the isolates were Leishmania infantum [1] , [2] . Furthermore , HIV-1 infection increases the risk of developing VL in endemic areas , enhances the probability of relapse and is also associated with a poorer anti-parasitic drug response [1] , [2] . It has been proposed that HIV-1 and Leishmania act in a deadly synergy . HIV-1/Leishmania co-infection exerts cumulative deficiency of the cellular immune response since both agents harm similar immune resources , such as macrophages and dendritic cells , for their reciprocal benefit [3]–[5] . Dendritic cells ( DCs ) are versatile antigen-presenting cells , recognizing and capturing invading pathogens and establishing a vital bridge between innate and adaptive immunity . DCs also play a key role in HIV-1 infection and dissemination . Indeed , it has been reported that HIV-1 is efficiently transferred from immature DCs to CD4+ T cells via different routes [6]–[9] . For example , after the cells capture and bind HIV-1 , a rapid transfer ( i . e . early transfer or trans infection ) occurs when the virus on the surface of immature DCs or located within endosomal compartments is transported to the DC/T-cell synapse and transmitted directly to CD4+ T cells . This event is followed by a second phase ( i . e . late transfer or cis infection ) that is dependent on productive virus infection of immature DCs and on the eventual transfer of progeny viruses to CD4+ T cells . It has been proposed that after virus capture or uptake , immature DCs ( iDCs ) located in submucosal tissues migrate to lymphoid tissues and become mature DCs ( mDCs ) [10]–[12] , which can potently present nominal antigens to CD4+ T cells in lymphoid tissues . Although the possible multifaceted interactions between HIV-1 and Leishmania have been studied in macrophages [13]–[15] , there is still very little information on the consequences of infection of DCs with these two microorganisms . In macrophages , Leishmania increases both HIV-1 gene transcription and release of progeny virus through production of proinflammatory cytokines . On the other hand , some studies have suggested that DCs can be infected by either Leishmania promastigotes or amastigotes [16] , [17] . It has also been reported that competition between HIV-1 and Leishmania amastigotes for DC-SIGN binding has an impact on the HIV-1 life cycle when DCs were first exposed to the parasite before inoculation with HIV-1 [18] . Proinflammatory cytokines play an important role in both innate and adaptive immune responses against viruses and intracellular pathogens . Interestingly , TNF-α has been reported to stimulate HIV-1 replication in the promonocytic U1 cell line through NF-κB activation and subsequent trans-activation of the viral regulatory elements [19] . In addition , IL-6 has been reported to stimulate HIV-1 replication in macrophages again via a NF-κB-mediated signal transduction pathway [20]–[22] . An increased production of IL-6 and TNF-α has been detected in Leishmania-infected macrophages [23] , [24] . Thus , in the case of HIV-1/Leishmania co-infected patients , it is likely that HIV-1 uses a part of the Leishmania-mediated cytokine network to its own advantage . In the present work , we investigated the influence of Leishmania infection on the biology of HIV-1 in physiologically relevant human DCs . We demonstrate here for the first time that Leishmania infantum amastigotes promote HIV-1 replication in both DCs and autologous CD4+ T cells when these two cell subpopulations are cultured together . The mechanism responsible for this up-regulatory effect is linked with secretion of parasite-induced soluble factors by DCs .
Human primary DCs were generated from monocytes obtained from the blood of healthy donors , in accordance with the guidelines of the Bioethics Committee of the Centre Hospitalier de l'Université Laval Research Center . A written consent was obtained from all blood donors . Recombinant human interleukin-2 ( rhIL-2 ) , efavirenz ( EFV ) and azidothymidine ( AZT ) were obtained from the NIH AIDS Repository Reagent Program ( Germantown , MD ) . Interferon-gamma ( IFN-γ ) and IL-4 were purchased from R&D Systems ( Minneapolis , MN ) , whereas granulocyte macrophage–colony stimulating factor ( GM-CSF ) was a generous gift from Cangene ( Winnipeg , MB ) . Lipopolysaccharide ( LPS ) and phytohemagglutinin-L ( PHA-L ) were purchased from Sigma ( St-Louis , MO ) . The culture medium consisted of RPMI-1640 supplemented with 10% foetal bovine serum ( FBS ) , penicillin G ( 100 U/mL ) , streptomycin ( 100 U/mL ) and glutamine ( 2 mM ) , which were all purchased from Wisent ( St-Bruno , QC ) , and primocine ( Amaxa Biosystems , Gaithersburg , MD ) . In brief , CD14+ cells ( i . e . monocytes ) were isolated from peripheral blood mononuclear cells , using a monocyte-positive selection kit according to the manufacturer's instructions ( CD14-positive selection kit; StemCell Technologies Inc . , Vancouver , BC ) and cultured in six-well plates ( 106 cells/ml ) . Immature DCs ( iDCs ) were generated from these purified monocytes by a treatment with GM-CSF ( 1 , 000 U/ml ) and IL-4 ( 200 U/ml ) for 7 days . The maturation of iDCs was induced on the fifth day by culturing them for 48 h with the above-described cytokines supplemented with IFN-γ ( 1 , 000 U/ml ) and LPS ( 100 ng/ml ) . The final phenotype of iDCs and mature monocyte-derived DCs ( mDCs ) was monitored by flow cytometry ( data not shown ) . For example , iDCs expressed HLA-DR , CD86 , DC-SIGN , CD1a and low levels of CD14 , whereas mDCs expressed CD83 and high levels of ICAM-1 , HLA-DR , and CD86 but lower levels of DC-SIGN and CD14 compared to iDCs ( data not shown ) . iDCs were considered effectively differentiated based on the loss of CD14 and acquisition of CD1a and DC-SIGN ( data not shown ) . Autologous CD4+ T cells were isolated with a negative selection kit ( StemCell Technologies ) and activated ( 2×106 cells/ml ) with the mitogenic agent PHA-L ( 1 µg/ml ) and rhIL-2 ( 30 U/ml ) for 48 h prior to their use . Viruses were produced upon transient calcium phosphate transfection of 293T cells either with pNL4-3Balenv , an R5-tropic infectious molecular clone of HIV-1 [25] , or pNLBalHSA-IRES ( see below for more details ) . Virus stocks were normalized for virion content by using a sensitive in-house , double-antibody sandwich enzyme-linked immunosorbent assay specific for the major viral core p24 protein [26] . Viral preparations underwent a single freeze-thaw cycle before being used in subsequent experiments . The pNLBalHSA-IRES molecular construct has been described previously [27] and was obtained by replacing the eGFP gene in the NLENG1-IRES vector ( NL4-3 backbone ) with the coding sequence for mouse heat stable antigen ( HSA ) and replacing the X4-tropic env gene of NL4-3 with that of the R5-using env gene of NL4-3Balenv . The Leishmania infantum strain MHOM/MA/67/ITMAP-263 was maintained at 27°C by a weekly passage in RPMI-1640 supplemented with 10% FBS , buffered with 25 mM HEPES and 2 mM NaHCO3 , containing 5 µg/ml haemin and antibiotics . Axenic Leishmania infantum amastigotes were differentiated in vitro from stationary-phase promastigotes . The culture and maintenance of axenic amastigotes have been described previously [28] . These amastigotes showed morphological , biochemical and biological characteristics similar to those of amastigotes isolated in vivo [28] . DCs ( both iDCs and mDCs ) were first pulsed with NL4-3Balenv ( 10 ng of p24/105 cells ) for 60 min at 37°C and unbound virus was eliminated by extensive washes with phosphate-buffered saline ( PBS ) . Next , DCs were exposed to Leishmania infantum promastigotes or axenic amastigotes at a parasite/cell ratio of 10∶1 for up to 4 h and free parasites were washed out with warm PBS . Cells were then incubated with activated CD4+ T cells at a 1∶3 ratio ( DCs∶CD4+ T cells ) . Viral production was assayed by measuring the cell-free p24 content at different time intervals . In some experiments , DCs were treated either with EFV ( 50 nM ) or AZT ( 10 µM ) for 30 min before pulsing with virions . In virus transfer experiments aimed at evaluating the percentage of virus-infected CD4+ T cells , we used NLBalHSA-IRES virus . Three days after initiation of the coculture , cells were stained either with an isotype-matched irrelevant control antibody or anti-CD3 ( American Type Culture Collection , Manassas , VA ) followed by an FITC-conjugated goat anti-mouse IgG ( Jackson ImmunoResearch Laboratories , West Grove , PA ) and biotin-conjugated anti-HSA followed by R-PE-labelled streptavidin ( the latter two from BD Pharmingen , San Diego , CA ) . Stained cells were fixed with 2% paraformaldehyde for 30 min at 4°C and then analyzed by flow cytometry ( Epics ELITE ESP; Coulter Electronics , Burlington , VA ) . In some HIV-1 transfer studies , permeable cell inserts with polycarbonate membranes ( Transwell , Corning Inc . , Lowell , MA ) ( pore size: 1 µm ) were used to separate DCs and CD4+ T cells . iDCs were exposed ( or not ) to parasites for 3 h , washed with PBS and transferred into a culture plate . Next , activated autologous CD4+ T cells were pulsed with HIV-1 for 2 h , washed with PBS and either cocultured directly with iDCs or transferred into permeable cell inserts and cocultured with iDCs for 3 days . Virus production was assessed by estimating the p24 content . De novo virus production in iDCs was monitored by incubating the cells for 1 h at 37°C with NL4-3Balenv ( 10 ng of p24/105 cells ) . After three washes with PBS , the cells were maintained in complete culture medium without autologous CD4+ T cells . Virus production was estimated by quantifying the p24 content in cell-free culture supernatants . In some studies , autologous activated CD4+ T cells were inoculated with NL4-3Balenv for 2 h and excess virus washed off with PBS . Next , iDCs were either left untreated or exposed to Leishmania infantum amastigotes for 3 h , washed , and cocultured with HIV-1-exposed autologous CD4+ T cells at a 1∶3 ratio . Virus production was determined by measuring p24 levels in the culture supernatants . To analyze the effect of DC-derived soluble factors on viral production in CD4+ T cells , 0 . 22 µm-filtered culture supernatants ( Millipore , Billerica , MA ) were used in parallel experiments . Supernatants were harvested from unstimulated ( control ) or amastigote-pulsed iDCs after 24 h of culture . Next , activated autologous CD4+ T cells were inoculated with NL4-3 Balenv for 2 h and excess virus was washed off with PBS . Infected CD4+ T cells were then incubated for 3 days with filtered supernatants from DCs . A commercial multiplex cytokine/chemokine assay that can detect and quantify different cytokines and chemokines ( i . e . IL-1α , IL-2 , IL-4 , IL-6 , IL-10 , IL-12 , IL-15 , TNF-α , IFN-γ , MIP-1α , MIP-1β and RANTES ) through the use of the Luminex 100 apparatus was purchased from Bio-Rad ( Wilmington , DE ) . The Luminex technology is a bead array cytometric analyzer designed to study numerous analytes simultaneously by using spectrally distinct beads in a single well of a microtiter plate , using very small sample volumes ( i . e . as little as 25 µl ) . Briefly , iDCs were either left unexposed or exposed to Leishmania infantum amastigotes for 4 h and free parasites were removed with warm PBS . Cell-free supernatants were harvested from unstimulated ( control ) or Leishmania-exposed iDCs after 24 h of culture . Quantification was achieved by measuring concentrations of the studied cytokines/chemokines in cell-free supernatants according to the manufacturer's instructions ( Bio-Rad ) . The TNF-α inhibitor pentoxifylline ( PTX ) ( 1- ( 5′-oxohexyl ) -3 , 7-dimethylxanthine; Sigma-Aldrich; 100 µM final concentration ) , neutralizing anti-TNF-α ( R&D Systems; 500 ng/ml final concentration ) and neutralizing anti-IL-6 ( R&D Systems; 10 µg/ml final concentration ) were added to the culture medium when initiating virus transmission or during acute infection studies and then every 2 or 3 days with each medium change . The statistical significance of the results was defined by performing a one-way ANOVA analysis of variance with Dunnett's post-tests to compare treated versus untreated control samples . All analyses were performed on raw data ( i . e . p24 concentrations ) . P values lower than 0 . 05 were considered statistically significant . InStat software ( version 3 . 05; GraphPad Software ) was used for all analyses .
In an attempt to provide precious information on possible relationships between Leishmania and HIV-1 , we studied the effect of Leishmania infantum on HIV-1 replication in the context of cocultures constituted of DCs and autologous CD4+ T cells . To this end , both iDCs and mDCs were first pulsed with R5-tropic virus and next exposed to Leishmania infantum amastigotes and promastigotes . Thereafter , DCs were cocultured with autologous CD4+ T cells . A statistically significant enhancement of virus replication was observed when iDCs were exposed to the amastigote form of the parasite ( Fig . 1A ) . Similar observations were made when coculture experiments were performed with mDCs ( Fig . 1B ) . Since promastigotes and killed amastigotes have not elicited any significant response in our experimental setup , all further experiments were performed only with the most pathologically-relevant stage of the Leishmania parasite in the context of HIV-1 infection , i . e . live amastigotes . The successful infection of DCs with parasites was monitored using Leishmania infantum-based reporter amastigotes as we described previously [15] . Laser-scanning confocal microscopy was used to detect infection with GFP-encoding parasites while a luminometer was used to measure infection with luciferase-encoding amastigotes ( data not shown ) . To define the mechanism ( s ) by which the parasite can promote HIV-1 replication in cocultures , iDCs were treated with the antiretroviral drug EFV during the virus pulsing period . This treatment will abrogate de novo virus production in iDCS without affecting transfer of viruses located on their surface or within their endosomal apparatus and the subsequent replication in CD4+ T cells . As expected , treatment with EFV reduces HIV-1 replication in cocultured cells ( Fig . 2 ) and completely abolished virus production in iDCs cultured alone ( data not shown ) . However , the parasite-mediated increase in HIV-1 transmission was detected in untreated and EFV-treated cells , therefore suggesting that Leishmania amastigotes are promoting virus production in both iDCs and CD4+ T cells . Similar observations were made when experiments were carried out with AZT ( data not shown ) , which is another potent reverse transcriptase inhibitor . Next , to confirm that the parasite is affecting de novo virus production in iDCs , these cells were first exposed to HIV-1 and next to Leishmania promastigotes , and maintained in culture for 3 days before addition of autologous CD4+ T cells . The rationale for this experimental setup is based on the previous report that endosome-associated virions in DCs remain infectious for no more than 2 days [9] . An increase in HIV-1 transfer was still seen in the case of Leishmania-treated samples ( Fig . 3A ) . To substantiate what seems to be a more productive HIV-1 infection of iDCs in the presence of Leishmania amastigotes , cells were pulsed with fully competent virus in the absence or presence of amastigotes and virus replication in iDCs cultured alone was evaluated over time . Results depicted in Fig . 3B indicate that the parasite induces a significant increase in de novo virus production in iDCs . To provide additional information on how the presence of amastigotes can modulate virus replication in a coculture system made of iDCs and autologous CD4+ T cells , we assessed whether the observed increase in virus production is reflected by a corresponding augmentation in the number of CD4+ T cells productively infected with HIV-1 . To this end , coculture experiments were carried out with NLBalHSA-IRES virus which , upon acute infection of target cells , will produce all viral proteins and also a cell surface reporter molecule ( i . e . the murine heat-stable antigen/HSA ) . Unlike most reporter viruses , fully infectious NLBalHSA-IRES virions have no deletions in the env , vpr or nef genes , and allow for the easy detection of productively infected cells through the surface expression of the HSA molecule [27] . Pulsing iDCs with NLBalHSA-IRES virus followed by a coculture step with autologous CD4+ T cells for 3 days resulted in a proportion of CD4+ T cells productively infected with HIV-1 ranging from 7 . 8 to 9 . 9% as monitored by flow cytometry measuring the relative percentage of HSA and CD3 double-positive cells ( Fig . 4 ) . This percentage of virus-infected cells was not affected when amastigotes were present during the pulsing step of iDCs ( i . e . percentages of HSA-expressing cells ranging from 8 . 1% to 10 . 2% ) ( mean of 9 . 03% for cells infected with HIV-1 only compared to 9 . 36% for cells inoculated with both HIV-1 and Leishmania: P = 0 . 73 ) . This observation suggests that Leishmania amastigotes do not affect the number of CD4+ T cells which are productively infected with HIV-1 , at least at an early time period following initiation of the coculture ( i . e . 3 days ) . Therefore , it can be proposed that Leishmania is most likely affecting virus production in the CD4+ T cell population by up-regulating virus gene expression . To validate that HIV-1 production in CD4+ T cells is enhanced upon a coculture with Leishmania amastigotes-exposed iDCs , we performed another set of experiments where autologous CD4+ T cells were first pulsed with HIV-1 and then cocultured with iDCs either left untreated or inoculated with parasites . Results from Fig . 5A demonstrate that viral replication in CD4+ T cells is augmented by a coculture step with Leishmania-loaded iDCs . Moreover , we investigated whether a cell-to-cell contact is necessary to achieve the observed phenomenon . To do so , we used permeable cell supports with a membrane pore size of 1 µm , which allows the crossing of virions and soluble factors but not that of cells [29] . When iDCs were separated from CD4+ T cells by a permeable membrane , HIV-1 replication decreased in comparison to the contact-favoured cocultured cells ( Fig . 5A ) . However , virus production was still augmented by Leishmania amastigotes even when iDCs and CD4+ T cells were not allowed to interact with each other . Thus , these results suggest that a soluble factor ( s ) secreted by Leishmania-pulsed DCs drives HIV-1 replication in cocultured cells . Furthermore , when HIV-1-inoculated autologous CD4+ T cells were incubated with cell-free supernatants from Leishmania-exposed iDCs , an increase in virus replication was seen in CD4+ T cells compared to those incubated with cell-free supernatants from unexposed iDCs ( Fig . 5B ) . Interestingly , heat-inactivated supernatants from Leishmania-exposed iDCs had no effect on viral replication in CD4+ T cells . These observations suggest that the Leishmania-directed augmentation in HIV-1 production in cocultures made of iDCs and autologous CD4+ T cells is due to a heat-sensitive soluble factor ( s ) secreted by iDCs . In order to identify the Leishmania amastigote-induced soluble factor ( s ) responsible for promoting virus replication in both DCs and CD4+ T cells , we quantified the levels of several cytokines and chemokines produced by unexposed and Leishmania-exposed iDCs using the Multi Analyte Profiling bead array technology from Luminex . Supernatants from Leishmania-pulsed iDCs display higher levels of various cytokines and chemokines ( i . e . IL-1α , IL-6 , IL-10 , TNF-α , MIP-1α , MIP-1β and RANTES ) than those from untreated iDCs ( Fig . 6 ) . However , because of the strong variability between the samples tested , only IL-6 and TNF-α concentrations were increased in a statistically significant manner by Leishmania infantum amastigotes . It has been previously reported that treatment of HIV-1-infected cells with IL-6 and TNF-α results in a significant induction of virus gene expression in a variety of cell systems directly through nuclear translocation of NF-κB [20]–[22] , [30] . Therefore , in an attempt to define the relative contribution of these two proinflammatory cytokines to the Leishmania-mediated stimulatory effect on virus production , transfer studies were performed with PTX , a phosphodiesterase inhibitor that abolishes TNF-α production [31] , and blocking antibodies . Data from Fig . 7A show a direct involvement of both IL-6 and TNF-α in the Leishmania-dependent augmentation of HIV-1 replication in cocultured cells . Interestingly , virus production was enhanced upon addition of exogenous IL-6 and TNF-α to cocultured cells , which further implicate both cytokines in the process . Similar patterns were seen when iDCs were cultured alone in absence of autologous CD4+ T cells and treated with the listed reagents ( Fig . 7B ) . Altogether , these data indicate that Leishmania amastigotes mediate a more important production of IL-6 and TNF-α by iDCs , which in turn stimulates HIV-1 replication in both iDCs and CD4+ T cells .
The multifaceted interplay between HIV-1 and Leishmania in macrophages has been previously addressed [13] , [15] . However , the macrophage is not the only cell type known to harbour both pathogens . Indeed , previous studies have established that DCs are also susceptible to HIV-1 and Leishmania infection ( i . e . primarily by amastigotes and marginally by promastigotes ) [18] , [32] , [33] . Unfortunately , there is a paucity of data on the consequences of DCs infection with these two microorganisms . A previous study has reported that Leishmania reduces HIV-1 capture and transfer by DCs upon pre-incubation with Leishmania amastigotes because both pathogens use the cell surface molecule DC-SIGN as a portal of entry in these professional antigen-presenting cells [18] . In the present work , we show that an initial pulsing of DCs with HIV-1 followed by exposure to the amastigote form of the parasite leads to an increased HIV-1 replication in both DCs and CD4+ T cells when these two cell subtypes are cocultured together . Interestingly , the promastigote developmental stage of the parasite does not affect virus production under similar experimental procedures . Previous studies have concluded that numerous glycoconjugates that are differentially expressed on the cell surface of the parasite in its promastigote stage ( e . g . lipophosphoglycan , leishmanolysin/gp63 , low molecular weight glycoinositolphospholipids and a membrane proteophosphoglycan ) could be involved in the binding of parasites to target cells [34] , but we can conclude that these promastigote-specific molecules have no direct role in the parasite-mediated effect on HIV-1 biology in the DC/CD4+ T cell milieu . Moreover , our results are in agreement with other studies showing that amastigotes are internalized by DCs more efficiently than promastigotes [18] , [32] , [33] , therefore validating the hypothesis that engulfment of Leishmania parasites is controlled by several factors , such as DC subtypes , Leishmania species and parasite stages [16] , [17] , [35] . It is now well established that DCs can efficiently transmit HIV-1 to neighbouring CD4+ T cells through both trans- and cis-infection pathways . Trans-infection mediated by DCs can occur across infectious synapses [8] , [9] , [11] and via exocytosis of virus-associated exosomes [36] , whereas the cis-infection route relies on de novo virus infection of DCs and long-term transmission of HIV-1 [7] , [9] , [37] , [38] . It is conceivable that both of these mechanisms may coexist and contribute to viral dissemination . In our experimental model , we provide evidence that Leishmania amastigotes are affecting HIV-1 replication at different levels in a coculture system consisting of DCs ( either iDCs or mDCs ) and autologous CD4+ T cells . Indeed , we demonstrate that the parasite increases virus production in both DCs and CD4+ T cells when these two cell types are cocultured together . Moreover , we report that the percentage of CD4+ T cells productively infected by HIV-1 does not seem to be affected by Leishmania amastigotes at an early time point after the initiation of the coculture ( i . e . 3 days ) . It can be proposed that the parasite is not modulating the viral transfer process between DCs and CD4+ T cells but is instead positively regulating virus gene expression in each cell type . Interestingly , our data indicate that a close contact between Leishmania-pulsed DCs and CD4+ T cells is not mandatory to stimulate HIV-1 production , therefore suggesting that the observed increase in virus replication is achieved through the secretion of a soluble factor ( s ) . Elevated plasma protein levels of proinflammatory cytokines and chemokines such as IL-1 , IL-6 and TNF-α have been detected in VL patients [39] . Based on this information , it can be proposed that Leishmania may modulate HIV-1 replication in our experimental coculture cell system by promoting secretion of proinflammatory cytokines known to amplify virus gene expression . This hypothesis was corroborated by a multiplex fluorescent microfluid immunoassay , which revealed that exposure of DCs to Leishmania amastigotes increases secretion of IL-6 and TNF-α . Coculture studies carried out in the presence of neutralizing antibodies specific for IL-6 or TNF-α showed that these two cytokines have a major role in the observed Leishmania-induced increase in HIV-1 replication in cocultured cells . A previous study has shown that lipophosphoglycan , a major surface molecule of Leishmania , acts as an effective activator of HIV-1 expression in a T lymphoid cell line latently infected with HIV-1 through secretion of TNF-α [40] . It is of interest to note that TNF-α seems to be involved in granuloma formation and the control of parasite growth [41] . For example , anti-TNF-α treatment resulted in the reactivation of VL in patients being treated for arthritis , suggesting a protective role of TNF-α against Leishmania infection [42] . On the other hand , when produced at very high levels , TNF-α might have a disease-enhancing effect . For example , one study identified a link between VL and an allelic polymorphism associated with elevated serum levels of TNF-α [43] . High levels of TNF-α promote the generation of IL-10-producing T cells as a homeostatic response to excessive inflammation [44] . It is also well-established that TNF-α significantly increases HIV-1 replication in cells of the macrophage lineage through nuclear translocation of NF-κB [19] , [45] , [46] . Infection of monocyte-derived macrophages ( MDMs ) with HIV-1 also induces TNF-α secretion resulting in a positive autocrine loop that enhances virus production [47] . Thus , it can be suggested that the Leishmania-mediated release of TNF-α by DCs may function in an autocrine/paracrine manner to induce virus gene expression in cells harboring HIV-1 . The same applies to IL-6 as it has been reported to augment HIV-1 replication in MDMs as well as in the latently infected U1 monocytoid cell line . Furthermore , IL-6 has been shown to potentiate TNF-α-induced upregulation of HIV-1 production and induction of NF-κB [21] , [22] . Previous studies have shown that PTX , as an effective inhibitor of protein kinase C , protein kinase A and NF-κB , selectively inhibits TNF-α production , as well as HIV-1 transcription and virus production [48] , [49] . We show in our study that the parasite-dependent enhancement in HIV-1 production in cocultured cells is no longer seen in presence of PTX . Our observations are in line with a previous report describing that PTX , a methylxanthine derivative that inhibits TNF-α synthesis , diminishes plasma viral load and improves cell-mediated immunity in HIV-1-infected individuals [50] . There is a possibility that PTX might also negatively affect NF-κB signaling in addition to TNF-α production , which could in turn reduce virus gene expression . However , the contribution of TNF-α in the parasite-dependent augmentation of virus production in cocultured cells is confirmed when using blocking anti-TNF-α antibodies . Results presented herein thus show that Leishmania amastigotes alter the complex cellular cytokine network and promote secretion of the proinflammatory cytokines IL-6 and TNF-α . Although our data suggest a major role for IL-6 and TNF-α in the Leishmania-induced effect on HIV-1 replication in a coculture system constituted of DCs and CD4+ T cells , we cannot eliminate the possibility that other soluble factors are involved , either upstream or downstream from them . HIV-1 and Leishmania are the cause of a vast array of immunological disorders . Both infections alter the predominant cellular immune response through complex cytokine-mediated mechanisms that confer susceptibility to both infections . DCs are proposed to have a dominant role in the early events of HIV-1 transmission by transporting the virus from peripheral sites to lymphoid compartments . Moreover , DCs act also as a natural and stable cellular reservoir in the natural course of Leishmania infection . The present study illustrates that the effect of Leishmania on HIV-1 replication in a coculture system composed of DCs and CD4+ T cells largely depends on the developmental stage of the parasite and the chronology of infection of DCs by the two pathogens . We demonstrate that Leishmania amastigotes enhance the process of HIV-1 replication by favouring a higher production of the proinflammatory cytokines IL-6 and TNF-α . Data from the current study lead us to speculate about the putative role played by Leishmania in the context of a co-infection with HIV-1 . The various immunological disturbances caused by Leishmania in DCs , a cell population recognized as both a virus and parasite reservoir , can be considered as detrimental to the host by promoting HIV-1 replication and progression of the disease . This assumption is validated by clinical studies showing that leishmaniasis enhances the viral load and reduces life expectancy in HIV-1-infected patients [51]–[53] . In conclusion , this analysis provides novel insights into the complex interconnections between HIV-1 and Leishmania and presents unique information that may facilitate the development of more effective therapeutic strategies aimed at controlling disease progression in dually infected individuals .
|
Visceral leishmaniasis ( VL ) is a potentially deadly parasitic disease afflicting millions worldwide . Although itself an important infectious illness , VL has also emerged as an opportunistic disease among patients infected with HIV-1 . This is partly due to the increasing overlap between urban regions of high HIV-1 transmission and areas where Leishmania is endemic . Furthermore , VL increases the development and clinical progression of AIDS-related diseases . Conversely , HIV-1-infected individuals are at greater risk of developing VL or suffering relapse . Finally , HIV-1 and Leishmania can both productively infect cells of the macrophage-dendritic cell lineage , resulting in a cumulative deficiency of the immune response . We therefore studied the effect of Leishmania infantum on HIV-1 production when dendritic cells ( DCs ) are cocultured with autologous CD4+ T cells . We show that amastigotes promote virus replication in both DCs and lymphocytes , due to a parasite-mediated production of soluble factors by DCs . Micro-beads array analyses indicate that Leishmania infantum amastigotes infection induces a higher secretion of several cytokines in these cells , and use of specific neutralizing antibodies revealed that the Leishmania-induced increase in HIV-1 replication is due to IL-6 and TNF-α . These findings suggest that Leishmania's presence within DC/T-cell conjugates leads to an enhanced HIV-1 production .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"Methods",
"Results",
"Discussion"
] |
[
"immunology/cellular",
"microbiology",
"and",
"pathogenesis",
"infectious",
"diseases/neglected",
"tropical",
"diseases",
"virology/immunodeficiency",
"viruses",
"virology/mechanisms",
"of",
"resistance",
"and",
"susceptibility,",
"including",
"host",
"genetics",
"infectious",
"diseases/hiv",
"infection",
"and",
"aids",
"infectious",
"diseases/protozoal",
"infections"
] |
2009
|
Leishmania infantum Amastigotes Enhance HIV-1 Production in Cocultures of Human Dendritic Cells and CD4+ T Cells by Inducing Secretion of IL-6 and TNF-α
|
Traditional protein annotation methods describe known domains with probabilistic models representing consensus among homologous domain sequences . However , when relevant signals become too weak to be identified by a global consensus , attempts for annotation fail . Here we address the fundamental question of domain identification for highly divergent proteins . By using high performance computing , we demonstrate that the limits of state-of-the-art annotation methods can be bypassed . We design a new strategy based on the observation that many structural and functional protein constraints are not globally conserved through all species but might be locally conserved in separate clades . We propose a novel exploitation of the large amount of data available: 1 . for each known protein domain , several probabilistic clade-centered models are constructed from a large and differentiated panel of homologous sequences , 2 . a decision-making protocol combines outcomes obtained from multiple models , 3 . a multi-criteria optimization algorithm finds the most likely protein architecture . The method is evaluated for domain and architecture prediction over several datasets and statistical testing hypotheses . Its performance is compared against HMMScan and HHblits , two widely used search methods based on sequence-profile and profile-profile comparison . Due to their closeness to actual protein sequences , clade-centered models are shown to be more specific and functionally predictive than the broadly used consensus models . Based on them , we improved annotation of Plasmodium falciparum protein sequences on a scale not previously possible . We successfully predict at least one domain for 72% of P . falciparum proteins against 63% achieved previously , corresponding to 30% of improvement over the total number of Pfam domain predictions on the whole genome . The method is applicable to any genome and opens new avenues to tackle evolutionary questions such as the reconstruction of ancient domain duplications , the reconstruction of the history of protein architectures , and the estimation of protein domain age . Website and software: http://www . lcqb . upmc . fr/CLADE .
The evolutionary history of eukaryotic proteins involves rapid sequence divergence , addition and deletion of protein domains , fusion and fission of genes . This implies that protein repertoires of distantly related species differ greatly ( new architectures , that is combinations of domains , are many ) , while domain repertoires do not ( new domains are few ) [1] . To account for the great diversity of domain contexts in eukaryotes , an effort was made to categorize coding regions into protein domains and domain families . An important contribution issued by this effort is the Pfam database [2] , a large collection of protein families , each represented by multiple sequence alignments ( MSAs ) and hidden Markov models ( HMMs ) . Pfam also provides protein domain architectures and higher-level groupings of related domains ( clans ) . It highlighted that different architectures give rise to the diversity of proteins found in nature , and that identifying domains of a protein can provide insights into its function . Hence , the complexity of protein annotation can be simplified by focusing on domains , even though the problem of unraveling domain organization ( domain architecture ) still remains . Proteins sharing more than 30% of sequence identity have a high probability also to share the same fold [3 , 4] . Thus , since fold and function of a protein have generally an intimate relationship [5] , strong sequence similarity is exploited by conventional alignment methods to reconstruct families of functionally related proteins and to accomplish genome annotations . Unfortunately , the complete sequencing of several organisms differing in physiology , habitat and genetics , as Plasmodium falciparum , brought to light the weakness of homology-based approaches to annotation [6–8] . This limitation is challenged even more by the large prokaryotic and eukaryotic metagenomic samples generated today . Radically new homology-based annotation approaches combine information from physico-chemical properties of sequences and conserved amino acid positions in MSAs , together with sophisticated inference methods to extract a general profile ( typically a profile HMM , in short pHMMs [9 , 10] ) of a protein domain family and use it as a signature for homology detection [8 , 11–17] . This profile represents a consensus of signals characterizing a given domain in a multitude of different species . We shall speak of sequence consensus model , in short SCM . Conversely , other approaches [18–20] associate to each protein domain family several different profiles , built from a sample of diversified homologous sequences . The resulting set of profiles , for all protein domain families , is approximately six or seven times larger than the number of domain families , depending on the method . For instance , SUPERFAMILY [18] constructs 15 438 models for 1 962 SCOP superfamilies [21] , and Gene3D [20 , 22] constructs 16 933 models for 2 738 domain families . Such sets of domain families are rather small compared to the number of distinct Pfam domains ( 14 831 for Pfam version 27 ) and one would like to generate and handle multiple models representing all Pfam domains . The co-occurrence of domains within a protein was also shown to be very powerful to accurately identify domains in divergent protein sequences [23] and especially for the P . falciparum genome [24–27] . Nevertheless , the 37% of P . falciparum proteins still completely lack domain annotation and one of the main reasons is that relevant signals in sequences might become too weak to be identified by consensus if sequence divergence is too important or if the pool of sequences is biased ( too small or overrepresented by certain clades ) . Another reason is that proteins without predicted domains might belong to novel families completely missing in Pfam and other databases . Based on the observation that structural and functional constraints might not be globally conserved through all species , we propose a novel pipeline , called CLADE ( CLoser sequences for Annotations Directed by Evolution ) , that identifies domains in proteins by using all known Pfam domains and the large quantity of available genomic data spanning through a large panel of species . The idea is to “decompose” the signal of consensus shared by homologous sequences , collected at the scale of the entire phylogenetic tree , into several consensus signals coming from homologous sequences collected at the scale of species within clades ( we shall speak of clade-centered models , in short CCM ) , and possibly of species that are phylogenetically distant from the genome considered . To do this , we construct several profiles for each Pfam domain , starting from a large and differentiated panel of homologous sequences in a protein domain family , and we use CCMs and SCMs to search for homologous sequences in the genome to annotate . The outcomes of these models are processed and transformed into features used to train a meta-classifier , that is a Support Vector Machine ( SVM ) [28] , that assigns a confidence score to each domain prediction . Based on this score ( defined later ) and on other properties , such as domain co-occurrence , CLADE finds the most probable architecture for each protein sequence by using DAMA [27] , a tool that finds best domain architectures based on multi-objective optimization criteria . By using high performance computing ( HPC ) , CLADE demonstrates that the limits in annotation reached by current methods can be bypassed . In fact , HPC makes it possible to construct and explore a large number of profiles ( a few millions ) and to search , within them , for the appropriate evolutionary patterns that match the protein sequences to annotate . The idea to explore a large space of profiles and to combine the information coming from them for the prediction of domains within proteins has never been developed before and turned out to be a winning strategy that opens up new hopes and directions for the development of accurate annotation systems . In CLADE , each domain is represented by about three hundred and fifty profiles , a number that is hundreds times larger than what has been previously proposed [18–20] . In total , about 2 . 5 million profiles are used to annotate a genome . The information issued by these models is merged and CLADE finds agreement among models , takes into consideration their phylogenetic origin , and combines potential annotations in an accurate reconstruction of domain architectures . Both the profile-based search and the analysis of the resulting annotations require HCP . By using grid computing , we could annotate proteins lying in the twilight zone [29] of the P . falciparum genome , remarkably rich in A and T , and advance on the fundamental question of how to identify domains for proteins that are highly diverged . The P . falciparum protein annotation represents a very difficult case study . More than 44% of open reading frames in this genome remains without any putative annotation [30–32] in PlasmoDB v11 . 1 that reports 2 464 proteins with unknown or hypothetical function over 5 542 genes . Our methodology shows a striking improvement over current annotation approaches . No specific property of the P . falciparum genome , besides its localization within the Alveolata clade , is used and the method can be applied to any other genome . We show that because of their closeness to actual protein sequences , CCMs are more specific and more functionally predictive than the broader Pfam family models , based on consensus . The accurate domain annotation reached with CCMs has important implications also for protein evolution studies . In fact , we show that CCMs provide novel information that can be exploited to explore the landscape of protein domain evolution . We shall argue that CCMs can help to trace ancient domain duplications , to reconstruct the history of protein architectures , and to better estimate protein domain age .
To predict domains in protein sequences , CLADE pipeline is organized in three main steps ( Fig 1 ) . The input is a set of protein sequences coming from the same genome or from different ones . In the latter case , each sequence is accompanied by its NCBI taxon code . To demonstrate that “multi-source” domain modeling is more appropriate than “mono-source” domain modeling for capturing remote homology , we consider several datasets: 1 . three datasets of single-domain sequences constructed from the SCOP database [21] as gold standard , 2 . four datasets of randomly generated single-domain sequences satisfying different statistical hypotheses , 3 . two datasets of randomly generated multi-domain sequences satisfying different statistical hypotheses . We evaluate the improvement in using an annotation that exploits multi-source domain modeling by comparing the performance of CLADE , that favors the multi-source strategy , to the performance of HMMScan and HHblits , that favor the mono-source one . We recall that HMMScan and HHblits are widely used search methods based on sequence-profile and profile-profile comparison , respectively . We estimate the false discovery rate ( FDR ) for the three methods . Also , with the SCOP datasets , we show that agreement among models is an important feature present in CLADE , and missing in existing multi-source domain approaches , like SUPERFAMILY . Note that CLADE skips its third step ( DAMA ) on datasets 1 and 2 because they comprise single-domain sequences . Upheld by CLADE performance on the SCOP benchmark datasets and the FDR tests , we checked whether CLADE could significantly contribute to the identification of domains within the full set of P . falciparum proteins . A large number ( 2464 ) of P . falciparum proteins has no identified domain in PlasmoDB and it remains with no domain identification ( 2068 ) even after the analysis of existing in silico predictive methods [2 , 24 , 26] . We performed a large scale domain prediction and compared CLADE results against annotations obtained with HMMScan . All evaluations reported below were realized with the same CLADE parameters ( with an E-value cut-off at 1e-3 and by adopting a specific SVM probability cut-off for each domain ) . See section “CLADE pipeline , parameter settings and tools used in CLADE” in Methods .
When sequences in a protein domain family are too divergent , signals of homology are easier to trace with CCMs rather than with models based on the consensus of homologous sequences spanning the entire phylogenetic tree ( reminiscent ideas were presented in [7 , 18–20 , 45] ) . CLADE shows that , by combining in a unique tool CCMs and consensus models , the predictive power can be highly reinforced . In CLADE , CCMs have been constructed for domains in the Pfam database . In this respect , it is worth to highlight that CCMs are basically different from: 1 . Pfam models characterizing Pfam clans [46] whose purpose is to group different families together , possibly new protein families that appear to have arisen from a single evolutionary origin; 2 . Pfam models characterizing Pfam domain families ( called SCMs in CLADE ) . Compared to both Pfam family models , CCMs are much closer to actual protein sequences , therefore more specific and more functionally predictive . In fact , one can think of CCMs as forming a new layer of models situated at the very bottom of the Pfam hierarchy and associated to FULL sequences within Pfam domain families . The example reported in Fig 8C and 8E illustrates how CCMs are closer to sequences than Pfam family models . Standard approaches , like Pfam , are limited to consensus within seeds only , and CCMs showed to model evolutionary information differently . Below , we propose several ways to revisit basic evolutionary questions based on CCMs . CCMs can be used to analyze ancient duplications . Modern sequences evolved in the context of full genome and local duplications . The fate of the duplicated domains could be different for several reasons . The most important one is domain organization within architectures , where the accomplishment of different cooperative functional purposes might induce duplicated domains to reach possibly very divergent sequence profiles . CCM can help to model these parallel profiles ( created by domain divergence ) better than a consensus approach . The example on clathrin domains illustrated in Fig 8 highlights the parallel co-existence in the Metazoa and Amoebozoa clades of CCMs characterized by two distinguished profiles , suggesting that an ancient duplication took place before the Metazoa-Amoebozoa phylogenetic divergence and gave origin to the two kinds of domains observed today in the two clades . The existence of two distinguished copies ( fitting the two profiles ) of the clathrin domain in the P . falciparum genome , reveals that this duplication is even older and places it before the Metazoa-Amoebozoa-Alveolata divergence . CCMs can be used for large scale explorations of parallel profiles evolution . Multiple CCMs can be used to analyse domain evolution . Whether the actual number of evolutionary pathways for a domain family is relatively small or not remains an open question . The large number of CCMs associated to a domain D characterizes the evolutionary landscape of D , and highlights the viability of different evolutionary processes . The mathematical description of the models , as probabilistic profiles , can be used to explicitly address quantitative questions on the landscape variability . For instance , a measure of how different CCMs are , one from the other , can be developed and used to bring an estimation on the number of distinct evolutionary pathways associated to a domain . Do different domains have a sensibly different number of evolutionary pathways associated to them ? Does the distribution of distances between probabilistic profiles of a given domain have a small/large variance ? Can we suggest a graph-like relational structure among CCMs and exploit the structure of the graph to infer functional consequences ? These questions do not have an answer today but they seem fundamental to the understanding of the evolutionary landscapes we observe . CCMs and their impact on functional annotation . The interest of an accurate genomic mapping of protein domains and protein architectures is multiple . It directly implies the possibility to develop: 1 . a more precise functional analysis of domains and architectures within genomes; 2 . a comparative analysis of domains and architectures between species within clades . Based on it , pangenomic differences of phylogenetically close microbial species ( strains or genomes ) can be defined at the domain level , and species variability carefully assessed . By using domains as the building blocks of proteins functional activity , we can assert that the presence of the same domains within different architectures in different species/strains might guarantee a similar functional activity for the organism . In this sense , a relaxed notion of pangenomic variability can be defined , closer in spirit to the functional activity of the species . Similar observations apply to microbial communities , where domains , more than proteins , appear to be useful building blocks for functional annotation; 3 . a comparative analysis of domains and architectures between species across distant clades . This could help to improve estimating the age of domains and architectures [47 , 48]; 4 . an accurate identification of gene homology between pairs of genomes . This will directly benefit synteny blocks reconstruction and chromosomal rearrangement analyses; 5 . an improved tracing of gene acquisition for bacterial species , where lateral gene transfer is much present . This will imply a more precise reconstruction of reticulate evolution . CCMs and domain architectures identification . The way protein architectures form is an important factor to understand protein evolution . A quantification of the elementary events affecting protein architectures , such as domain ( s ) insertion/deletion , duplication and exchange , was done [49] but , yet , little is known about the relationships between these elementary events [50] and the molecular mechanisms they originate from . Finer domain mapping ( obtained with more precise annotation tools as CLADE ) on all proteins of completely sequenced genomes will contribute precise information on the evolution of protein architectures . This means , for instance , a more precise estimation of the rate of insertion , deletion , duplication and exchange of domains within proteins in a given species . In general , it would be interesting: 1 . to establish whether the process of generation of an architecture follows constraints or not , 2 . to pinpoint such constraints , if they exist , and 3 . to verify whether they are species specific or not . This information turns out to be useful in the context of phylogenetic profiles prediction [51] . The role of the Alveolata and other clades in P . falciparum domain annotation . An a posteriori analysis of our predictions highlights that: 1 . species in the Alveolata clade are preferably chosen for domain identification ( 54% ) , 2 . a significant number of identifications ( 46% ) are suggested by species that lie outside the Alveolata clade ( that is , far from P . falciparum ) and yet providing acceptable E-values for predicted domains ( Fig 5C ) . The first point confirms the strength and importance of considering phylogenetic signals in annotation , and the second point highlights the limitations of the idea of phylogenetic proximity . Besides the observation that many P . falciparum sequences are more easily identifiable by CCMs generated by phylogenetically distant species rather than CCMs coming from Alveolata species , other observations strongly reinforce the interest in looking at different clades while identifying domains: 1 . multi-domain proteins identified with CCMs originated in different clades , suggest processes of domain evolution that are independent within each protein ( see the three domains identified with models constructed from Alveolata , Amaebozoa and Cryptophyta sequences in Fig 8 ) ; 2 . new domain families are periodically added to databases like Pfam and annotation becomes gradually more precise as a function of this addition ( see the impact of the Pfam27 enrichment on the Chlamydomonas reinhardtii protein architecture compared to Pfam24 in Fig 7A and 7C ) ; 3 . CCMs coming from bacterial and archaeal species should allow to check for ancient lateral gene transfer . On CLADE methodological improvements . On the methodological side , the use of agreement among models ( handled by SVMs ) and of co-occurrence ( handled by DAMA ) in CLADE improves predictions up to 19 . 6% on the P . falciparum sequences ( S4 Table ) , over a score system based on best E-values ( CLADEBEv ) and this highlights that criteria other than sequence similarity , play a key role in the identification of protein domains . The usage of multiple criteria to reach agreement among models , allows for new predictions and plays an important role in the evaluation of the confidence in a prediction , with improved scores that can help the biologist to annotate sequences . Several ways could be envisaged to improve further our methodological approach . These methodological improvements are general , independent from specific genome characteristics , and guarantee the strategy to be applied to any genome . They will likely be able to address at least a part of the 28% of P . falciparum proteins that are still missing a domain identification and the protein architectures that should be enriched with new domains ( among the 2394 single domain proteins , 1214 of them contain a domain that covers less than the half of the protein length , and for these proteins we expect multiple domains to lie together ) . Some unannotated proteins in P . falciparum likely contain completely novel domains , which are evolutionarily unrelated to domains present in the existing domain family databases . If that is so , CLADE would not find them , at least not until existing databases grow enough to cover more of that domain space . CCMs and computational power . By using HPC , CLADE demonstrates to push the limits in annotation reached by current methods . The first step of CLADE , generating CCMs , is the highly expensive one . It is performed once for all genomes to be annotated . All CCMs used to realize P . falciparum annotation have been constructed in 3 . 7 months of computer time by using 250 CPUs ( and they are made publicly available ) . The two subsequent steps ( 2 and 3 ) , dedicated to genome annotation , are relatively fast . For example , CLADE steps 2 and 3 ran in about 1 hr on 100 CPUs for the entire P . falciparum genome . CLADE domain library is expected to be regularly updated on new domains appearing in databases . This means that only CCMs for new domains need to be constructed and added to the existing library . This step can be realized independently from steps 2 and 3 . In years to come , the expected improvements in HPC and in CLADE implementation ( with a thoughtful selection of domain models ) will render CLADE more computationally accessible .
Our method extends Pfam , an important collection of protein domains , that has been widely used for annotating proteins with unknown function . We use Pfam v27 ( Pfam27 , downloaded from http://pfam . sanger . ac . uk ) , containing 14 831 protein domains . In order to assess the performance of our method , we apply it to the set of all P . falciparum proteins . For this , we use PlasmoDB ( http://PlasmoDB . org ) , that is the official repository of the P . falciparum proteins used as a reference database by malaria researchers . PlasmoDB v11 . 1 contains 5 542 proteins . We used the UniProtKB database [56]: 1 . to extract NCBI taxonomy for sequences and the list of known domain architectures . 2 . to recover the Pfam domain organization of all proteins in UniProt 15 . 6 ( Swiss-Prot 57 . 6 and SP-TrEMBL 40 . 6 ) ; we downloaded the dataset ftp://ftp . ebi . ac . uk/pub/databases/Pfam/releases/Pfam27 . 0/swisspfam . gz and used it to analyze co-occurrence in CLADE annotations . ( In our tables and figures , “Cooc” , abbreviating “predictions supported by domain co-occurrence” , and “Cooc-CCM” , abbreviating “predictions identified by a CCM and supported by domain co-occurrence” , count architectures whose domain pairs belong to already known architectures . ) A reference list of clades has been extracted from NCBI ( http://www . ncbi . nlm . nih . gov/taxonomy ) and used for selecting a representative set of sequences in the construction of CCMs . Clades have been specified for Bacteria , Archaea , Viruses and Eukaryotes in S1 Table . We used the SCOP v1 . 75 [41] database to compare CLADE , based on the multi-source strategy , with the mono-source strategy of HMMScan [2 , 33] and HHblits [17] . Also , the SCOP datasets allowed us to compare CLADE with the computational strategy employed in SUPERFAMILY , a system that builds multiple hidden Markov models , for each protein superfamily , to realize sequence search . The SUPERFAMILY sequence search method is built on 1 962 superfamilies ( from classes a to g ) , while CLADE relies on Pfam27 containing 14 831 protein domains . To realize the comparative analyses , we considered SCOP domains whose associated sequences are coming from at least 10 species , and constructed a three testing sets from the ASTRAL95 dataset , containing 255 domain families and 8 633 sequences with at most 95% sequence identity , from ASTRAL30 , made of 66 domain families and 1 251 sequences with at most 30% sequence identity , and from ASTRAL10 made of 18 domain families and 306 sequences with at most 10% sequence identity . ASTRAL95 , ASTRAL30 and ASTRAL10 are subsets of SCOP [21] and can be downloaded from http://scop . berkeley . edu/astral/subsets/ver=1 . 75 . Tests with HHblits required the use of HHdb , a database built from the UniProt and NCBI NR databases , and provided in the HH-suite package [17 , 42] . HMMScan was run with default parameters and curated inclusion thresholds . The option –cut_ga , for model-specific thresholding ( using profile’s GA gathering cutoffs to set all thresholding ) , was used . HMMScan is included in the HMMer 3 . 0 package [33] downloadable at http://hmmer . janelia . org/software . HHblits was run with default parameters . It is part of the HH-suite package—version 2 . 0 . 15 [17 , 42] downloadable at https://github . com/soedinglab/hh-suite . First , we detail how Pfam methodology works and how we modify it by including additional models , called clade-centered models ( CCM ) . Then , we describe how to combine those models to produce reliable predictions . In the final step , we apply DAMA , an algorithm especially designed to take into account domain co-occurrence , to find the most likely domain architecture . To realize the comparison experiment on the three SCOP datasets , we used a leave-one-out strategy as follows . Given a domain family FD in one of the ASTRAL datasets , we considered the set of n sequences , coming from different species and associated to FD in ASTRAL , to create n test-sets for FD . Each test-set takes n − 1 sequences for training and leaves one sequence out for the test . Then , we tested whether the sequence that was left out could be annotated by a model ( or models ) constructed without using it , and counted the correct identification of the domain as a true positive ( TP ) , the identification of an erroneous domain as a false positive ( FP ) and the identification of no domain as a false negative ( FN ) . Note that a domain is “correctly” identified when it belongs to the domain family FD . This same procedure , was implemented for all systems we wanted to compare: HMMScan , HHblits , CLADE , CLADEBEv , CLADE_HHblits , CLADEBEvHHblits . For HMMScan ( run with default parameters ) , each test-set of n − 1 training sequences was first aligned with Clustal W ( version 2 . 1 ) [65 , 66] . Then we used hmmbuild to construct a probabilistic model ( a profile ) from the alignment . To compare profiles and test sequences we used hmmsearch that employs a composition bias filter by default for eliminating false positives . All tools can be found in the HMMer package—version 3 . For HHblits ( run with default parameters ) , we also constructed a profile from each test-set of n − 1 aligned sequences ( it is the same alignment as for the HMMScan experiment ) but using hhmake . In HHblits experiment , test sequences must be represented by profiles and we used hhblits on HHdb ( the HHblits database ) to construct them . Profiles were compared with hhsearch [42] . All tools can be found in the HH-suite package—version 2 . 0 . 15 . For CLADE , we considered the profile constructed from the test-set of n − 1 aligned sequences in the HMMScan experiment , and we constructed additional profiles , one for each sequence in the n − 1 set . This was done by using PSI-BLAST and the NR database ( see details in Methods ) . To combine the outputs of the n profiles , we trained a SVM by using as positive set the same n − 1 sequences and as negative set other sequences never used in training nor test . The negative sequences were chosen randomly into the same SCOP dataset of the positive ones , and by avoiding sequences sharing the same SCOP fold . The SVM was trained with an equal number of positive and negative sequences . The third step of CLADE , handling architectures with DAMA , was not used . For CLADEBEv , the version of CLADE that does not include the SVM filter and that considers a score system based on best E-values only , the procedure is the same as for CLADE . The domain-specific cut-offs were learned based on E-values of positive and negative training sets . For CLADE_HHblits , we carried out the same profile construction as for CLADE , but we replaced PSI-BLAST and HMMScan profiles by HHblits models . Like before , we trained a SVM for combining profile outputs . The only difference is that we used profile-profile comparison to generate the meta-features for training the SVM instead of using sequence-profile comparison like in CLADE . The domain specific cut-offs used in CLADE were computed for the HHblits models using the same learning procedure as that employed for CLADE . For CLADEBEvHHblits , the procedure is the same as for CLADE_HHblits , with cut-offs computed for HHblits models that are domain specific but based on E-values ( as done for CLADEBEv ) . Estimating the number of false predictions is an essential step for evaluating the performance of domain identification methods . The basic principle is to estimate the probability that a potential domain has been randomly predicted . We computed the False Discovery Rate ( FDR ) in two different ways , based on different random models of sequence generation . The key idea is that domain predictions on random sequences arise by chance alone , that predictions on real sequences give us the total number of predictions ( true or false ) , and that their ratio approximate the false discovery rate . For evaluating CLADE ( but also HMMScan and DAMA ) , we run it on real sequences concatenated to its reshuffled ones . The first random model takes a protein and generates 20 different reshuffling of the protein sequence producing new sequences that have the same residue content of the original one and the same length . We call this model “1-mer” . The second random model takes a protein and generates 20 different reshuffling of k-mers in the original sequence , for k = 4 . We call this model “4-mer” . The idea behind this last model is that small k-mers within a protein sequence might be more likely to occur than random k-mers , since protein sequences might contain repetitive patterns ( for instance , blocks of hydrophobic amino-acids or other compositional biases ) . Given the original set of protein sequences P and its associated shuffled sequences S , let P + S be the set of concatenated sequences . Note that the set P + S is a set containing 20 times more sequences than P because from each sequence in P , we generated 20 sequences . Then , we computed the number of CLADE domain predictions within the P-portion ( saying R ) and the number of predictions within the S-portion ( saying A ) of the P + S sequences , and set the FDR = A/R for the dataset . This calculation is repeated 20 times , with respect to 20 different reshuffling and the FDR for the 1-mer experiment is considered to be the average of the FDRs of the 20 datasets . The same for the 4-mer experiment . The same strategy was used in [26 , 27] ( see section “FDR curves” in Methods ) . The random reshuffling was realized with the perl function List::Util::shuffle ( ) . The performances of CLADE , CLADEBEv ( a version without the SVM filter , introduced in the second step of CLADE , and considering a score system based on best E-values only ) , HMMScan [33] and HHblits [17 , 42] have been evaluated by using two standard measures: positive predictive value PPV = TP/ ( TP + FP ) ( also called precision ) and sensitivity Sen = TP/ ( TP + FN ) ( also known as recall ) , where TP , FP and FN are true positives , false positives and false negatives , respectively . To give the overall performance for each method , we computed the F-score ( also called F-measure ) combining PPV and Sen , and defined as F-score = 2 * PPV * Sen/ ( PPV + Sen ) . The F-score can be interpreted as the harmonic mean of PPV and Sen , reaching its best value at 1 and worst score at 0 . TP , FP and FN are defined as follows: let s be a protein sequence , A be its domain architecture and T be the evaluated method; a true positive is a domain in A that is correctly predicted by T , a false positive is a domain detected by T that overlaps a different domain in A , and a false negative is a domain in A that is not detected by T . The method T can detect other domains along s that do not overlap domains in A , and we shall refer to them as “additional” domains . CLADE is a pipeline that involves several different tools , and its formal time complexity is hard to establish . The model construction step takes relatively long time due to the time of generation of more than 2 millions ( 2 389 235 ) models defining the CLADE library . Namely , each domain construction takes about 30 minutes , and the overall model construction step can be realised in about 3 , 7 months on 250 CPUs . Once the models are constructed , domain identification is fast ( that is , less than an hour on 100 CPUs for about 5000 proteins ) . CLADE software and the entire library of CCMs used for the applications presented in this article are available at the address: http://www . lcqb . upmc . fr/CLADE . The CCMs that we generated can be used for annotating any genome and they avoid running the first step of CLADE again . For P . falciparum , CLADE website provides access to a downloadable xls file containing the full list of annotations for the 5542 P . falciparum proteins ( AllDomains . xls ) . The file also contains the annotations obtained with HHblits and HMMScan , and for each hit , it reports its position , the PlasmoDb accession number , the Pfam domain name , the Pfam clan ( if any ) , and the E-value . The list of disagreeing hits between CLADE and HHblits/HMMScan is also given . A HHblits/HMMScan hit disagrees with CLADE for two reasons: 1 . either the hit does not overlap CLADE hits , 2 . or the hit overlaps ( with an overlapping of any size ) a CLADE hit of a different domain and a different clan .
|
Current sequence databases contain hundreds of billions of nucleotides coding for genes and a classification of these sequences is a primary problem in genomics . A reasonable way to organize these sequences is through their predicted domains , but the identification of domains in very divergent sequences , spanning the entire phylogenetic tree of species , is a difficult problem . By generating multiple probabilistic models for a domain , describing the spread of evolutionary patterns in different phylogenetic clades , we can effectively explore domains that are likely to be coded in gene sequences . Through a machine learning approach and optimization techniques , coding for expected evolutionary constraints , we filter the many possibilities of domain identification found for a gene and propose the most likely domain architecture associated to it . The application of this novel approach to the full genome of Plasmodium falciparum , to a dataset of sequences from three SCOP datasets highlights the interest of exploring multiple pathways of domain evolution in the aim of extracting biological information from genomic sequences . Our new computational approach was developed with the hope of providing a novel tier of accurate and precise tools that complement existing tools such as HMMer , HHblits and PSI-BLAST , by exploring in a novel way the large amount of sequence data available . The existence of powerful databases for sequences , domains and architectures help make this hope a reality .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"sequencing",
"techniques",
"taxonomy",
"parasite",
"groups",
"plasmodium",
"parasitology",
"apicomplexa",
"phylogenetics",
"data",
"management",
"genome",
"analysis",
"sequence",
"motif",
"analysis",
"molecular",
"biology",
"techniques",
"research",
"and",
"analysis",
"methods",
"sequence",
"analysis",
"computer",
"and",
"information",
"sciences",
"genomics",
"sequence",
"alignment",
"proteins",
"biological",
"databases",
"molecular",
"evolution",
"evolutionary",
"systematics",
"molecular",
"biology",
"biochemistry",
"sequence",
"databases",
"database",
"and",
"informatics",
"methods",
"protein",
"domains",
"genetics",
"biology",
"and",
"life",
"sciences",
"computational",
"biology",
"evolutionary",
"biology",
"genomic",
"databases"
] |
2016
|
Improvement in Protein Domain Identification Is Reached by Breaking Consensus, with the Agreement of Many Profiles and Domain Co-occurrence
|
It has been proposed that the sensorimotor system uses a loss ( cost ) function to evaluate potential movements in the presence of random noise . Here we test this idea in the context of both error-based and reinforcement-based learning . In a reaching task , we laterally shifted a cursor relative to true hand position using a skewed probability distribution . This skewed probability distribution had its mean and mode separated , allowing us to dissociate the optimal predictions of an error-based loss function ( corresponding to the mean of the lateral shifts ) and a reinforcement-based loss function ( corresponding to the mode ) . We then examined how the sensorimotor system uses error feedback and reinforcement feedback , in isolation and combination , when deciding where to aim the hand during a reach . We found that participants compensated differently to the same skewed lateral shift distribution depending on the form of feedback they received . When provided with error feedback , participants compensated based on the mean of the skewed noise . When provided with reinforcement feedback , participants compensated based on the mode . Participants receiving both error and reinforcement feedback continued to compensate based on the mean while repeatedly missing the target , despite receiving auditory , visual and monetary reinforcement feedback that rewarded hitting the target . Our work shows that reinforcement-based and error-based learning are separable and can occur independently . Further , when error and reinforcement feedback are in conflict , the sensorimotor system heavily weights error feedback over reinforcement feedback .
Error feedback and reinforcement feedback can each guide motor adaptation in a visuomotor rotation task [1] . It has been proposed that error feedback and reinforcement feedback engage different neural mechanisms [2] . Indeed , depending on the form of feedback used during adaptation , newly acquired motor commands differ in terms of generalizability [1] and retention [3] . For error-based learning , it is suggested that adaptation occurs by minimizing error to update an internal model [4] . For reinforcement-based learning , it is proposed that adaptation is model-free and occurs by sampling motor outputs to find a set that maximizes the probability of task success [5] . Importantly , both these forms of learning can occur in the presence of internally [6 , 7] and externally [8 , 9] generated random variability ( noise ) . Loss functions are central to several current computational theories of sensorimotor control [10] . An error-based loss function describes the relationship between potential movements and the associated costs of error [9] . In other words , an error-based loss function describes how errors of different magnitudes are penalized . A reinforcement-based loss function describes the relationship between potential movements and the associated probabilities of reward ( or punishment ) [7] . The idea that the sensorimotor system uses a loss function to select a statistically optimal movement in the presence of noise has been examined in tasks involving error feedback [9 , 11] , reinforcement feedback [1] , and both error and reinforcement feedback together [1 , 3 , 7 , 12] . In tasks involving error feedback , it has been proposed that the sensorimotor system may use an error-based loss function to select movements [9] . As an example , here we will describe how an absolute error loss function and a squared error loss function can be used to select aim location during a game of darts . In this context , error refers to the distance of an individual dart to the bullseye . Let us assume that , after several throws , when attempting to aim the darts at a particular location that there is some spread , or distribution , of darts on a board . If the chosen strategy is to minimize absolute error around the bulls-eye , one should adjust their aim until the sum of the absolute distances of the darts is at its lowest possible value . This strategy corresponds to selecting a single aim location that minimizes the cost ( output ) of an absolute error loss function ( i . e . , ∑|errori|1 ) . This absolute error loss function linearly weights individual error magnitudes and would result in the median of the dart distribution being directly over the bullseye . Similarly , if the chosen strategy is to minimize squared error around the bullseye , one should adjust their aim such that the sum of the squared distances of the darts is at its lowest possible value . This strategy corresponds to selecting a single aim location that minimizes the cost of a squared error loss function ( i . e . , ∑|errori|2 ) . The squared error loss function places a heavier emphasis on minimizing large errors relative to smaller errors , in a quadratic fashion , and would result in the mean of the dart distribution being directly over the bullseye . Using tasks that involve noisy error feedback , some researchers have reported that sensorimotor behavior is best represented with an error-based loss function where the exponent on the error term is between 1 ( absolute error ) and 2 ( squared error ) [9 , 13] , while others report that an exponent of 2 best fits behavior [11 , 14 , 15] . Based on these works , it is possible that the error-based loss function that most aligns with behavior may to some extent be task dependent . The concept of loss functions also extends to sensorimotor tasks involving reinforcement feedback , where the goal is maximize task success [5] . The optimal movement that maximizes the probability of success can be found by minimizing the 0-1 loss function [9 , 16] . Again , we can describe this reinforcement-based loss function using a distribution of darts on a board . With the 0-1 loss function , every dart that hits the bulls-eye is assigned a value of 0 and each dart that misses the bulls-eye is assigned a value of 1 . To minimize this loss function , one should adjust their aim such that the greatest number of darts hit the bulls-eye . This strategy maximizes the probability of success ( by minimizing failure ) and corresponds to placing the mode of the distribution of darts directly over the bullseye . This loss function can easily be extended to account for graded reinforcement feedback , where either the magnitude [7 , 17] or probability [18] of reinforcement feedback varies according to some function that is externally imposed by the experimenter . It has recently been shown that the sensorimotor system can maximize task success when using only binary reinforcement feedback . Shadmehr and colleagues used a visuomotor rotation task , where the true target was displaced from the displayed target by some small amount [1 , 18] . In line with a reinforcement-based loss function and without any error feedback , they found that participants were able to adapt where they aimed their hand using only reinforcement feedback that signaled whether or not the true target had been hit . Researchers have also explored how reinforcement feedback and error feedback are used when provided simultaneously [1 , 3 , 7 , 12 , 17 , 19] . This has been done for a range of tasks and has yielded mixed results . In studies reported by Trommershäuser and colleagues [7; 20–23] , participants performed rapid reaches with continuous visual feedback of their hand ( visual error feedback ) to a large rewarding target ( positive reinforcement ) with an overlapping punishment region ( negative reinforcement ) . Participants learned to aim to a location that maximized reward . Conversely , others have provided evidence that continuous error feedback dominates over , or perhaps suppresses , reinforcement feedback during a visuomotor rotation task [3 , 12 , 24] . Izawa and Shadmehr ( 2011 ) suggested that with a decrease in error feedback quality , the sensorimotor system might increase its reliance on reinforcement feedback . However , a feature of these experiments is that they did not separate the predictions for where participants should aim their hand when receiving reinforcement feedback or error feedback . Such separation would provide a powerful way to investigate how the sensorimotor system weights the relative influence of error feedback and reinforcement feedback when they are provided in combination . Here , we designed two experimental reaching tasks that separate the predictions of error-based and reinforcement-based loss functions on where to aim the hand . In doing so , we promoted dissociation in behavior simply by manipulating the form of feedback provided to participants . Participants reached to visual targets without vision of their arm . Unbeknownst to participants , visual feedback of their hand ( represented by a cursor ) was laterally shifted from trial to trial by an amount drawn from a skewed probability distribution . Skewed lateral shift probability distributions allowed us to separate the predictions of error-based and reinforcement-based loss functions on where to aim the hand . For example , a squared error loss function would predict that we should aim our hand to a location that corresponds to the mean of the skewed lateral shift probability distribution . Conversely , a reinforcement loss function would predict that we should aim our hand to a location that corresponds to the mode . Critically , skewed distributions separate several statistics , such as the mean and mode , that align with the optimal predictions of error-based and reinforcement based loss functions . Thus , by laterally shifting feedback using skewed noise and observing where participants reached , we were able to directly test how the sensorimotor system weights the relative influence of reinforcement feedback and error feedback when deciding where to aim the hand . In Experiment 1 we tested how reinforcement feedback and error feedback influence where participants aimed their hand . We predicted that participants receiving only error feedback would minimize some form of error . We also predicted that participants receiving both error and reinforcement feedback would increasing rely on reinforcement feedback with a decrease in error feedback quality . Such a strategy predicts a different pattern of compensation for participants who received both forms of feedback when compared to those who only received error feedback . Surprisingly , we found that both error-only feedback and error plus reinforcement feedback resulted in participants minimizing approximately squared error . In Experiment 2 we used a modified task to verify that reinforcement feedback alone was capable of influencing where to aim the hand . Indeed , we found that participants who received only reinforcement feedback maximized the probability of hitting the target . However , we again found that participants minimized approximately squared error when both error and reinforcement feedback were present . Taken together , our results describe how the sensorimotor system uses error feedback and reinforcement feedback , in isolation and combination , when deciding where to aim the hand .
Participants performed 2000 reaching movements in a horizontal plane ( Fig 1 ) . They were instructed to “hit the target” . A cursor that represented the true hand position disappeared once the hand left the home position . On each trial , the unseen cursor was then laterally shifted by an amount drawn from a skewed probability distribution . Participants were randomly assigned to one of three groups ( n = 10 per group ) . The ErrorSR group and ErrorSL group both received error feedback that was laterally shifted by a right-skewed ( RS; Fig 2A ) or left-skewed ( SL ) probability distribution , respectively . The third group , Reinforcement + ErrorSR , received error feedback and reinforcement feedback that were both laterally shifted by a SR probability distribution . Importantly , the skewed lateral shift probability distributions were designed to separate the mean and the mode , corresponding with the optimal solutions of error-based and reinforcement-based loss functions , respectively . This separation allowed us to investigate how the sensorimotor system weights the relative influence of reinforcement feedback and error feedback when deciding where to aim the hand . We hypothesized that with a decrease in error feedback quality the sensorimotor system would increase its reliance on reinforcement feedback . To manipulate error feedback quality , halfway through each reach the laterally shifted cursor was either not presented ( ς∞ ) or briefly presented ( for 100ms ) as a single dot ( ς0mm ) , a medium cloud of dots ( ς15mm ) or a large cloud of dots ( ς30mm ) , before disappearing once again ( Figs 1 and 2 ) . The medium and large clouds were composed of 25 dots , such that the dots were distributed according to a bivariate normal distribution with a standard deviation of 15mm and 30mm , respectively . Participants then attempted to hit the target by accounting for the laterally shifted error feedback ( ς0mm , ς15mm , ς30mm , ς∞ ) they had experienced mid-reach . All participants received additional error feedback ( also a single dot ) at the end of the reach on trials in which single dot ( ς0mm ) error feedback was presented mid-reach [8] . Participants in the Reinforcement + ErrorSR group were presented with reinforcement feedback only on trials in which the error feedback was presented as a single dot ( ς0mm ) . The reinforcement feedback was binary ( the target doubled in size , a pleasant sound was played over a loudspeaker , and participants received 2¢ CAD ) and was presented when the laterally shifted cursor hit the target . On each trial , we estimated how a participant compensated for the lateral shift by recording their hand location relative to the displayed target ( see Fig 1 ) . This was done for the four different levels of error feedback quality and lateral shift magnitudes . The average compensatory behavior of each group is shown in Fig 3 . It can be seen that , with very little visual uncertainty about the magnitude of a lateral shift ( ς0mm ) , all groups had a pattern of compensation that was well matched to the true magnitude of the shift . As error feedback quality decreased from little uncertainty ( ς0mm ) to some uncertainty ( ς15mm and ς30mm ) to complete uncertainty ( ς∞ ) , participants’ pattern of compensation became increasingly less sensitive to the true magnitude of the lateral shift . This is consistent with Bayesian inference in that participants were increasing their reliance on their prior with a decrease in error feedback quality . Interestingly , the average compensation of each group , even for participants who received both reinforcement and error feedback , corresponded quite closely to the predictions made by a Bayesian model whose loss function minimized squared error ( compare Figs 3A–3C to 2E ) . To quantify the extent to which reinforcement feedback had an influence on behavior , we performed three separate analyses . First , we compared the average compensatory behavior between groups . Second , we used a Bayesian model to characterize how error feedback is used to guide behavior , and the extent to which reinforcement feedback influenced compensation . Finally , we used a simple linear model to characterize how the relationship between compensation and lateral shift is modulated by error signal quality , and the extent to which reinforcement feedback influenced compensation . All three analyses supported the idea that reinforcement feedback did not influence behavior when it was presented in combination with error feedback . Below , we describe each group’s compensatory behavior and the results of the Bayesian model . For brevity , detailed results of the linear model are presented in S2 Data . In Experiment 1 we found that when both error feedback and reinforcement feedback were presented in combination , participant behavior seemed only driven by error feedback . There is a possibility that the reinforcement feedback we used was not capable of influencing behavior . To test this , in Experiment 2 some participants only received reinforcement feedback , without error feedback . It is important to note that these participants only received reinforcement feedback at the end of each movement . This represents a difference in experimental design from Experiment 1 , whose participants ( ErrorSR , ErrorSL , Reinforcement + ErrorSR ) often received feedback twice in a single movement—mid-reach and as they passed by the target . To properly control for this , in Experiment 2 we tested two additional groups that received only error feedback , or error plus reinforcement feedback . Importantly , however , all participants in Experiment 2 only received feedback once per trial , at the end of movement . This ensured that the frequency and location of feedback received by the three groups was the same . As a consequence of not providing error feedback mid-reach and only providing feedback at the target , compensation to the skewed lateral shift probability distribution in Experiment 2 reflects a trial-by-trial updating of where to aim the hand . This differs from Experiment 1 in which compensation reflects both online ( via mid-reach feedback ) and trial-by-trial ( via target feedback ) updating of where to aim the hand . In the context of a Bayesian framework , this would indicate in Experiment 2 that the prior representation of lateral shifts is updated after a trial is completed , instead of both during and after a trial is complete as in Experiment 1 . One group of participants only received reinforcement feedback ( Reinforcement ) , a second group received only error feedback ( Error ) , and a third group received both error and reinforcement feedback ( Reinforcement + Error ) . In total , there were ninety participants ( 30 per group ) . All participants performed 500 reaching movements in a horizontal plane . They were instructed to “hit the target” . On every trial , a cursor that represented the true hand position disappeared once the hand left the home position . The unseen cursor was laterally shifted by an amount drawn from a skewed-right ( SR; n = 15 per group; Fig 4A ) or skewed-left ( SL , n = 15 per group ) probability distribution . Binary reinforcement feedback occurred when the laterally shifted cursor hit the target ( the target doubled in size , a pleasant sound was played over a loudspeaker , and participants received 2 ¢ CAD ) . Error feedback was presented as a single dot at the end of the reach , at the location where the laterally shifted cursor passed by or through the target . For each reach we recorded each participant’s pattern of compensation , that is , how laterally displaced his or her hand was relative to the displayed target ( Fig 1 ) . We calculated the compensation location that would maximize the probability of hitting the target ( x a i m m a x ( h i t s ) ; Eqs 7 , 8 and 12 and Fig 4C ) . This calculation incorporated a measure of movement variability at the target , which was larger in this experiment , relative to Experiment 1 , given that there was no mid-reach error feedback ( see Methods for further details ) . We also calculated the compensation location that would minimize squared error of cursor positions about the target ( x a i m m i n ( e r r o r 2 ) ; Eqs 7–11 and Fig 4B ) . Fig 5 shows the pattern of compensation of a participant from each group . In response to the skewed lateral shift probability distribution , it can be seen that the Reinforcement participant learned to compensate by an amount that was on average close to maximizing the probability of hitting the target . Conversely , both the Error and the Reinforcement + Error participants had an average compensation that corresponded to minimizing approximately squared error . Fig 6A shows the average group compensatory behavior in response to the skewed lateral shift probability distribution . Compensation reached an asymptotic level after approximately 100 reaches ( bin 10 ) . Thus , for each participant , we averaged their last 400 trials to obtain a stable estimate of their behavior ( Fig 6B ) . However , the results reported below were robust to whether we averaged the last 100 , 200 , 300 or 400 trials ( Table 1 ) . To test whether the form of feedback and skew direction influenced behavior , we compared the average pattern of compensation between the three groups . We found that there was a significant main effect of group [F ( 2 , 84 ) = 8 . 928 , p < 0 . 001 , ω ^ G 2 = 0 . 150] . There was no significant main effect of skew direction [F ( 1 , 84 ) = 0 . 164 , p = 0 . 687 , ω ^ G 2 < 0 . 001] nor an interaction between group and skew direction [F ( 2 , 84 ) = 0 . 498 , p = 0 . 61 , ω ^ G 2 < 0 . 001] . Thus , under the influence of the same skewed lateral shift probability distribution , we found that different forms of feedback resulted in significantly different compensatory behavior . Specifically , we found a statistically reliable difference between the Reinforcement and Error groups ( p < 0 . 001 , one-tailed; θ ^ = 77 . 7 % ) . The Reinforcement group approached a compensatory position that would maximize their probability of hitting the target [x a i m m a x ( h i t s ) ] ( p = 0 . 081 , two-tailed; θ ^ = 63 . 3 %; CI[7 . 1 , 10 . 8mm] ) . Further , the Reinforcement group’s pattern of compensation was significantly different from one that corresponded to the minimization of squared error [x a i m m i n ( e r r o r 2 ) ] ( p < 0 . 001 , one-tailed; θ ^ = 83 . 3 % ) . These two findings suggest that the reinforcement feedback used in Experiments 1 and 2 is capable of influencing behavior in a way that aligns with a reinforcement-based loss function . One prediction in the context of reinforcement-based learning is that an individual’s movement variability should influence their pattern of compensation to the lateral shifts . We tested this for the Reinforcement participants in Experiment 2 . We characterized movement variability as the standard deviation of final hand position during the asymptotic phase of learning ( the last 400 trials ) . While at the group level , participants appeared to compensate by an amount that approached an optimal strategy ( see above ) , on an individual basis we did not find a statistically reliable relationship between movement variability and compensation ( R2 = 0 . 003 ) . This finding appears to differ from that reported by Trommershäuser et al . ( 2003a ) . However , as expanded on in the Discussion , there are many differences between their task and ours , and such findings are not uncommon [16 , 25] . Nevertheless , the group level data suggests the existence of a reinforcement-based loss function that maximizes the probability of hitting the target . As expected , the Error group minimized squared error ( Fig 6B ) . Their pattern of compensation was aligned with one that , on average , minimized squared error ( p = 0 . 795 , two-tailed; θ ^ = 50 . 0 % ) . It was also significantly different from one that maximized their probability of hitting the target ( p < 0 . 001 , one-tailed; θ ^ = 100 . 0 % ) . The results of Experiment 2 support the idea that the human sensorimotor system can update where to aim the hand during a reach by using only reinforcement-based feedback ( to maximize the probability of hitting the target ) or only error-based feedback ( to minimize approximately squared error ) . As in Experiment 1 , we again found that participants in the Reinforcement + Error group minimized squared error , and that reinforcement feedback did not influence behavior . Participants in the Reinforcement + Error group exhibited a pattern of compensation that was significantly different from the Reinforcement group ( p = 0 . 004 , two-tailed; θ ^ = 76 . 7 % ) , but was indistinguishable from participants in the Error group ( p = 0 . 922 , two-tailed; θ ^ = 52 . 2 % ) . Taken together , the results from both Experiment 1 and 2 support the idea that the sensorimotor system heavily weights error feedback over reinforcement feedback when updating where to aim the hand during a reaching task .
A key aspect of this study was the use of skewed noise to separate the optimal aim locations predicted by reinforcement-based and error-based loss functions . This allowed us to probe how the sensorimotor system uses reinforcement feedback and error feedback , in isolation and combination , to update where to aim the hand during a reaching task . We found that participants minimized approximately squared error when they received only error feedback . Participants maximized the probability of hitting the target when they received only reinforcement feedback . When both forms of feedback were presented in combination , participants minimized approximately squared error at the expense of maximizing the probability of hitting the target . This finding suggests that the sensorimotor system heavily weights error feedback over reinforcement feedback when deciding to aim the hand . In both Experiments , we found that the sensorimotor system minimized approximately squared error when using error feedback to guide reaching movements . This agrees well with previous work that examined how humans adapt to a small range of asymmetrical or multimodal noise during proprioceptive [11 , 14] and visual [15] tasks . In Experiment 1 and 2 , we also used small ranges of asymmetrical noise that respectively spanned 3cm and 2 . 8cm of the workspace . There is some evidence that as noise exceeds this range , the sensorimotor system is less sensitive to large errors [9 , 26] . While in the present study our data points to an error-based loss function based on squared error , other studies have focused on other mathematical forms . Another commonly examined loss function is the inverted-Gaussian [9 , 13 , 16] , which places greater emphasis on penalizing smaller errors and less emphasis on penalizing larger errors . Sensinger and colleagues ( 2015 ) used a biofeedback task that involved controlling a myoelectric signal corrupted with skewed noise similar to that used by Körding and Wolpert ( 2004b ) . They then examined several different loss functions and their corresponding best-fit parameters given the data . The parameters of a loss function define how errors of different sizes are weighted relative to one another . For the inverted-Gaussian loss function they found its best-fit parameter was much larger ( 9 times ) than that found by Körding and Wolpert ( 2004b ) . They suggest that the inverted Gaussian loss function may not be generalizable across different motor tasks . They did , however , estimate a best-fit power loss function exponent of 1 . 69 , a value that was quite close to the figure of 1 . 72 estimated by Körding and Wolpert ( 2004b ) . In the present study we estimated an average best-fit power loss function exponent closer to 2 . 0 , which was not significantly different from 1 . 72 . These differences may be due to the range of noise and possibility the shape of the skewed noise . Nevertheless , and similar to others [11 , 14 , 15] , we were able to explain 80% to 89% of the variability in our data ( see S1 Data ) using a power loss function that minimized approximately squared error ( i . e . , αopt ≈ 2 . 0 ) . In the current paper , we use a Bayesian framework to interpret and model how the sensorimotor system uses error feedback and reinforcement feedback when deciding where to aim the hand . This framework combines prior experience and current sensory information , such as sensory cues [27] and sensory uncertainty [8] , in a statistically optimal fashion . By accounting for both prior and current information , the Bayesian framework has successfully explained a broad range of phenomena , such as reduced movement variability [8 , 28] , perceptual illusions [29] and online feedback control [30] . An alternative computational framework for error-based learning has been instrumental to our understanding of how the sensorimotor system learns to adapt on a trial-by-trial basis [31–35] . Of these models , the ones that account for sensorimotor noise [32 , 33 , 35] have been termed , ‘aim point correction’ models [36] . van Beers ( 2009 ) extended upon this framework with the ‘planned aim point correction’ model . This model separates central movement planning noise and peripheral movement execution noise . This model was able to explain reach adaptation patterns in a naturalistic task while demonstrating that the sensorimotor rate of learning was optimal given the properties of planning and execution noise . It has been successfully applied to explain differences in novice and expert dart throwers [37] , and can account for both learning in task-relevant dimensions and exploratory ( random walk ) behaviour in task-irrelevant dimensions [38] . Aim point correction models , which can be derived from a Bayesian framework , are attractive because they are computationally tractable and learning is modeled using terms and constructs from sensorimotor control , such as planning noise , motor commands , efference copies , and execution noise [36] . In their current formulation , however , these models do not incorporate how the sensorimotor system responds to errors of differing magnitudes and different amounts of sensory uncertainty . We accounted for both of these factors with our Bayesian model , which was essential for testing our hypotheses . To further study how the sensorimotor system adapts on a trial-by-trial basis in experiments such as the ones used in this paper , a useful future direction would be incorporating the effects of sensory uncertainty and how errors of differing magnitudes are penalized . While it is most common to study adaptation while providing only error feedback , researchers have also examined adaptation in the context of both reinforcement and error feedback [1 , 3 , 7 , 12 , 20–23 , 39 , 40] . Using a visuomotor rotation task , Izawa and Shadmehr ( 2011 ) examined how the sensorimotor system uses both a reinforcement and error feedback when deciding where to aim the hand during a reach . They manipulated the quality of error feedback presented on each trial in the following three ways: first , by displaying the cursor both at mid-reach and also at the target ( mid-reach and target error-feedback condition ) ; second , by displaying the cursor only at the target ( target error-feedback condition ) ; and thirdly , by withholding visual feedback of the cursor completely ( no error-feedback condition ) . In each of these conditions , participants received reinforcement feedback for hitting the target ( the target expanded and a pleasant sound was played over a loudspeaker ) . They modeled participants’ aiming behavior using a modified Kalman filter , which increasingly relied on reinforcement feedback as the quality of error feedback decreased . However , in the Izawa and Shadmehr ( 2011 ) experiment , the predicted aim location for minimizing error and maximizing the probability of hitting the target overlapped , making it difficult to determine the extent to which participants’ adaptation was driven by error feedback versus reinforcement feedback . Izawa and Shadmehr ( 2011 ) found greater movement variability in trials that provided feedback only at the target when compared to trials that provided feedback both mid-reach and at the target . Their model attributed a greater proportion of adaptation due to reinforcement feedback on trials in which error feedback was provided at the target , compared to trials in which error feedback was provided both mid-reach and at the target . While this is certainly a possibility , an alternative explanation for these behavioral differences is that the participants receiving feedback both mid-reach and at the target , unlike those receiving feedback only at the target , were able to compensate for accumulated sensorimotor noise error they sensed mid-reach . That is , it is difficult to determine whether behavioral differences between conditions were a result of reinforcement feedback , differences in the amount of ( or location of ) feedback , or both . In the present study we designed experiments aimed at resolving both of these potential issues . First , we were able to separate the optimal aim location of error-based and reinforcement-based loss functions . Second , for each group in Experiment 2 we equated the amount of and location of feedback by providing it only at the target . We found no differences in compensation between participants who received error feedback and participants who received both error and reinforcement feedback . This finding aligns with the work of Vaswani et al . ( 2015 ) , who similarly found that reinforcement feedback did not influence behaviour when it was combined with error feedback . This suggests that the sensorimotor system heavily weights error feedback , when available , over competing reinforcement feedback . Trommershäuser and colleagues ( 2003a ) provided evidence that humans can use reinforcement feedback to adjust where they aim their hand during a reaching task . In their study , participants reached to a screen displaying a rewarding target area ( positive reinforcement: monetary gain ) and an overlapping penalty area ( negative reinforcement: monetary loss ) . Thus , participants received both reinforcement feedback for hitting the target , and error feedback that indicated where they touched the screen relative to the target . The reward and penalty for hitting these respective areas were held constant for a given block of trials . In their task , internal sensorimotor noise created a smooth and differentiable reinforcement landscape within and surrounding the reward and penalty areas . They found participants accounted for their internal sensorimotor noise and aimed towards the peak of this reinforcement landscape to maximize the probability of hitting the target . In our task , we found that participants receiving only reinforcement feedback maximized the probability of hitting the target , but participants who received both reinforcement and error feedback minimized squared error . Here , we suggest two potential reasons why participants receiving both reinforcement and error feedback continued to minimize error in our task , and appeared not to be driven by reinforcement feedback . First , it may be related to how the reinforcement region was spatially defined [10] . Second , the finding may only be present during implicit learning [41] . In the Trommershäuser et al . ( 2003a ) experiments , the reinforcement regions were static and clearly defined in the workplace by the bounds of the penalty and reward areas . Thus , participants may have used error feedback to guide their aim towards the location they explicitly learned would maximize the probability of hitting the target . In our task , a stationary target was displayed on the screen . However , due to the lateral shifts , this target did not always represent the true reinforcement region . Moreover , similar to other experiments that used a small range of external noise and in which participants were unaware of their average change in reaching aim direction [8] , our task was most likely implicitly learned [41] . Thus , unlike Trommershäuser et al . ( 2003a ) , our participants may have had to implicitly build a representation of the probabilistically varying reinforcement regions . In the absence of any error feedback , we found that participants receiving only reinforcement feedback compensated by an amount that approached the optimal compensation that maximized the probability of hitting the target . In accordance with a loss function that maximizes the probability of hitting the target , this suggests that the sensorimotor system was able to learn and compensate for the probabilistically provided reinforcement feedback . Despite an average compensation across participants that did not differ from the optimal solution , we did not find a correlation between final compensation and movement variability on an individual basis . This differs from Trommershäuser et al . ( 2003a ) , but as mentioned above there are many differences in experiment design between their task and ours . However , optimal performance across participants with suboptimal performance on an individual basis has been previously reported when using complex distributions ( i . e . , skewed or bimodal distributions ) to perturb feedback [16 , 25] . From a probabilistic viewpoint , suboptimal performance on an individual basis may partially be explained by individual differences in exploratory movement variability [42] , as well as inaccurate [10] , approximate [15] or stochastic [16] representations of the prior , likelihood and posterior . To produce accurate goal-directed movements , our nervous system must account for uncertainty and nonlinearities present in our environment [8 , 9 , 11] , in the biomechanics of our body [43 , 44] , and in our nervous system [6 , 45–47] . By using skewed probability distributions , we were able to separate the optimal aim locations predicted by reinforcement-based and error-based loss functions . The results of our experiments demonstrate that by changing the form of feedback provided to participants , reinforcement-based and error-based learning are dissociable and can occur independently . Further , we also found that when both a reinforcement and error feedback are available , the sensorimotor system relies heavily , perhaps exclusively , on error feedback when deciding where to aim the hand during targeted reaching . Our work consolidates and builds upon previous research to provide insight into how the sensorimotor system performs model-based and model-free learning .
30 individuals ( 20 . 2 ± 2 . 7 years ) participated in Experiment 1 and 90 individuals ( 20 . 6 ± 2 . 4 years ) participated in Experiment 2 . All participants were healthy , right-handed and provided informed consent to procedures approved by Western University’s Ethics Board . In both experiments , participants performed right-handed reaching movements in a horizontal plane while grasping the handle of a robotic arm ( InMotion2 , Interactive Motion Technologies , Massachusetts , United States; Fig 1 ) . A semi-silvered mirror occluded vision of the arm and projected images from an LCD screen onto a horizontal plane aligned with the shoulder . These images included visual targets and , in certain conditions , real-time visual feedback of either the true or laterally shifted hand position . The workspace was calibrated such that the centroid of unperturbed visual feedback matched the center of the robotic handle . An air-sled supported each participant’s right arm and provided minimal friction with the desk surface . An algorithm controlled the robot’s torque motors and compensated for the dynamical properties of the robotic arm . Hand position was recorded at 600Hz and stored for offline analysis . We performed all offline analyses with custom Python ( 2 . 7 . 9 ) scripts . SPSS ( version 21 . 0; IBM , Armond , NY ) was used for analysis of variance ( ANOVA ) tests . We used Greenhouse-Geisser adjustments to correct for violations of sphericity . Effect sizes for each ANOVA were calculated using generalized omega squared ( ω ^ G 2 ) , where values of ω ^ G 2 equal to 0 . 02 , 0 . 13 , and 0 . 26 were considered small , medium and large effects [56 , 57] . Follow-up comparisons ( one-sampled , two-sampled or paired ) were computed using nonparametric bootstrap hypothesis tests ( n = 1 , 000 , 000 ) . These tests provide more reliable p-value estimates than traditional comparison tests ( e . g . , t-tests ) . Briefly , they make no parameter assumptions ( e . g . , normality ) and are less sensitive to unequal sample sizes or unequal variances . One-tailed tests were used when we had a priori predictions based on our Bayesian model , such as when we predicted significantly different compensation between participants only receiving error feedback and those receiving only reinforcement feedback . For all other comparisons we used two-tailed tests . Holm-Bonferroni corrections were used to allow for multiple comparisons [58] . 95th percentile confidence intervals ( CI ) were calculated by bootstrapping . Effect sizes for follow-up comparisons were made using the common language effect size statistic ( θ ^ ) , where values of θ ^ equal to 56 . 0% , 64 . 0% and 71 . 0% were respectively considered small , medium and large effects [59 , 60] . Significance was set to p < 0 . 05 .
|
Whether serving a tennis ball on a gusty day or walking over an unpredictable surface , the human nervous system has a remarkable ability to account for uncertainty when performing goal-directed actions . Here we address how different types of feedback , error and reinforcement , are used to guide such behavior during sensorimotor learning . Using a task that dissociates the optimal predictions of error-based and reinforcement-based learning , we show that the human sensorimotor system uses two distinct loss functions when deciding where to aim the hand during a reach—one that minimizes error and another that maximizes success . Interestingly , when both of these forms of feedback are available our nervous system heavily weights error feedback over reinforcement feedback .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"learning",
"control",
"theory",
"medicine",
"and",
"health",
"sciences",
"engineering",
"and",
"technology",
"nervous",
"system",
"social",
"sciences",
"limbs",
"(anatomy)",
"neuroscience",
"learning",
"and",
"memory",
"control",
"engineering",
"probability",
"distribution",
"mathematics",
"cognitive",
"psychology",
"systems",
"science",
"statistical",
"distributions",
"musculoskeletal",
"system",
"computer",
"and",
"information",
"sciences",
"hands",
"probability",
"theory",
"arms",
"psychology",
"anatomy",
"biology",
"and",
"life",
"sciences",
"sensory",
"systems",
"physical",
"sciences",
"cognitive",
"science"
] |
2017
|
Dissociating error-based and reinforcement-based loss functions during sensorimotor learning
|
Leishmaniasis is a parasitic infection that afflicts approximately 12 million people worldwide . There are several limitations to the approved drug therapies for leishmaniasis , including moderate to severe toxicity , growing drug resistance , and the need for extended dosing . Moreover , miltefosine is currently the only orally available drug therapy for this infection . We addressed the pressing need for new therapies by pursuing a two-step phenotypic screen to discover novel , potent , and orally bioavailable antileishmanials . First , we conducted a high-throughput screen ( HTS ) of roughly 600 , 000 small molecules for growth inhibition against the promastigote form of the parasite life cycle using the nucleic acid binding dye SYBR Green I . This screen identified approximately 2 , 700 compounds that inhibited growth by over 65% at a single point concentration of 10 μM . We next used this 2700 compound focused library to identify compounds that were highly potent against the disease-causing intra-macrophage amastigote form and exhibited limited toxicity toward the host macrophages . This two-step screening strategy uncovered nine unique chemical scaffolds within our collection , including two previously described antileishmanials . We further profiled two of the novel compounds for in vitro absorption , distribution , metabolism , excretion , and in vivo pharmacokinetics . Both compounds proved orally bioavailable , affording plasma exposures above the half-maximal effective concentration ( EC50 ) concentration for at least 12 hours . Both compounds were efficacious when administered orally in a murine model of cutaneous leishmaniasis . One of the two compounds exerted potent activity against trypanosomes , which are kinetoplastid parasites related to Leishmania species . Therefore , this compound could help control multiple parasitic diseases . The promising pharmacokinetic profile and significant in vivo efficacy observed from our HTS hits highlight the utility of our two-step phenotypic screening strategy and strongly suggest that medicinal chemistry optimization of these newly identified scaffolds will lead to promising candidates for an orally available anti-parasitic drug .
Leishmaniasis constitutes a spectrum of diseases that range in severity from self-healing to fatal . The disease can present as self-healing but potentially disfiguring cutaneous leishmaniasis [1]; metastatic and highly disfiguring mucocutaneous leishmaniasis [2]; or fatal visceral leishmaniasis [3] , where the parasite targets internal organs such as the liver , spleen , and bone marrow . Different species and strains of Leishmania parasites cause these distinct pathologies . The severity of the disease also depends upon host factors such as immune status [4] . An estimated 12 million individuals are infected with leishmaniasis worldwide , with a widespread geographic range that spans from India to the Mediterranean countries , to North and South America [5] . All Leishmania species have a life cycle that includes motile promastigotes that reside in the gut of the sand fly vector and non-motile amastigotes that live in the phagolysosomal vesicles of mammalian host macrophages [5] . Despite the disease’s prevalence , the current antileishmanial drug therapies are inadequate [6] . Since the 1940s , standard therapies for leishmaniasis include pentavalent antimonials , such as sodium stibogluconate ( Pentostam ) and meglumine antimonate ( Glucantime ) , which are administered daily over the course of 20–30 days . Both drugs are subject to widespread resistance and are highly toxic such that treatment alone can lead to mortality [7] . The diamidine pentamidine , which has similar disadvantages , has been another drug of choice to treat cutaneous leishmaniasis for several decades . Newer drugs include amphotericin B , especially in liposomal formulation ( AmBisome ) , the aminoglycoside paromomycin , and the phospholipid miltefosine [8 , 9] , which received FDA approval in 2014 . However , none of these drugs is even close to optimal . They all have moderate to high toxicity , need to be administered over multiple weeks , and suffer from increasing drug resistance . Only miltefosine , a known teratogen that is unsuitable for pregnant patients , can be administered orally [10] . Leishmaniasis has been characterized as ‘a major health problem , and there is no satisfactory treatment so far’ [6] . Hence there is an urgent need for novel therapies that are safe , potent , orally bioavailable , have a low cost of goods , and are effective against drug-resistant strains of Leishmania parasites . Although a major bottleneck in progress had been the paucity of lead compounds [11] that offer the potential of becoming new antileishmanial drugs , the situation has improved recently with the application of phenotypic screening and the associated identification of multiple lead series [12] . Phenotypic screens measure the effects of a compound on intact cells rather than an isolated target ( i . e . , biochemical enzymatic assay ) [13 , 14] . Active compounds generated from whole cell-based phenotypic screens generally offer favorable cell permeability and solubility that can facilitate compound development . One limitation with this approach is that the mechanism of action of new compounds is typically unknown . Nonetheless , phenotypic screens have the complementary advantage that they can identify compounds that act therapeutically against pathways that were previously not known to be critical for parasite viability [15] . Prior phenotypic screens have predominantly used the promastigote form of the parasite , which can be readily cultured in vitro but is not the disease-causing form of the parasite . This approach has the advantage of being able to accommodate large numbers of compounds , such as the 200 , 000-compound library that Sharlow and colleagues screened [16] . Investigators have also used axenic amastigotes [17 , 18] , which are more relevant to the disease but are nevertheless a host cell-free system that only imperfectly approximates intracellular amastigotes . Most scientists agree that assays that use intramacrophage amastigotes are the most physiologically relevant assays even though they offer lower throughput . Researchers have started employing a two-stage approach involving an initial screen of promastigotes or axenic amastigotes and a secondary step to confirm the hits by screening them against intracellular amastigotes [19–21] . This approach allows the screen to be carried out with a facile high throughput approach followed by a second , more stringent , test of the primary hits for efficacy against the disease-causing intra-macrophage parasites . The advent of high-content microscopic approaches has enabled the direct screening of compounds against amastigotes growing inside cultured mammalian macrophages [22–26] . This method can eliminate compounds that act against promastigotes while leaving amastigotes unaffected . This method is also useful for identifying compounds that target amastigotes but not promastigotes . However , this assay is technically much more complicated to undertake than assays that use promastigotes or axenic amastigotes [19] . Although one can screen large libraries with sufficient time and effort , the screens published to date have all employed smaller libraries , such as the 26 , 500-compound library used in Siqueira-Neto et al . ’s report [26] , or the focused libraries of Medicines for Malaria Box [27] , and the microbial extracts collection [28] . Although many of the hits identified in the above screens have not yet been advanced to testing in animal models of leishmaniasis [29 , 30] , some promising leads have been identified , and various organizations are currently conducting medicinal chemistry programs . For example , the Drugs for Neglected Diseases Initiative ( DNDi ) is subjecting several chemotypes such as the nitroimidazoles and oxaboroles [31] to both in vitro and in vivo evaluation as orally deliverable antileishmanials in mice and hamsters . Furthermore , the Genome Institute of the Novartis Research Foundation ( GNF ) has identified a selective inhibitor of the kinetoplastid proteasome , GNF6702 , which is active against several parasite species [32] . In addition to these advances , there is substantial benefit to providing a continued robust pipeline of lead compounds for the development of safe , potent , and orally bioavailable antileishmanials that could considerably improve the current sub-optimal armamentarium for leishmaniasis . In this paper we report a screen of roughly 600 , 000 compounds for growth inhibition of L . mexicana promastigotes from several libraries , namely the St . Jude Children’s Research Hospital Chemical Biology & Therapeutics ( CBT ) library [33] and the Tres Cantos Antimalarial Set [34] . Two top hits from this screen , compounds 4 and 5 , exhibited promising pharmacokinetic profiles that were substantially efficacious in a L . mexicana murine model of cutaneous leishmaniasis when delivered by oral gavage at a dose of 25–30 mg/kg over 10 days . Together these results suggest that compounds 4 and 5 are promising new starting points for the development of orally bioavailable antileishmanial drugs .
Animal work was approved by the Oregon Health & Science Institutional Animal Care and Use Committee under protocol #IS00002639 under adherence to the Animal Welfare Act regulations and Public Health Service Policy for the Humane Care and Use of Laboratory Animals or by the St . Jude Children’s Research Hospital Institutional Animal Care and Use Committee under protocol #477 in compliance with the Animal Welfare Act and rules articulated by the Public Health Service Policy for the Humane Care and Use of Laboratory Animals . The current CBT library consists of roughly 600 , 000 unique molecules purchased from a variety of commercial sources . The library breaks into four major sets: approved drugs ( ~1 , 100 compounds ) ; other known bioactives ( ~2 , 500 compounds ) ; focused sets directed at defined targets , including G protein coupled receptors , kinases , proteases , and phosphatases ( ~45 , 000 compounds ) ; and the diversity collection , which is the largest component of this library . All samples in the CBT library were carefully chosen to provide a balanced , functionally diverse collection suitable for discovery of chemical matter active against a wide variety of targets and for phenotypic screening [35 , 36] . In particular , the diversity subset has been designed using a maximally diverse cluster philosophy so that the population is made up of multiple clusters , each containing a series of related compounds , where the clusters are diverse with respect to one another . First , commercially available compounds were filtered using a combination of physiochemical metrics to improve bioavailability , and functional group metrics to reduce the probability of non-specific or artifact effects . The former is guided by the correlation of physiochemical parameters with bioactivity , as opposed to oral availability [36] . The latter is guided by implementation of the Vertex ‘Rapid Elimination of Swill’ model [37–39] , which utilizes a numeric scoring method with each functional group being assigned a score from –5 ( always excluded ) to 0 ( never excluded ) and allowing an aggregate score of –2 before elimination . Next , the filtered compound list was used to generate maximally diverse clusters . In order to do this , the compounds were reduced to core fragments ( or ‘scaffolds ) using the method of Bemis and Murcko [40] , and the compound clusters were then prioritized for purchase based on the balance of cluster diversity from the existing library as assessed by Tanimoto similarity and the presence of a reasonable number of analogs within a cluster . From 5 to 20 compounds per cluster were required , with preference for clusters of more than 20 compounds , from which a maximum of only 20 representative compounds were purchased . All materials were purchased from commercial suppliers and used without further purification . All hits subjected to further study were repurchased and identity and purity were assessed by ultra-performance liquid chromatography ( UPLC ) using an H-class Waters Acquity system . Data were acquired using Masslynx v . 4 . 1 and analyzed using the Openlynx software suite . The total flow rate was 1 . 0 mL/min and gradient program started at 90% A ( 0 . 1% formic acid in H2O ) and was changed to 95% B ( 0 . 1% formic acid in acetonitrile ) and then to 90% A . A full scan ranging from m/z 110 to 1000 in 0 . 2 s was used to acquire MS data . Compound identity was confirmed by low-resolution mass spectrometry and purity was assessed by ultraviolet spectroscopy and evaporative light scattering . All samples were required to exhibit > 85% purity . Into each well of 384-well microplates ( black polystyrene , clear bottom , tissue culture treated , Corning ) , 15 μl of medium ( DME-L [41] plus 100 μM xanthine and 10% heat inactivated fetal calf serum ) was dispensed with a liquid dispenser ( Matrix Wellmate , Thermo Scientific ) . Stock compounds , dissolved in DMSO at a fixed concentration of 10 mM , were pin-transferred ( V&P Scientific ) into the microplate to the desired final concentration using an automated robot arm . To each well of the plates , 15 μl of L . mexicana promastigotes ( strain MNYZ/BZ/62/M379 , 2 x 106/mL ) was added with the Wellmate dispenser . Microplates were incubated ( Liconic ) at 28°C and 5% CO2 for 72 h . After incubation , 10 μl of lysis/dye solution ( 5X SYBR Green I , 5% Triton X-100 in PBS ) was added to each well . Plates were shaken at 1000 rpm , incubated at room temperature for 20 min , and fluorescence signal measured ( excitation 485 nm , emission 535 nm ) with the Envision plate reader ( PerkinElmer ) . L . mexicana ( MNYZ/BZ/62/M379 ) or L . donovani ( LdBob strain ) [42] parasites expressing the Renilla luciferase gene from a rRNA gene locus were used to infect J774A . 1 macrophages . Growth of intracellular amastigotes was measured using a luminescence assay , as detailed previously [43] . The growth inhibitor activities of compounds were tested against bloodstream form Lister 427 T . brucei in 96-well plates containing 1 X 105 parasites per well in 0 . 2 ml HMI-11 medium ( Gibco/Thermo Fisher ) [44] . Compounds ( 2 μl volumes in DMSO ) were added to the parasites using serial 3-fold dilutions to cover a range of concentrations from about 10 μM to 1 nM . After 48 h incubation at 37°C under a humidified 5% CO2 atmosphere , 10 μl of 10% Triton X-100 and 100 X stock SYBR Green I ( Sigma-Aldrich ) in PBS was added and florescence measured ( excitation 497 nm; emission 520 nm ) after 1 h incubation in the dark using a Spectra Max Gemini XPS fluorimeter ( Molecular Devices ) . Data were log transformed and EC50 values were determined using GraphPad Prism 6 ( GraphPad Software ) . In the absence of growth inhibiting compounds , the parasites grew from an initial density of 5 X 105 cells/mL to ~3 X 106 cells/mL . Trypanosoma cruzi CL Brener epimastigotes were obtained from Dr . Fred Buckner of the Department of Medicine at the University of Washington . T . cruzi epimastigotes were grown in liver infusion tryptose medium and seeded in a 96-well plate at 105 epimastigotes per well in 100 μl medium . For each well , 1 μl of compound in DMSO at 100X the desired final concentration was added . Epimastigotes were exposed to a range of compound concentrations from 10 μM to 1 nM to determine EC50 . Plates were incubated at 26°C for 72 h , then 10 μl 50X SYBR Green in 1% Triton X-100 was added to each well followed by incubation with shaking at room temperature for 30 minutes . Fluorescence was read ( excitation 485nm , emission 535nm ) with the Victor2 multiplate reader ( PerkinElmer ) . All data processing and visualization were performed using GraphPad Prism 6 software . Methods for determination of liver microsomal stability , solubility , permeability of artificial membranes , Caco-2 cell permeability , stability in simulated gastric fluid , binding to mouse serum proteins , and in vivo pharmacokinetic studies are reported in Supporting Information . The BJ cell line was purchased from the American Type Culture Collection ( ATCC , Manassas , VA ) and cultured according to recommendations . Cell culture media were purchased from ATCC . Cells were routinely tested for mycoplasma contamination using the MycoAlert Mycoplasma Detection Kit ( Lonza ) . Cells were grown to 80% confluence , collected , and plated in 25 μL of medium per well in 384-well plates ( Costar 3712 ) . Compounds were diluted as described above and transferred to cells using a pin tool ( V&P Scientific ) equipped with FP1S50 pins resulting in final compound concentrations of 25 μM , and the plates incubated for 72 h at 37°C in 5% CO2 . CellTiter-Glo ( Promega ) detection reagent was added following the manufacturer’s instructions , and luminescence was measured using an EnVision ( PerkinElmer ) plate reader . Data were log transformed and EC50 values were determined using GraphPad Prism 6 ( GraphPad Software ) . Cytotoxicity of compounds to J774 . A1 macrophages was determined by dose-response curves as described previously [43] . In the absence of growth inhibitors or DMSO , the macrophages increased in number ~6-fold over 96 h in Minimum Essential Medium , employed for both macrophage infections and the toxicity assays . Drug doses were chosen based on pilot toxicology and pharmacokinetic studies . Female BALB/C mice ( compound 4 ) or C57BL6 ( compound 5 ) of 17–21 grams were purchased from Charles River Laboratories ( Wilmington , MA ) . Food and water were provided ad libitum . Two mice were used as control and another 5 mice were dosed daily via oral gavage ( 25 mg/kg with compound 4 and 50 mg/kg with compound 5 ) . Every day blood was collected by retro-orbital bleed from one animal from the treatment group for pharmacokinetics . Because compound 5 induced seizures when delivered at 50 mg/kg , blood glucose was simultaneously measured with a glucose meter ( Alpha track ) to determine whether reduced sugar levels could be a cause of this toxicity . Each mouse received two blood collections and glucose measurements over the 10-day course of treatment . Female BALB/c mice ( ~20 g ) were injected in one hind footpad with 1 x 106 stationary phase promastigotes suspended in 25 μl of phosphate buffered saline ( PBS ) . Four weeks after infection , when a small cutaneous lesion was visible in the injected footpad , cohorts of five mice were treated with either compound or vehicle alone ( 90 μl ) , delivered daily for 10 consecutive days by oral gavage using a 20-gauge x 30 mm disposable plastic feeding needle . Vehicle consisted of 10/10/40/39 mixture of ethanol/ ( PG ) /PEG400/PBS plus 1% ( weight/volume ) 2HβCD ( PG is propylene glycol , PEG is polyethylene glycol , 2HβCD is 2-hydroxy-β-cyclodextran ) . The daily dose for each compound was: compound 4 , 25 mg/kg; compound 5 , 30 mg/kg; miltefosine , 20 mg/kg . The width of the footpad ( top to bottom ) was measured with calipers before injection of parasites ( day 0 ) and weekly from weeks 4–12 . The width of the uninfected contralateral footpad was also measured each week , and its width was subtracted from that of the infected footpad to determine lesion size . All research involving animals was carried out with the approval of the Institutional Animal Care and Use Committee of either St . Jude Children’s Research Hospital or the Oregon Health & Sciences University . The study was conducted adhering to the guidelines for animal husbandry of each institution .
A summary of the HTS workflow and quality control data is shown in Fig 1 . The initial promastigote screen was performed with the St . Jude Chemical Biology & Therapeutics ( CBT ) library consisting of 596 , 414 compounds . Library compounds were filtered by several computational methods [35 , 36] to remove those likely to have undesirable physical or biological properties and biased towards oral bioavailability . In this way we focused the collection on compounds most likely to be effective in cellular models of activity , without structural features that would pose a challenge to drug development [33] . In the primary screen , compounds were applied to promastigotes of L . mexicana at a fixed concentration of 10 μM , and parasite proliferation was monitored , following a 72 h incubation , by quantifying total DNA content after lysis using the nucleic acid binding dye SYBR Green I [45] . The raw data for the HTS campaign are summarized in Fig 1B as a scatterplot of normalized percent growth inhibition relative to the control drug pentamidine , which gives 100% inhibition of proliferation under these conditions . The scatterplot demonstrated ample signal separation between the positive ( green ) and negative ( red ) controls throughout the HTS campaign and a well-defined activity distribution of test compounds ( blue and black ) . The fidelity and quality of the HTS assay were assessed using two metrics: Z-prime and EC50 of the control ( pentamidine ) that were calculated for each screening plate . The entire screen produced a median Z-prime value of 0 . 81 ( interquartile range: 0 . 75–0 . 85 , Fig 1C ) and a consistent EC50 value of pentamidine ( median 2 . 3 μM , interquartile range: 1 . 7–3 . 1 ) indicating the assay was consistent throughout the screen . The assay’s discriminatory power was assessed using Receiver Operator Characteristic ( ROC ) analysis [46] as described [33] . This method helped define an optimal cutoff for designating primary hits by balancing the likelihood of a true positive with acquiring a reasonable total number of hits . Briefly , compounds were stochastically selected from the HTS screening set to sample the primary assay results according to the distribution of observed activities ( ranging from 0 to 100% activity ) . The selected compounds were plated in a 10-point dose-response and re-evaluated in the HTS assay . True positives were defined as any compound yielding a well-behaved , saturating sigmoidal curve in the dose-response assay . The ROC curve , shown in Fig 1D , demonstrated that the assay has good discriminatory power , with an area under the curve ( AUC ) of 0 . 89 ( a perfect assay would have an AUC 1 . 0 , whereas a random assay has an AUC of 0 . 5 ) . Based on this analysis , a cut-off value of > 65% inhibition was chosen , resulting in 2 , 703 primary hits with an expected true positive rate of 85% . It is worth noting that a significant number of true hits likely remain in the group of compounds exhibiting growth inhibition of lower the 65% cut-off activity , and these compounds were not considered in this manuscript . To confirm the activity of the primary hits and improve confidence that they would be reasonable starting points for drug development , a variety of secondary screens and analyses were employed ( Fig 1A ) . First , EC50 values were determined against the promastigotes using a 10-point dose-response , run in triplicate , with concentrations ranging from 0 . 0005–25 μM . Compounds that reproducibly exhibited EC50 activity lower than 2 μM were considered validated hits . In parallel , mammalian cell growth inhibition was determined using in vitro proliferation assays with normal human fibroblasts ( BJ cells ) . Compounds inhibiting proliferation of BJ cells at concentrations lower than 20 μM were deprioritized . To further triage the hits , we carried out a chemical structure analysis for the 2 , 703 primary hit compounds utilizing topology mapping and clustering methodology [16] . We identified a wide range of chemotypes , including several scaffolds with potential structure-activity relationships ( SARs ) , based on their dose response activity ( Fig 2 ) . Validated hits were then culled by eliminating scaffolds with less favorable drug development properties such as charged planar structures , reactive electrophilic warheads , known pan-assay interference motifs ( PAINS ) [47] , and compounds displaying gross rule of five noncompliance [48] . Finally , we prioritized scaffolds with the possibility of facile chemical modification to generate a substantial number of structural analogs for future SAR and structure-property relationship ( SPR ) studies . Among the 2703 hits , 230 compounds exhibited both an EC50 of < 2 μM for L . mexicana promastigotes and a TI > 5 ( based on mammalian fibroblast toxicity ) . These were chosen as candidates for further study . From these 230 compounds , we were able to repurchase 113 from commercial vendors . These compounds were characterized for purity by ultra-performance liquid chromatography using ultraviolet spectroscopy and evaporative light scattering detection [49] and identity by mass spectrometry . All validated compounds were profiled for activity against intracellular amastigotes , the disease-causing stage of the life cycle . Intracellular amastigote activity was determined using a strain of L . mexicana in which the Renilla luciferase gene was integrated into the rRNA locus [43] , allowing robust expression for measuring amastigote growth within cultured macrophages [50] . All 113 compounds were applied at 1 μM concentration for 96 h to J774A . 1 macrophages infected with L . mexicana luciferase-expressing parasites . Of the compounds tested 55 inhibited amastigote growth by > 70% . Next , we generated dose-response curves for these 55 compounds against intracellular amastigotes and independently against J774A . 1 macrophages to establish the relative potency of each compound against the pathogen and its host cell . Those that had EC50 values < 1 μM and TI values > 10 for macrophages were selected from the 55 compounds , as suggested for lead identification for leishmaniasis [12] , and several compounds were then removed due to known biological liabilities of scaffolds ( manual curation , Fig 1 ) . The nine remaining compounds , each representing a unique chemical scaffold ( compounds 1–9 , Fig 3 ) , were designated top hits ( Fig 1 ) . Notably , this screening strategy successfully identified several known antileishmanial scaffolds , including compounds 1 , 2 , 3 , and 4 , thus providing further validation of the screen . The alkaloid cephaeline ( 1 ) , a known irritant of gastric mucosa and component of ipecac , has been shown to be potent against L . mexicana and L . donovani intracellular amastigotes [51] . Another known scaffold , the quinazoline-2 , 4-diaminoquinazolines , represented by compound 2 , has been studied extensively and shown to have activity against L . donovani , and L . amazonensis [52 , 53] . Compound 2 is also present in the malaria box of compounds active against Plasmodium falciparum and has been shown to have activity against L . infantum [54] . We also found a member of the 2 , 4-diaminopyrimidine scaffold , compound 3 , some of which are selective against L . major amastigotes , with EC50 values in the low μM range and in once case with a therapeutic index ( TI ) as high as 130 [55] . The compounds in that study share the 2 , 4-diaminopyrimidine scaffold with compound 3 , but they differ in having a benzyl substitution at the 5 position of the pyrimidine ring rather than modifications on the 2- and 4-amino substituents that are present in compound 3 . Finally , various 4H-chromen-4-ones , similar to compound 4 , are active against L . major [56] . Of the validated scaffolds included in the HTS campaign , the three that exhibited the widest SAR range ( 7–88 fold ) were the 2 , 4-diaminoquinazolines , 2 , 4-diaminopyrimidines , and 4H-chromen-4-ones ( Fig 2 ) . Potency of the nine compounds against intracellular amastigotes of L . donovani was also quantified to assess each compound’s potential to control this agent of fatal visceral leishmaniasis ( Fig 3 ) . We have recently reported that compound 5 is also potent against another kinetoplastid parasite , the bloodstream form of Trypanosoma brucei ( EC50 value of 0 . 027 μM ) [58] and active in vivo in a murine model of African trypanosomiasis ( manuscript in preparation ) . Thus , we also evaluated the other compounds for activity against the related pathogen Trypanosoma brucei . As noted for the broad spectrum kinetoplastid proteasome inhibitor GNF6702 [32] , compounds exhibiting activity against multiple parasites are especially interesting , as such scaffolds can be explored for therapies against multiple neglected parasitic diseases . All nine compounds were potent ( EC50 < 0 . 6 μM ) against both the L . mexicana and L . donovani intracellular amastigotes . Often , potency correlated well between the two species , although there were significant differences for some compounds ( e . g . , compounds 2 , 4 , 8 , and 9 ) . While none of the compounds affected the proliferation of BJ cells at concentrations as high as 20 μM , most of the compounds reduced viability of macrophages with half-maximal lethal dose ( LD50 ) values around 1–10 μM . Only compounds 2 and 4 demonstrated no reduction in viability in dose-response studies against the host macrophage J774A . 1 , suggesting these compounds may afford the best selectivity for inhibiting parasite growth relative to toxicity toward the host macrophage or other mammalian cells . Thus , all of the nine compounds tested afforded favorable therapeutic indices ( > 50 ) , except compound 3 . Notably , compounds 5 , 8 , and 9 exhibited good potency ( < 0 . 3 μM ) against bloodstream form T . brucei . To determine whether any of the top hits might also be effective against the related kinetoplastid parasite T . cruzi , we performed dose-response curves with compounds 4 , 5 , 8 , and 9 against epimastigotes and found either no inhibition ( 4 , n = 3 ) or EC50 values of 0 . 086 ± 0 . 03 μM ( 5 , n = 4 ) , 0 . 33 μM ( 8 , n = 1 ) , and 2 . 1 ± 0 . 07 μM ( 9 , n = 2 ) , respectively . Hence , each of these latter scaffolds is of potentially high interest for development of drugs against multiple species of kinetoplastid parasites . Together , these data suggest the seven compounds not previously reported to possess antileishmanial activity ( only 1 and 2 have been documented previously ) can be good starting points for discovering new antileishmanials . Herein , we chose to further profile compounds 4 and 5 , representing the 4H-chromen-4-ones and p-chloronitrobenzamides scaffolds , respectively . Compound 4 was chosen for its distinct lack of toxicity against host macrophages and compound 5 was chosen for its cross-species potency . We suggest that similar studies could be undertaken using the other validated compounds from our two-stage phenotypic screening campaign . In order to evaluate compounds 4 and 5 for in vivo studies , we measured the in vitro ADME physiochemical properties likely to be predictive of oral bioavailability ( Table 1 ) . First , we looked at solubility in an aqueous buffer ( pH = 7 . 4 ) and ability to cross an artificial ( parallel artificial membrane permeability , PAMPA ) or cellular ( Caco-2 ) membrane . Compound 4 exhibited good solubility ( 67 μM ) and moderate membrane permeability ( Table 1 ) suggesting a high predicted absorption across the intestinal epithelium ( ~85% ) , and low probability of being a substrate of the drug resistance pumps expressed by Caco-2 cells ( efflux ratio < 2 ) . Compound 5 showed moderate permeability in both the PAMPA and Caco-2 assays as well as an acceptable efflux ratio of 1 . 92 ( Table 1 ) . Compound 5 exhibits low aqueous solubility ( 0 . 3 μM ) but we anticipated that this could be compensated by formulation for delivery [59] . Next , we investigated the stability of both compounds in simulated gastric fluid and in microsomal models of oxidative metabolism . Both compounds exhibited high stability in simulated gastric fluid ( t1/2 > 24 h ) and demonstrated good metabolic stability ( t1/2 > 4 h for all species ) in liver microsome preparations from mouse , rat , and human . Compounds 4 and 5 also showed modest ( <50% ) binding to mouse plasma proteins , below the level of the positive control drug propranolol ( Table 2 ) . Criteria that have been suggested as promising for an orally bioavailable compound include: aqueous solubility > 1 μM but ideally > 100 μM [60] , PAMPA permeability coefficient of > 1x10-5 cm/sec represents high permeability , Caco-2 cell permeability coefficient of > 1x10-6 cm/sec [60] , Caco-2 cell efflux ratio < 2 represents no efflux , gastric stability > 24 h , t1/2 in microsomes > 30 min [12] . However , these values only represent broad guidelines , and many efficacious drugs violate them . Overall , the in vitro ADME data suggest that the scaffolds of compounds 4 and 5 would be appropriate for development into orally bioavailable antileishmanial compounds . To further evaluate the potential of 4 and 5 in vivo , we performed preliminary single oral dose pharmacokinetic studies in mice . Following a single oral gavage ( PO ) of 4 in mice at 25 mg/kg ( Fig 4 ) the plasma concentration remained above its EC50 of 0 . 08 μM for approximately 20 h . Compound 4 reached a peak plasma concentration ( Cmax ) of 3 . 2 μM within 1 h ( tmax ) of dosing , afforded an AUC of 16 . 7 μM . h , and an elimination half-life ( t1/2 ) of 3 h ( Table 3 ) . Following PO dosing of 5 at 50 mg/kg ( Fig 4 , Table 3 ) , the plasma concentration remained above the EC50 of 0 . 022 μM for roughly 48 h , with a Cmax of 6 . 49 μM , a tmax of 4 h , an AUC of 83 . 2 μM*h , and a t1/2 of 7 . 1 h . Thus , both compounds exhibited good plasma exposure and sustained plasma concentrations above an efficacious dose ( EC50 ) for more than 12 h following a single oral gavage dosing using our standard formulation ( 10/10/40/39 , EtOH/PG/PEG/PBS ( 7 . 4 ) ( v/v ) and 1% ( w/v ) HβCD ) . These results strongly suggested that both compounds were appropriate candidates for efficacy evaluation in the murine model of cutaneous leishmaniosis . Next , we sought to determine the allowable dosing range for our efficacy model by carrying out dose-ranging tolerability studies . When compound 4 was dosed by oral gavage at 50 mg/kg , half of the animals exhibited seizure-like behavior . Blood chemistries revealed a very low glucose level in plasma ( 23–40 mg/dl for treated mice compared to 185–251 mg/dl for untreated mice ) . This observation might suggest blockage of a kidney and/or an intestinal glucose transporter . When we repeated the same experiments at 25 mg/kg , no seizures were seen and the glucose level of each animal remained within normal limits at both Cmax and Cmin . Daily oral administrations of 4 at 25 mg/kg were well-tolerated in all study animals , no significant changes in either clinical chemistry or complete blood counts were observed , and there were no other test article-related effects noted in the liver or any other tissues . For compound 5 dosed at 50 mg/kg in mice , a 10-day toxicity study revealed that animals reduced food intake and lost more than 10% of their weight overtime . The observed suppressed appetite was resolved by dose reduction to 30 mg/kg . Thus , we employed 25 mg/kg of 4 and 30 mg/kg of 5 in the efficacy model . No weight loss or other toxicity was observed at these doses . Next we assessed the potential of compounds 4 and 5 to control disease in a murine model of cutaneous leishmaniasis [30] . We infected BALB/c mice with L . mexicana via footpad injections on day zero , allowed incipient lesions to develop for four weeks , and then treated cohorts of five animals with each compound for 10 consecutive days by oral gavage . In addition , five mice were treated with 20 mg/kg of the only orally available approved antileishmanial drug , miltefosine , as a positive control and with vehicle alone as a negative control . Footpad widths were measured from 4–12 weeks post-infection ( Fig 5 ) . For vehicle-treated mice , the lesions grew steadily up to 2 mm width , at which time mice were euthanized . Miltefosine reduced lesion size from the initial dimension and was able to maintain growth inhibition for eight weeks following treatment . Both compounds 4 and 5 controlled lesion size at dimensions similar to that at the time of compound dosing ( 4 weeks ) until week 9 , well after stopping oral administration . After week 9 , the footpad lesions began to increase in size . Hence , oral dosing of both compound 4 and 5 controlled disease progression during the dosing period and for a significant period of time after dosing stopped but neither was as efficacious as miltefosine . The partial control of virulence exhibited by our HTS hits , without any optimization , strongly suggests both compounds are novel early leads for the development of orally available antileishmanials . Given the synthetic tractability of these scaffolds [58 , 61] , we envision a rapid timeline for the development of optimized leads with enhanced therapeutic properties .
High throughput phenotypic screening offers a powerful tool to discover therapeutically relevant leads for drug discovery [62] . The HTS campaign described in this paper represented part of a larger effort to identify selective inhibitors of hexose transporters from various parasitic protozoa [45] , including a transgenic strain of L . mexicana , Δlmxgt1-3[pLmxGT2] [63] . The results reported here began as a second , adventitious outcome of that screen , where we identified 2 , 703 compounds that significantly inhibited the growth of promastigotes of the Δlmxgt1-3 parasites employed as the cellular expression vehicle for the hexose transporters . In this study , we leveraged this secondary outcome to identify novel orally available antileishmanials . We emphasize that since most of the top hits against L . mexicana are also potent against an agent of lethal visceral leishmaniasis , L . donovani ( Fig 3 ) , this screen is of potential therapeutic value for both cutaneous and visceral leishmaniasis . The initial screen against the L . mexicana promastigote form of the parasite was highly robust with a median Z value of 0 . 81 and an AUC of 0 . 893 for the ROC curve . In addition to the antileishmanial compounds that were carried through the secondary validation assays ( vide supra ) , the HTS identified a variety of inhibitors known to be active against various Leishmania strains: crystal violet ( EC50: 0 . 29 μM ) [64] , disulfiram ( EC50: 0 . 50 μM ) [65] , thiram ( EC50: 1 . 77 μM ) [65] , actinomycin D ( EC50: 0 . 36 μM ) [28] , anisomycin ( EC50: 0 . 58 μM ) [66] , and avicin ( EC50: 1 . 50 μM ) [16] . The rediscovery of these known inhibitors provided another level of validation and confirmed the screen’s ability to identify active antileishmanial compounds . Our motive behind the sequential screening of promastigotes followed by amastigotes was to eliminate promastigote-specific hits . In the process we identified multiple hits that were potent inhibitors of both promastigote and amastigote growth and removed compounds that inhibited growth of either host macrophages ( J774A . 1 ) or normal fibroblasts ( BJ cells ) ( Fig 3 ) . Hence , while there has been much discussion about the relative merits of screens employing promastigotes , axenic amastigotes , and intracellular amastigotes ( see Introduction ) , the sequential approach employed here sidesteps that debate and identified multiple scaffolds with potential for further development toward orally bioavailable antileishmanial drugs . Three scaffolds from the sequential screen stood out as promising candidates due to the wide range of SAR inherent in our screening data set ( Fig 2 ) , the high potency of certain exemplars , and good TI values ( Fig 3 ) : 2 , 4-diaminoquinazolines ( 2 ) , 2 , 4-diaminopyrimidines ( 3 ) , and 4H-chromen-4-ones ( 4 ) . The 2 , 4-diaminoquinazolines have been disclosed previously as potential antileishmanials [52 , 53 , 67] . This scaffold has been explored by medicinal chemistry , and one compound was identified that exhibited an EC50 of 0 . 15 μM against L . donovani amastigotes and a TI of 100 [52 , 53 , 67] . These studies also demonstrated that Leishmania dihydrofolate reductase ( DHFR ) is inhibited by 2 , 4-diaminoquinazolines , highlighting this essential enzyme as one target for this class of antileishmanials . Similarly , 2 , 4-diaminopyrimidines have been shown to have μM potency against Leishmania amastigotes [55] . Compound 3 is more potent than the previously studied 2 , 4-diaminopyrimidines [55] , and it has substitutions on the 2 , 4-amino groups , unlike previously characterized 2 , 4-diaminopyrimidines . These results suggest that substitution at these positions may be important for potency and imply that additional modifications at these sites may be worth exploring . Furthermore , 2 , 4-diaminopyrimidines are structurally related to classical DHFR inhibitors such as pyrimethamine and trimethoprim [55] that also selectively inhibit the essential Leishmania DHFR , providing a potential molecular target for this family of antileishmanials . However , two distinct enzymes in L . major , DHFR and pteridine reductase 1 ( PTR1 ) , can reduce folate , and amplification of the PTR1 gene can confer methotrexate resistance upon the parasite by metabolically circumventing inhibition of DHFR by this antifolate [68] . Hence , effective inhibitors of DHFR may also need to inhibit PTR1 , thus complicating chemotherapy against this target . 4H-chromen-4-ones [56] , and related chroman-4-ones [69] , have been demonstrated to have activity against both T . brucei and L . major , and they bind to and inhibit the critical [70] enzyme PTR1 that is present in kinetoplastid parasites , but not in mammals . These results with structurally related compounds suggest that PTR1 may be a principle target of compound 4 . However , the compounds tested in this previous work had an aromatic substituent at the 2 position of the chromen-4-one ring rather than at the 3 position . Those compounds exhibited much lower potency against T . brucei and L . major , with EC50 values in the micromolar range , compared to compound 4 against L . mexicana or L . donovani amastigotes ( Fig 3 ) . Furthermore , compound 4 has lower toxicity , a higher TI ( Fig 3 , no inhibition of J774A . 1 macrophages up to 10 μM concentration ) , and greatly superior pharmacokinetic properties ( Fig 4 , Table 3 ) compared to compounds tested by Borsari et al . [56] , where the top hit exhibited a half-life of 7 . 6 min in mice . These observations suggest that structural features present in compound 4 may provide a route for developing this scaffold toward more optimal lead compounds against Leishmania parasites . Variants of the isoflavone scaffold present in 4 have been employed as dietary supplements and are known to have phytoestrogen and antioxidant properties [71] . Thus , there has been a long-standing interest in the development of synthetic routes to access these desirable properties [72 , 73] . More recently , rapid synthetic routes to access highly substituted hydroxylated isoflavones such as 4 have been published [61] . Compound 5 has been identified previously [58] by our laboratory as active against all Trypanosoma species in vitro and efficacious against T . congolense and T . b . rhodesiense in vivo ( manuscript in preparation ) . This scaffold is especially interesting , since it may act on a common cellular target found among kinetoplastids and could be developed as both antileishmanial and antitrypanosomal drugs . In addition , the in-house experience with the scaffold in in vivo models inspired confidence regarding its oral activity . Therefore , we chose to progress compounds 4 and 5 for further in vitro ADME and in vivo pharmacokinetic and pharmacodynamic testing . We note that compounds 8 and 9 also exhibit significant potency toward T . brucei and may therefore be of special interest for future investigations . Compounds 4 and 5 were both able to partially control the size of an incipient cutaneous lesion when delivered orally at 25–30 mg/kg for 10 days , compared to mice that received vehicle alone . However , they were not as efficacious as the currently employed oral drug miltefosine . This efficacy in vivo indicates that improvements will be required to further address the potential of these scaffolds for drug development . In particular , analogs that exhibit higher potency and/or lower toxicity in animals may achieve greater efficacy or allow higher dosing . The ability to extensively modify both scaffolds by medicinal chemistry offers the potential to generate libraries of analogs of each lead whose members can then be tested for improved ADME , PK , toxicity , and in vivo efficacy in animal models of both cutaneous and visceral leishmaniasis . Additionally , the combination of potency , selectivity , and identification of DHFR as a potential molecular target all suggest that further exploration of the 2 , 4-diaminoquinazoline and 2 , 4-diaminopyrimidine scaffolds , represented by compounds 2 and 3 respectively , may be warranted . Overall , this work demonstrates how sequential screening of promastigotes , which are especially amenable to HTS assay development , followed by hit validation in the disease causing intramacrophage amastigotes can be used to successfully identify novel antileishmanial scaffolds . The promising pharmacokinetic profile and significant in vivo efficacy of our newly identified scaffolds strongly suggests that additional medicinal chemistry optimization may yield orally available anti-parasitic drugs .
|
Leishmaniasis , caused by the protozoa of the Leishmania species , represents a spectrum of diseases that afflicts roughly 12 million individuals worldwide . Current drug therapies for this parasitic disease are suboptimal because they are toxic , expensive , difficult to administer , and subject to drug resistance . In order to identify new and improved drug candidates , we screened a large library of small molecules for compounds that inhibit parasitic growth inside mammalian host macrophages , and have low toxicity toward the macrophages . We discovered two compounds that significantly impaired disease progression when administered orally in an animal model of cutaneous leishmaniasis . The promising pharmacokinetic and in vivo efficacy profile of the compounds make them attractive starting points for pharmaceutical development .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"methods",
"Results",
"Discussion"
] |
[
"blood",
"cells",
"medicine",
"and",
"health",
"sciences",
"immune",
"cells",
"immunology",
"tropical",
"diseases",
"microbiology",
"parasitic",
"diseases",
"protozoan",
"life",
"cycles",
"animal",
"models",
"developmental",
"biology",
"model",
"organisms",
"experimental",
"organism",
"systems",
"neglected",
"tropical",
"diseases",
"molecular",
"biology",
"techniques",
"pharmacology",
"promastigotes",
"research",
"and",
"analysis",
"methods",
"infectious",
"diseases",
"white",
"blood",
"cells",
"zoonoses",
"animal",
"cells",
"life",
"cycles",
"mouse",
"models",
"protozoan",
"infections",
"molecular",
"biology",
"amastigotes",
"pharmacokinetics",
"molecular",
"biology",
"assays",
"and",
"analysis",
"techniques",
"cell",
"biology",
"library",
"screening",
"leishmaniasis",
"biology",
"and",
"life",
"sciences",
"protozoology",
"cellular",
"types",
"macrophages"
] |
2017
|
Discovery of novel, orally bioavailable, antileishmanial compounds using phenotypic screening
|
The methionine salvage pathway is responsible for regenerating methionine from its derivative , methylthioadenosine . The complete set of enzymes of the methionine pathway has been previously described in bacteria . Despite its importance , the pathway has only been fully described in one eukaryotic organism , yeast . Here we use a computational approach to identify the enzymes of the methionine salvage pathway in another eukaryote , Tetrahymena thermophila . In this organism , the pathway has two fused genes , MTNAK and MTNBD . Each of these fusions involves two different genes whose products catalyze two different single steps of the pathway in other organisms . One of the fusion proteins , mtnBD , is formed by enzymes that catalyze non-consecutive steps in the pathway , mtnB and mtnD . Interestingly the gene that codes for the intervening enzyme in the pathway , mtnC , is missing from the genome of Tetrahymena . We used complementation tests in yeast to show that the fusion of mtnB and mtnD from Tetrahymena is able to do in one step what yeast does in three , since it can rescue yeast knockouts of mtnB , mtnC , or mtnD . Fusion genes have proved to be very useful in aiding phylogenetic reconstructions and in the functional characterization of genes . Our results highlight another characteristic of fusion proteins , namely that these proteins can serve as biochemical shortcuts , allowing organisms to completely bypass steps in biochemical pathways .
The amino acid methionine and its derivative S-adenosylmethionine ( SAM ) are essential substrates in a variety of cellular reactions , including protein and nucleic acid methylation and polypeptide , polyamine and ethylene syntheses [1]–[5] . Ethylene , the hormone that regulates plant growth and development , is synthesized when the aminobutyrate group of SAM is released as 1-aminocyclopropane-1-carboxylic acid and then oxidized to form ethylene . SAM contributes to the synthesis of polyamines by providing the source of the groups spermidine and spermine [6] . The synthesis of polyamines and ethylene from the precursor SAM also results in the production of the byproduct 5′-methylthioadenosine ( MTA ) , which can be converted back to methionine via the methionine salvage pathway . The methionine salvage pathway is highly conserved in all domains of life [4] , and its proper function is particularly important in cells that generate large quantities of polyamines and ethylene [7] , [8] . While the complete methionine salvage pathway has been known for some years in bacteria [4] , [5] , only recently were all of the enzymes of the pathway characterized in an eukaryotic organism , the yeast Saccharomyces cerevisiae [9] . Sekowska et al . [4] surveyed the complete genome sequences of several bacteria to determine the enzymes of the methionine salvage pathway ( Figure 1 ) . In bacteria there are two main variations in the pathway . Some bacteria , like Pseudomonas aeruginosa , convert MTA into S-methyl-5thio-D-ribose-1-phosphate ( MTRP ) in a single step using the enzyme MTA phosphorylase ( MtnP; in this paper we will use the nomenclature for the enzymes in the pathway proposed in [4] ) . This same step is catalyzed by two separate enzymes in other bacteria , like Bacillus subtillis . In these organisms , mtnN , a hydrolase , first removes the adenine from MTA , converting it into S-methyl-5-thio-D-ribose ( MTR ) . This compound is then converted into MTRP by the addition of a phosphate group by mtnK , a kinase . The next step in the pathway involves the use of an isomerase , mtnA , to convert MTRP into S-methyl-5-thio-D-ribulose-1-phosphate ( MTRuP ) . MTRuP is then dehydrated by a dehydratase , mtnB , generating 2 , 3-diketo-5-methylthiopentyl-1-phosphate . In some bacteria , like Klebsiella pneumoniae , this diketone is converted into 1 , 2-dihydroxy-3-keto-5-methylthiopentene by an enolase-phosphatase , mtnC . In other organisms , like Bacillus subtillis , this reaction is carried out in two steps . The diketone is first converted into 2-hydroxy-3-keto-5-methylthiopentenyl-1-phosphate by an enolase , mtnW , and a phosphatase , mtnX , finally adds the phosphate group to the resulting molecule , forming 1 , 2-dihydroxy-3-keto-5-methylthiopentene . The last two steps of the pathway are the formation of 4-methylthio-2-oxobutanoate ( MOB ) by the action of mtnD , a dioxygenase , followed by the transamination of this product to methionine . The last transamination step can be performed by many different transaminases depending on the organism [4] , [9] . When a specific enzyme exists to catalyze this step , it is called mtnE; however , because of the variability of the enzymes that can potentially catalyze this step in vivo , we will use the term mtnE to refer to any transaminase that catalyses this last step . The whole pathway , including the two main variations , is depicted in Figure 1 . Pirkov et al . [9] used a computational and experimental approach to show that , in yeast , MTA is converted to MOB by the enzymes Meu1 ( mtnP ) , Mri1 ( mtnA ) , Mde1 ( mtnB ) , Utr4 ( mtnC ) and Adi1 ( mtnD ) . The final transaminase step can be catalyzed by several transaminases in yeast . For example , both the aromatic amino acid transaminases ( Aro8 and Aro9 ) and the branched chain amino acid transaminases ( Bat1 and Bat2 ) are capable of transaminating MOB into methionine [9] . For simplicity , in the remainder of this paper we will refer to the yeast enzymes using the pathway nomenclature proposed in [4]; thus we will refer to Meu1 as mtnP , Mri1 as mtnA , and so forth . Sekowska et al . [4] and Ashida et al . [5] noted that in Arabidopsis thaliana mtnB and mtnC are coded for by a single gene and are most likely part of a fusion protein that performs both functions . We propose to call this protein mtnBC . Ashida et al . [5] also noted that the genes coding for mtnB and mtnD are fused in Tetrahymena thermophila , suggesting that a single protein can perform the functions of mtnB and mtnD in this organism . We propose to call this protein mtnBD . It is important to stress that the alternative pathways described in [4] are catalyzed by different enzymes . Although mtnN and mtnK together catalyze the same reaction as mtnP , they share no sequence similarity with mtnP . The same is true for mtnW and mtnX . While together they catalyze the same reaction as mtnC , they share no sequence similarities with mtnC . The fused proteins in Tetrahymena and plants are different in that they contain the same domains as their corresponding components . For example , the N-terminus of the mtnBD enzyme in Tetrahymena shares a strong sequence similarity with the entire sequence of mtnB , while the C-terminus shares a strong similarity to all of the sequence of mtnD . The same is true for the mtnBC enzyme in Arabidopsis . Here we investigated the methionine salvage pathway in Tetrahymena thermophila . We used similarity searches to show that in addition to the fusion of mtnB and mtnD , this organism has another fusion protein in the methionine salvage pathway , a fusion of the mtnK and mtnA enzymes ( which we will refer to as mtnAK ) . Homologs to proteins catalyzing all the steps of the methionine salvage pathway are present in Tetrahymena , with the exception of mtnC , the protein that catalyzes the step between those of mtnB and mtnD in the pathway . The lack of an mtnC homolog in the Tetrahymena genome led us to hypothesize that the fusion protein mtnBD might perform the function of mtnC in addition to those of mtnB and mtnD in this organism . We performed experiments that show that this hypothesis is correct and that the fusion mtnBD is a trifunctional enzyme able to catalyze three steps in the salvage pathway , those of mtnB , mtnC and mtnD . The two identified fusions make Tetrahymena able to recycle MTA into methionine with the use of only four enzymes , as opposed to the six enzymes required by yeast or the eight required by B . subtillis ( Figure 1 ) .
We used the sequences of the methionine salvage pathway enzymes from yeast and Bacillus subtillis as queries in blastp searches to find homologs in Tetrahymena . Because the pathway is slightly different in these two organisms – yeast uses mtnP and mtnC , while B . subtillis uses mtnN and mtnK , and mtnW and mtnX ( Figure 1 ) – some genes are unique to each organism . Table 1 shows the best blastp hit for each of the genes in Tetrahymena . Because the last step in the pathway can be catalyzed by many different transaminases in different organisms [4] , [9] , we did not search specifically for homologs of these genes in Tetrahymena . However , Tetrahymena has homologs to both Bat1 and Bat2 ( TTHERM_00765280; e = 6e-79 to both Bat1 and Bat2 ) and to Aro8 and Aro9 ( TTHERM_00140940; e = 2e-9 to Aro9 and 1e-5 to Aro8 ) , shown to be the enzymes responsible for the last step of the pathway , the MOB transamination reaction , in yeast [9] . The results from the blast searches suggest that Tetrahymena has homologs to mtnN , mtnK , mtnA , mtnB and mtnD ( Table 1 ) . A single Tetrahymena protein ( XP_001031773 ) is the best hit to both mtnK and mtnA , and another protein ( XP_001025046 ) is the best hit to both mtnB and mtnD . Further blast searches using each of these Tetrahymena proteins as queries showed that the C-terminus of XP_001031773 has strong sequence similarity to the sequence of mtnA and that its N-terminus has strong sequence similarity to the sequence of mtnK . The same is true for XP_001025046; its N-terminus has strong sequence similarity to the sequence of mtnB , and its C-terminus has strong similarity to the sequence of mtnD . These results indicate that mtnA and mtnK might form a fusion protein ( mtnAK ) , and mtnB and mtnD might form another fusion protein ( mtnBD ) in Tetrahymena . To our knowledge , the fusion mtnAK has not been previously described in any organism . To further verify if the genes identified as part of the methionine salvage pathway in Tetrahymena are true orthologs to the yeast and B . subtillis genes , we ran blastp searches against yeast and B . subtillis proteins in the RefSeq database using the Tetrahymena proteins as queries . In each case , the best hit to the Tetrahymena gene in the other genome was the original gene we used in the first search , indicating that all genes identified are reciprocal best blast hits – a good indication that the genes are orthologs . In the case of the fusion genes , the two best hits in yeast and B . subtillis were the components of the fusion , as expected . The absence of a homolog to mtnP implies that Tetrahymena hydrolyzes the adenine from MTA using mtnN , instead of using mtnP to convert MTA into MTRP like yeast does . It seems that the resulting molecule could be converted directly to MTRuP since we identified what seems to be a fusion of the mtnK and mtnA enzymes . As previously described [5] , mtnB and mtnD also seem to be fused in Tetrahymena . Surprisingly , we could not find Tetrahymena homologs to either mtnC , or to mtnW or mtnX , the enzymes responsible for converting 2 , 3-diketo-5-methylthiopentyl-1-phosphate into 1 , 2-dihydroxy-3-keto-5-methylthiopentene . The lack of homologs to this step in the pathway led us to hypothesize that the fusion of mtnB with mtnD might be able to convert MTRuP into MOB without the help of mtnC or mtnW-mtnX . It should be noted that although the yeast mtnP protein doesn't bring any Tetrahymena hits , the human mtnP ( accession: NP_002442 ) brings two Tetrahymena hits with evalues of 1×10−14 ( XP_001020972 ) and 2×10−10 ( XP_001020274 ) , respectively . These Tetrahymena genes , however , are other types of purine nucleoside phosphorylases . XP_001020972 is an inosine and guanosine-specific phosphorylase; its best hit in humans is to a nucleoside phosphorylase ( NP_000261; evalue = 2×10−61 ) and not to mtnP ( evalue = 1×10−14 ) ; its best hit in yeast is to NP_013310 ( evalue = 4×10−48 ) and not to mtnP ( evalue = 0 . 019 ) . The best hits of the other Tetrahymena nucleoside phosphorylase , XP_001020274 , are NP_013310 in yeast ( evalue = 5×10−45 ) and NP_000261 ( evalue = 5×10−54 ) in humans , and not the mtnP enzymes in these organisms ( yeast mtnP , evalue = 2 . 5; human mtnP , evalue = 2×10−10 ) . Thus , it seems that although Tetrahymena has at least two purine nucleoside phosphorylase proteins , neither of these share functional homology with mtnP ( they are not best reciprocal blast hits ) . The results of our blast searches suggested that the mtnK and mtnA enzymes are fused in Tetrahymena , as both proteins hit the same Tetrahymena gene . Based on the nomenclature proposed by Sekowska et al . ( 2004 ) , we will refer to this enzyme as mtnAK . The Tetrahymena protein is 779 amino acids long . Approximately the first 350 amino acids hit mtnA; mtnA in B . subtillis is 353 amino acids long , and its alignment to the N-terminus of mtnAK has a percent identity of 38% and spans residues 17 to 353 . The second half of mtnAK hits mtnK; mtnK in B . subtillis is 397 amino acids long , and its alignment to the C-terminus of mtnAK has a percent identity of 38% and spans residues 30 to 393 . To determine whether the Tetrahymena protein XP_001031773 is really a fusion of mtnK and mtnA or just an artifact of incorrect gene prediction , we performed a tblastn search of the protein sequence of mtnAK against the NCBI EST database for evidence that this protein is expressed in Tetrahymena . The Tetrahymena EST sequences TT1BI24TH ( acc: FF565362 ) and TT1BI24TV ( acc: FF565363 ) correspond to the 5′ and 3′ of a single cDNA clone and hit the N- and C-termini of the mtnAK fusion protein , with evalues of 1×10−125 and 2×10−125 , respectively ( Figure 2 ) . This strongly suggests that the genes coding for mtnK and mtnA are fused in Tetrahymena and code for an expressed fused protein , with the N-terminus corresponding to mtnA and the C-terminus corresponding to mtnK . We used the sequence of the mtnAK protein in Tetrahymena to search for other organisms that might have this fusion in the NR database from GenBank using blastp . The same fusion of mtnK and mtnA is present in at least some stramenopile and plant species like Ostreococcus tauri , Phaeodactylum tricornutum , Micromonas pusilla and Micromonas sp . RCC299 . Interestingly , in both sequenced Micromonas genomes this gene is annotated as a fusion of a translation initiation factor with methylthioribose kinase ( mtnK ) . This is due to the fact that methylthioribose-1-phosphate isomerase ( mtnA ) shares sequence similarities with eIF-2B-alpha [4] . However , because it is fused with another gene in the methionine salvage pathway and has strong sequence similarity with mtnA , this gene is most likely a fusion between mtnK and mtnA and not a translation initiation factor . Our blast searches confirmed the previously described fusion of mtnB and mtnD in Tetrahymena [5]; we will refer to this enzyme as mtnBD . To determine if this is a real fusion enzyme or just a gene prediction artifact , we searched the Tetrahymena ESTs in GenBank for evidence of transcripts of this gene . The searches did not return any Tetrahymena EST with significant similarity to the mtnBD fusion protein . Therefore , to confirm that mtnBD is a real fusion protein in Tetrahymena , we used a Tetrahymena cDNA library to determine the presence of a transcript of the predicted gene coding for the fusion protein . PCR results ( not shown ) indicated that the gene is transcribed as a single mRNA in Tetrahymena , and thus codes for an expressed fusion protein . We used the sequence of the mtnBD gene from Tetrahymena to search the NR database from GenBank in order to identify other organisms that might have this same gene fusion . Interestingly , this fusion gene is not present in any other organism in GenBank and is absent even in ciliate species closely related to Tetrahymena , like Paramecium tetraurelia . This species has a fully sequenced genome that completely lacks homologs to any enzyme in the methionine salvage pathway . Sequence similarities to mtnB and mtnD and the lack of an mtnC homolog suggest that the fusion protein , mtnBD , might be able to perform the functions of mtnC in addition to those of mtnB and mtnD in Tetrahymena . To determine if this hypothesis was correct , we investigated the in vivo function of the T . thermophila fused gene by doing complementation studies in yeast cells . We designed a synthetic gene that codes for the same amino acid sequence as the mtnBD fusion from Tetrahymena , but with codon usage optimized for expression in yeast cells . This gene will be referred to as SYN-MTNBD . We cloned SYN-MTNBD under the control of a GAL promoter in the pGREG505 plasmid ( Euroscarf ) . The plasmid ( pGREG505/SYN-MTNBD ) was transformed into three different S . cerevisiae met− strains ( met15Δ0 ) that had the genes that code for the enzymes mtnB , mtnC and mtnD of the methionine salvage pathway individually deleted . The goal was to determine if the SYN-MTNBD gene is capable of complementing each single knockout strain . As a negative control each strain was also transformed with the parent vector , pGREG505 . Although met− yeast strains usually cannot grow in media lacking methionine , they can grow if the media is supplemented with MTA at a concentration of at least 5 mM [9] , [10] . In such media , yeast cells are able to synthesize methionine from MTA using the methionine salvage pathway . The three deletion strains used in this experiment cannot grow in the absence of methionine even when the medium is supplemented with MTA since each strain lacks a gene essential to the functioning of the methionine salvage pathway and is unable to synthesize methionine from MTA ( Figure 1 ) . If our hypothesis was correct , and the mtnBD fusion replaces mtnB , mtnC and mtnD in Tetrahymena , a synthetic version of the T . thermophila mtnBD fusion protein would restore the methionine salvage pathway in the three yeast deletion strains . We predicted that , while the three deletion strains would fail to grow in media lacking methionine but containing MTA due to the deletions in the methionine salvage pathway , the three strains would grow in this media when the SYN-MTNBD gene is expressed . Figure 3 shows the results of the experiment . None of the strains grew on the negative control plate ( Figure 3A: −Met−MTA medium; 4 days ) ; and all strains grew on the positive control plate ( Figure 3B: +Met medium; 4days ) . Only the deletion strains transformed with the pGREG505/SYN-MTNBD plasmid grew on the experimental plate; the deletion strains transformed with the control parent vector ( pGREG505 ) did not grow ( Figure 3C: −Met+MTA; 4 days ) . As expected , the growth of the strains transformed with SYN-MTNBD was slower in the medium supplemented with MTA than in the +Met medium . Yeast strains with a similar genetic background grow slower in medium supplemented with MTA than in medium supplemented with methionine , even when they have an intact methionine salvage pathway [9] . This is probably a result of having to synthesize the methionine necessary for growth from MTA , instead of having it readily available in the medium . Our results show that a single fusion protein in the methionine salvage pathway of T . thermophila , mtnBD , accomplishes enzymatic functions that are generally performed by three enzymes , mtnB , mtnC and mtnD , in other organisms .
Interestingly , the fusion of mtnB and mtnD in Tetrahymena led to a gain of function of the fused protein , as evidenced by its ability to complement yeast mtnB , mtnC , or mtnD single knock-out strains . This demonstrates another potential advantage of fused proteins , the ability to short-circuit biochemical pathways . It is worth noting that mtnC can catalyze in one step a reaction that in other organisms is performed in two steps by mtnX and mtnW . In these organisms we could say that 1+1 = 4 , and that the fusion of two genes in Tetrahymena generates a tetrafunctional protein . Because mtnBD is able to catalyze the reaction of mtnC , this protein probably became superfluous in Tetrahymena and was eliminated during evolution . It is possible that in order for mtnBD to catalyze the enolase-phosphatase reaction of mtnC , another enolase-phosphatase , non-homologous to mtnC or to mtnW/mtnX , is needed . We believe this scenario to be highly unlikely , however , since this other enolase-phosphatase would have to be present in yeast and only functional when mtnBD is being expressed , as the absense of mtnC is lethal for yeast of this genetic background in −Met+MTA medium in the absense of mtnBD . Furthermore , in all eukaryotes in which we searched for the enzymes of the pathway using blast , either mtnC or mtnW/mtnX is present , with the exceptions of Tetrahymena and organisms that completely lack the pathway . Because Tetrahymena is the only organism in which the fusion gene is present and in which mtnC and mtnW/mtnX are missing , and because the fusion is able to complement the deletion of the enolase-phosphatase ( mtnC ) in a heterologous system , yeast , a much more likely explanation is that the Tetrahymena fusion gene somehow bypasses the enolase-phosphatase step or is able to catalyze it . It is important to stress that the fusion protein in Tetrahymena is composed of only the sequences of mtnB and mtnD , with a linker of about 10–15 amino acids in between . The fusion has no significant sequence similarity with either mtnC or mtnW/mtnX . It would be interesting to characterize how mtnBD allows the cell to bypass the enolase-phosphatase step in the pathway , or how it catalyzes this step , and to compare its mechanism to the mechanism of mtnC or mtnW/mtnX . MtnD is an interesting dioxigenase . In vivo , this enzyme catalyzes different reactions depending on the metal ion bound to the active site [16] , [17] . When bound to Ni2+ the enzyme catalyzes the formation of 3-Methylthiopropionate , formate and carbon monoxide , while when bound to Fe2+ it catalyzes the formation of 4-Methylthio-2-oxobutanoate and formate . Only when it is bound to Fe2+ is the enzyme involved in the methionine salvage pathway , as 4-Methylthio-2-oxobutanoate ( MOB ) is the substrate of mtnE in the last step of the pathway . The purpose of the off-pathway reaction is not known , and the product 3-Methylthiopropionate is cytotoxic [17] . It would be interesting to determine if the fusion enzyme mtnBD , besides being able to catalyze the three reactions in the methionine pathway , still retained the ability to bind Ni2+ and generate 3-Methylthiopropionate and carbon monoxide . It would also be interesting to determine if the mtnBD gene is more efficient than the combined three genes it replaces , mtnB , mtnC and mtnD , and whether it leads to more efficient methionine recycling in Tetrahymena .
The complete set of enzymes of the methionine salvage pathway has been described in various bacteria [3] , [4] and recently in S . cerevisiae [9] . We used the KEGG database ( Kyoto Encyclopedia of Genes and Genomes [18] ) to download the sequences of the methionine salvage pathway enzymes in Bacillus subtillis and yeast . In B . subtillis MTA is converted into methionine by mntN , mtnK , mtnA , mtnB , mtnW , mtnX , mtnD and a transaminase . In yeast the pathway involves mtnP , mtnA , mtnB , mtnC , mtnD and a transaminase . Because the last step of the pathway can be performed by a variety of transaminase enzymes , we excluded it from the analyses . Using each of the 7 B . subtillis enzymes and each of the 5 yeast enzymes as the queries , we performed blastp searches [19] of all the Tetrahymena protein sequences in RefSeq [20] . To eliminate the possibility that a homolog for a given protein exists in Tetrahymena but has not been annotated , we used tblastn to search the Tetrahymena genome sequence , using as queries each of the proteins that didn't return a significant hit in the blastp searches . None of these returned significant hits . In yeast , the genes coding for mtnB , mtnC and mtnD were identified by Pirkov et al . ( 2008 ) as , MRI1 ( YPR118W ) , MDE1 ( YJR024C ) and UTR4 ( YEL038W ) , respectively . The single null mutant strains used in this study , mri1Δ , mde1Δ and utr4Δ ( referred to in this study as mtnBΔ , mtnCΔ , and mtnDΔ , respectively ) and a WT strain , all in the BY4741 genetic background ( MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0 ) , were purchased from the single gene knockout collection from Open Biosystems . The synthetic gene ( SYN-MTNBD ) was synthesized and cloned into E . coli pUC57-simple expression vector by GeneScript Corporation . pUC57/SYN-MTNBD was digested with NotI and NheI , and the DNA fragment with the SYN-MTNBD gene was cloned into yeast expression vector pGREG505 ( purchased from EUROSCARF Frankfurt , Germany ) , cut with the same set of enzymes to create pGREG505/SYN-MTNBD . pGREG505/SYN-MTNBD has a LEU2 selectable marker and the SYN-MTNBD gene is under the control of the GAL promoter . pGREG505/SYN-MTNBD and pGREG505 were transformed into each of the three yeast deletion strains mtnBΔ , mtnCΔ and mtnDΔ . Transformants were selected by plating on synthetic media lacking Leucine ( −Leu plates ) . Leu+ transformants were picked and re-streaked to single colonies on −Leu plates . Liquid cultures were then inoculated and grown to saturation to generate frozen stock cultures of the transformed null mutant strains with pGREG505/SYN-MTNBD and pGREG505 . All Saccharomyces cerevisiae strains were maintained and grown in YPD ( yeast Nitrogen base , peptone , dextrose ) minimal media or the appropriate drop-out medium as specified in Pirkov et al . 2008 ( 1 . 7 g/L yeast nitrogen base without amino acids and ammonium sulfate , 0 . 5 g/L serine , 0 . 2 g/L aspartate , 0 . 17 g/L tryptophan , 0 . 12 g/L adenine , 0 . 1 g/L leucine , 0 . 02 g/L tyrosine , and 0 . 05 g/L of histidine , uracil , lysine , alanine , phenylalanine , tyrosine , isoleucine , valine and tyrosine ) with 2% glucose or galactose . Drop-out media were made by leaving out the corresponding amino acid or base . Since all strains used are met15Δ0 , they cannot grow in the absence of methionine ( Met ) . The media of the positive control plates used in the complementation studies were supplemented with 5 mM Met , and the experimental plates were supplemented with 5 mM MTA ( Sigma-Aldrich , St . Louis , MO ) . When expression of the cloned gene was necessary , glucose was replaced with galactose . Strains carrying the pGREG505 plasmids ( LEU2 marker ) were grown in −Leu drop-out media to prevent loss of the plasmids . The three yeast deletion strains mtnBΔ , mtnCΔ and mtnDΔ transformed with pGREG505/SYN-MTNBD or pGREG505 were streaked to single colonies on −Leu +Met agar plates to select for retention of the plasmids . Liquid cultures in 1 ml −Leu +Met were started by inoculating with a single colony and grown overnight in a roller drum at 30°C . Cells from the overnight cultures where spun down , washed , resuspended in −Leu −Met medium and put back for another overnight in the roller drum at 30°C . This step was performed to exhaust extracellular and intracellular pools of methionine [9] . After an overnight in −Leu −Met medium the number of cells/ml in each liquid culture was determined by counting with a hemacytometer . 50 µl serial dilution samples of each strain with 1×108 cells/ml , 1×107 cells/ml , 1×106 cells/ml , 1×105 cells/ml , 1×104 cells/ml , 1×103 cells/ml , 1×102 cells/ml and 1×10 cells/ml were transferred onto a 96-well plate . 3 µl samples were then pinned onto agar plates in triplicates using a 96-well floating Pin Replicator ( V&P Scientific , Inc . , San Diego , CA ) . Agar plates used: +Gal −Leu −Met in the negative controls , +Gal −Leu +Met in the positive controls , and +Gal −Leu −Met +MTA in the experimental plates . Pinned cells were grown at 30°C for four days and then visually analyzed and photographed with a Kodak IC440 imaging system ( Figure 2 ) . Because the strains are met15Δ0 , they cannot grow in the absence of methionine . Pirkov et al . [9] showed , however , that if supplemented with enough MTA these strains are able to grow in the absence of methionine by converting the MTA into methionine by means of the methionine salvage pathway . Each of our deletion strains has one of the genes in the pathway deleted and cannot convert MTA to methionine , and thus should not be able to grow in the presence of MTA , unless the SYN-MTNBD gene complements the gene deleted in the strain . All strains should grow in media supplemented with methionine ( positive control ) , and none of the strains should grow in −Met −MTA media ( negative control ) .
|
Fusion genes , composed of the complete sequence of two or more other genes , are excellent markers of evolution . In addition , fused genes are usually composed of genes with related functions , which makes them useful in inferring function when the function of one of their components is known . We detected a fusion gene in the eukaryotic organism Tetrahymena thermophila that , although composed of only two genes , seems to perform the function of three genes in this organism . To show that this is the case , we expressed the Tetrahymena fused gene in three different yeast strains , each lacking one of these three genes . The Tetrahymena gene was able to rescue the phenotype of all yeast strains , proving that it can perform the functions of the three genes in yeast . Our results highlight another important biochemical characteristic of fusion genes: they can serve as biological shortcuts , allowing a single fusion of two enzymes to functionally replace three independent enzymes .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"computational",
"biology/comparative",
"sequence",
"analysis",
"genetics",
"and",
"genomics/bioinformatics",
"genetics",
"and",
"genomics/microbial",
"evolution",
"and",
"genomics",
"genetics",
"and",
"genomics/gene",
"expression"
] |
2009
|
1+1 = 3: A Fusion of 2 Enzymes in the Methionine Salvage Pathway of Tetrahymena thermophila Creates a Trifunctional Enzyme That Catalyzes 3 Steps in the Pathway
|
The folding pathway and rate coefficients of the folding of a knotted protein are calculated for a potential energy function with minimal energetic frustration . A kinetic transition network is constructed using the discrete path sampling approach , and the resulting potential energy surface is visualized by constructing disconnectivity graphs . Owing to topological constraints , the low-lying portion of the landscape consists of three distinct regions , corresponding to the native knotted state and to configurations where either the N or C terminus is not yet folded into the knot . The fastest folding pathways from denatured states exhibit early formation of the N terminus portion of the knot and a rate-determining step where the C terminus is incorporated . The low-lying minima with the N terminus knotted and the C terminus free therefore constitute an off-pathway intermediate for this model . The insertion of both the N and C termini into the knot occurs late in the folding process , creating large energy barriers that are the rate limiting steps in the folding process . When compared to other protein folding proteins of a similar length , this system folds over six orders of magnitude more slowly .
The wide range of kinetics that characterizes protein folding has attracted interest from both experimentalists and theoreticians for decades [1] . Proteins fold on time scales that vary from microseconds to minutes [2] , even though the corresponding energy landscape directs folding towards its native state . This wide range of rates can be explained by the diverse size and shapes of the free energy barriers between the unfolded and folded ensembles , which are largely determined by the pattern of contacts , often called the protein topology . Some of the slowest folding proteins , such as the green fluorescent protein , are both long and have complicated contact patterns [3] . In the past 15 years , a set of proteins has been discovered that fold slowly and have knotted topologies [4] . Topological constraints can lead to large energy barriers that are difficult to characterize , so we employ a reduced description of the protein and explore the landscape using geometry optimization techniques . Topological problems have been investigated previously using G models [5] , and have shown that local unfolding may be required in some cases to organize the sequence of the folding of structural units . This type of potential energy landscape analysis with transition state theory has been used for describing kinetic phenomena in systems as diverse as molecular clusters , glasses , and proteins [6] . An advantage of this approach is that the topological constraint limits the utility of a simple energy based reaction coordinates ( like , see Methods section ) , which ordinarily work well to describe folding . We studied a tRNA methyltransferase ( PDB code 1UAM ) , which contains a deep trefoil knot in the C terminus domain [7] . The goal of this study was to estimate the fastest speed possible for folding a small knotted protein , and we therefore truncated the system to residues 78–135 to limit the number of atoms not included in the knot . Mathematically knots are defined in closed loops . In proteins links are used to connect the termini , and the structure is topologically classified by the determination of its Alexander polynomial [8] , [9] . Recently knotted proteins have been identified [10] and their kinetics explored . Knotted systems with a larger number of residues beyond the knot will exhibit slower kinetics , because of the need to break a larger number of contacts to fold properly . Protein models based on random contacts produce knotted proteins with greater frequency than is seen in protein structure databases [11] , [12] . Knotted proteins are likely avoided during evolution , while some have remained and are an evolutionary curiosity . Knotting can also occur in other biopolymers such as DNA , and these systems exhibit significantly slower kinetics [13] .
A connected path of minima and transition states between an unfolded structure and the native state was created with the discrete path sampling method ( DPS ) combined with the associative memory Hamiltonian [14] , [15] . After obtaining an initial connection , this database of structures was expanded using schemes to systematically reduce the length and barriers associated with the largest contribution to the rate coefficient [16]–[19] . Here , the SS superscript refers to an approximate formulation where the steady-state condition is applied to minima outside the product and reactant sets . This formulation provides a convenient framework for analysis because can be written as a sum over discrete paths [16] , [20] . Once this pathway appeared to have converged , the database was further refined by connecting undersampled minima with a large ratio of the free energy barrier to free energy difference from the global minimum [21] , [22] . This choice is motivated by the idea of optimizing folding kinetics for topologically constrained system in a similar way to the minimization of energetic frustration obtained by comparing the folding temperature to the glass transition temperature [23] , [24] . The resulting database contained 212054 minima and 206923 transition states , and the corresponding disconnectivity graph [25] , [26] is shown in Figure 1 , as rendered by VMD [27] . Here we remind the reader that every vertical line in the graph terminates at the energy of a local minimum , and that the minima are progressively connected together as the threshold energy , , increases , according to the lowest barrier between them . The graph exhibits three distinct color coded features corresponding to potential energy basins , with properly knotted minima occupying the lowest-lying states in the center of the figure . Branches corresponding to minima with the knotted topology are colored blue , while those with the C and N termini still free are colored green and red , respectively . The kinetic coefficients for interconverting minima within these basins are relatively fast , so that local equilibrium is achieved on the time scale of the slow kinetics determined by the barriers between the different basins . This figure represents an unusual folding energy landscape , where large energy barriers occur despite the lack of favorable non-native interactions in the Hamiltonian . The two higher energy sets of structures correspond to local minima where either the C or N terminus has the native topology , but the other terminus is still unknotted , and we will refer to these as N-free and C-free geometries , respectively ( Figures 2 and 3 ) . A useful descriptor of these ensembles is their structural overlap . The value ( a measure of structural similarity , see Methods section ) between the N-free and native minima is , between the C-free and native minima is , and between N-free and C-free minima is . The small variation in shows that only a few contacts are different , where a contact is restricted to be less than 9 Å in order to distinguish structures that have a high degree of similarity . Most of the contacts in each basin are identical , except for a few important differences near the termini . In the N-free minima these interactions are between residues 7–8 and 45–46 as shown in Figure 4 , while in the C-free minima they are between 48–52 and 30–34 as shown in Figure 5 , and define the interactions that prevent unphysical chain crossings . These non-native contacts are energetically neutral with respect to the interaction Hamiltonian ( see Methods section ) due to the native-only form of the Hamiltonian , but they affect the calculated pathways through an excluded volume repulsion due to the backbone interaction . A kinetic analysis of the DPS database using transition state theory requires choices for the value of and the mass associated with each site in the AMH potential . For simplicity a value of 12 atomic mass units ( amu ) was assigned to each site . To assign a value for , we compared the normal mode frequencies for the AMH potential with typical values associated with heavy atom motion in all-atom representations of proteins . This comparison suggests that ε should be around one kcal/mol . The discrete path that makes the largest contribution to the phenomenological two-state rate coefficient , which we use to define the overall reaction mechanism , exhibits the same qualitative features over a wide range of values for . The estimated rate coefficients themselves are more sensitive , as discussed below . If we take kcal/mol and the values of length and mass in the AMH potential as 1 Å and 12 amu then the reduced value of at room temperature is approximately 0 . 59 . The pre-exponential factor for each minimum-to-minimum rate coefficient scales as , while the reduced value for room temperature decreases linearly , lowering the corresponding Boltzmann factors exponentially . The choice of the reactant and product states can have a significant effect on the calculated rates . One way of selecting the states is to calculate an order parameter for all the local minima , and simply assign reactant and product states on this basis . However , an alternative method is possible using the known characteristics of the kinetic transition network and a self-consistency condition . Here we take the two endpoints that were used to calculate an initial path , and assign these minima to reactant and product sets . We then regroup the database using a recursive scheme [21] , combining free energy minima that can interconvert without encountering a barrier higher than a chosen threshold value , . This approach is attractive , because we require a separation of time scales for equilibration in the product and reactant regions , compared to the folding transition time , in order to recover a two-state description of the kinetics [20] , [28] , [29] . In this case , we expect to see a range of values for that give a similar value for the calculated rate coefficient . Rate coefficients were calculated for three different choices of the reactants , namely a fully unfolded minimum and low-lying minima from the C-free and N-free regions of the landscape . In each case a low-lying minimum from the set of knotted configurations was chosen as the product , and rate coefficients were calculated for a range of and values . Following the recursive regrouping of states according to the value of , mean first-passage times were calculated from each minimum in the reactant set using a graph transformation procedure [17] , [30] . A phenomenological two-state rate coefficient is then obtained using appropriate occupation probabilities for the starting minimum [16] , [17] , [20] , [30] . These values of tested ( 1 . 0 , 0 . 9 and 0 . 7 kcal/mol ) are close to the magnitude suggested by examining the normal mode frequencies . For kcal/mol the rate coefficient varies from 0 . 04 to 0 . 4 s for kcal/mol using a fully unfolded or C-free minimum as the reactant . Therefore the folding time is between 25 seconds and 2 . 5 seconds . A movie of the C-free minimum to the folded state is shown in Video S1 . With a low-lying N-free minimum as the reactant , the calculated rate coefficient is 0 . 02 s for kcal/mol , jumping to about 50 s for kcal/mol . So the folding time becomes 50 seconds to 0 . 02 seconds . A movie of the N-free minimum to the folded state is shown in Video S2 . The values obtained rise by about two orders of magnitude starting from N-free as reactant if we set kcal/mol . For each choice of reactant , the discrete path making the largest contribution to the rate coefficient [16] was extracted , and snapshots of the intervening structures are superimposed upon the energy as a function of integrated path length in Figures 6 , 7 , and 8 . Here the path length is defined from the Euclidean distance between successive configurations in the folding reaction . These pathways are based on the rate coefficients and associated free energy barriers calculated at , and the entropy terms that derive from alternative discrete paths through the stationary point database are all included in the estimates of the overall rate coefficients . This kinetic analysis suggests that the local maxima in the energy profiles shown in Figures 6 , 7 , and 8 generally correspond to kinetic bottlenecks . Starting from an unfolded state , we see that the N terminus first forms a loop that threads through the middle of the protein , and then opens ( Figure 6 ) . The structures involved in this process appear very similar to the corresponding event in the pathway starting from an N-free minimum in Figure 8 , aside from the state of the C terminus . The final folding events illustrated in Figure 6 are very similar to those shown in Figure 7 , with a bend forming at the C terminus , threading through the center of the structure , and straightening . The calculated rate coefficients are also very similar when the reactant is chosen as either the fully unfolded state or a C-free minimum , indicating that knotting of the C terminus is the rate-determining step for this model system . The region of configuration space corresponding to low-lying N-free minima is then interpreted as a kinetic trap , which would probably result in a distinct relaxation time scale .
In this study a truncated sequence from a tRNA methyltransferase was considered with a G model containing only the favorable interactions that are present in the global minimum . In contrast to previous minimally frustrated models , which exhibit only a single potential energy funnel [31]–[33] , the landscape for the knotted protein is divided into three distinct regions , corresponding to the correctly folded native state and to structures where either the C or N terminus are not knotted . The potential energy barriers between the lowest minima in these regions are relatively large , with values of order 15 to 20 in units of the associated memory Hamiltonian [34] , [35] parameter , for which we estimate a value of around one kcal/mol . The calculated rates for folding are therefore rather slow , in agreement with previous simulations of knotted proteins [36] , [37] . The folding reaction is hindered by the complex topology of this protein . Modeling these interactions and mechanisms in a realistic way requires new tools that prevent unphysical chain crossing events from occurring during the interpolation between structures that have an intervening chain . Details of the procedure are given in the Methods section , and an overview is provided here . Our initial aim was to avoid chain crossings by changing parameters of the potential . However , tightening the bond length constraints for covalent bonds does not solve the problem , because the interpolated images simply avoid the chain-crossing region . In the doubly-nudged [38] elastic band [39]–[41] ( DNEB ) method for identifying useful starting geometries for transition state refinement , a set of images are connected by harmonic springs , and the images can be equally spaced by increasing the corresponding force constant . However , this increase forces the chains into high energy structures that bracket an unphysical crossing . To avoid this situation it is necessary to construct a non-linear interpolation between the end points , and two strategies were implemented . To accelerate the energy evaluation , an elastic network potential was defined based on the two end points , with harmonic restraining potentials for atoms whose separation does not change . This geometrical analysis was also used to diagnose chain crossings for a linear interpolation between the end points , and to distinguish the chain that is moving from one that constitutes a barrier . When crossings were identified the potential was modified to shrink the end of the moving chain and add repulsive interactions to keep it away from the other chain . The DNEB images were then refined following standard procedures and the modified potential was morphed back into the AMH potential slowly enough for sites on one chain to move around the other chain , rather than through it . Overall , this procedure allows paths to be obtained that circumvent chain crossing , while retaining flexibility and providing a solution that is free of constraints . Both the C and N termini must effectively cross over a chain belonging to the central region of the protein in order to achieve the knotted topology . In the present model it is the C terminus crossing that appears to be the rate-determining step . The rate coefficients reported here are order of magnitude estimates , and correspond to a slow folding process , as expected from previous simulations [36] , [37] and experiments [42] , [43] . The precise energetics of this truncated model may differ from those of the full protein , but we expect the key steps in the folding and knotting pathways to be retained . Making a meaningful estimate of the scaling behavior with respect to chain length will be addressed in future work . For both the C and N termini the chain cross-over are achieved by formation of a loop , which then inserts through the center of the protein and straightens . While the chain is in the loop conformation the folding process could notionally be reversed by pulling the end of the chain , which is one definition of a slipknot topology , consistent with previous simulations [37] . The region of configuration space corresponding to N-free local minima , where the C terminus crossing occurs first , is therefore likely to give rise to a separate relaxation time scale . The folding pathways exhibit some interesting mechanistic features , which might be transferable to related systems of knotted proteins and polymers . In particular , both the C and N termini crossings are achieved by formation of a loop that threads through the main body of the protein . The order of the knotting of the chain and the folding of the protein may change as the length of the system changes and as the energy function becomes more realistic . Adding non-native interactions would likely lower the free energy barrier of folding [44] , but could also stabilize non-native structures and slow local refolding of the loops involved in chain crossings . Introducing non-additive cooperative contacts would increase the energy barriers and likely slow the kinetics [45] . Protein engineering studies of YibK suggest that knotting and formation of native structure are independent events that occur in sequence [46] . These experiments also suggest an early knotting event and slow development of native structure in the knotted region . Similar behavior is seen in DNA , where local unfolding speeds up diffusion of the knot along the polymer chain [47] . To provide meaningful comparisons with these observations will require simulations of longer systems , rather than the truncated sequence considered in the present work . When compared to other protein folding proteins of a similar length , which fold on the microsecond time scale [2] , this system folds over six orders of magnitude more slowly .
In order to describe an energy landscape with an exponential number of states , we reduce the atomistic detail of the system and discretize the energy landscape into minima and transition states . The associative memory Hamiltonian ( AMH ) protein model [34] , [35] , is a coarse-grained molecular mechanics potential inspired by the physics of protein folding . The energy functions consist of a polypeptide backbone term , , with a molecular interaction term , [48]–[53] . The number of atoms per residue is limited to three ( C , C , and O ) , except for glycine . The interaction parameter , which is the unit of energy , is defined by the native state energy excluding backbone contributions , , via ( 1 ) where is the number of residues . All temperatures are quoted in reduced units as . While creates self-avoiding peptide-like stereochemistry , introduces the majority of the attractive interactions that produce folding . Using the interactions described by , we define a pairwise additive G model [54] , [55] , which is biased toward the native basin . Such models have been shown to reproduce many features of the mechanism and kinetics of protein folding [56] , [57] . The interactions between residues were defined by , ( 2 ) where the distances in the Gaussian term are determined by the native state . The interactions are defined in this minimal model for residues with greater than three residues sequence separation between the atom pairs . The weights , , corresponding to the depths of the Gaussian wells , are set to ( 0 . 177 , 0 . 048 , 0 . 430 ) in order to approximately divide the interaction energy equally between the different distance classes , as suggested by previous theoretical models [58] . The width of the Gaussian , , is determined by the sequence separation as Å . The scaling factor is used to satisfy Eq . ( 1 ) . We measure the quality of the structures encountered with an order parameter , , which measures the sequence dependent structural similarity of two configurations . is calculated from Eq . ( 3 ) as a summation of pairwise differences between distances in a target and a reference structure ( usually the native state ) , normalized by the number of contacts , where is sequence length: ( 3 ) The resulting order parameter , , ranges from zero , when there is no similarity between structures at a pair level , to unity , which indicates an exact overlap . We made several changes to the original AMH backbone potential , , in the present work . Eliminating some compromises necessary for rapid molecular dynamics simulations allows the AMH potential to be used with geometry optimization methods to produce tightly converged stationary points . This tight convergence is necessary for the construction of a kinetic transition network [20] , [22] , [59]–[61] . The terms shown in Eq . ( 4 ) are used to reproduce the peptide-like conformations in the original molecular dynamics energy function: ( 4 ) For all calculations , we replaced the SHAKE method for bond constraints with a harmonic potential , , between the C-C , C-C , and C-O atoms . This replacement permits the location of local minima without requiring an internal coordinate transformation , and avoids discontinuous gradients [62] . The neighboring residues in sequence sterically limit the positions the backbone atoms can occupy , and this effect is reproduced with a Ramachandran potential , . The planarity of the trans peptide bond is ensured by another harmonic potential , . The chirality of the C centers is maintained using the scalar triple product between neighboring C , C , and N atoms , . Excluded volume repulsion between the backbone atoms is achieved with via a smooth step ( hyperbolic tangent ) function , , in order to have a continuous potential , and differs from the previous hard sphere potential in the AMH . For this Hamiltonian , we employed the discrete path sampling ( DPS ) approach to create databases of local minima and their intervening transition states , starting from two end points . To identify suitable endpoints , we used basin-hopping global optimization [63] , [64] to search for the global minimum of the energy landscape , and to create an unfolded conformation . We have previously shown how basin-hopping can be successfully combined with associative memory Hamiltonians for identifying low energy states , and high quality structures [62] . The discrete path sampling approach is a coarse-grained analogue of the transition path sampling method [29] , [65] , [66] , where geometry optimization tools are employed to refine a kinetic transition network . The network consists of local minima and transition states of the energy potential , where a transition state is defined as a stationary point with a single negative Hessian eigenvalue [67] . The connectivity is defined by approximate steepest-descent paths obtained by energy minimization following infinitesimal displacements parallel and anti-parallel to the eigenvector corresponding to the unique negative eigenvalue . A discrete path then refers to a series of minimum-to-minimum connections together with the intervening transition states . The original DPS formulation has been presented in detail elsewhere [16] , [20] , as have more recent developments [18] , [21] . The aim is to enlarge a database of connected stationary points starting from those in the initial path , by adding all the minima and transition states found during successive connection-making attempts for pairs of minima selected from the current database . The main challenge of DPS calculations is the characterisation of transition states . In contrast , energy minimization and identifying approximate steepest-descent pathways is straightforward; here we used the limited-memory Broyden–Fletcher–Goldfarb–Shanno ( LBFGS ) algorithm of Liu and Nocedal [68] , [69] . The transition state searches are connection attempts for a given pair of local minima . A doubly-nudged [38] elastic band [39]–[41] ( DNEB ) refinement of interpolated images was first run for each connection attempt , and the images corresponding to local energy maxima were then tightly converged using hybrid eigenvector-following [70] . The missing-connection algorithm [71] was employed to choose subsequent pairs of minima for further connection attempts [15] . To avoid unphysical chain crossed transition states , we made two changes in methodology to generate physical interpolations for finding potential transition state structures . We define two new potentials , which maintains chain connectivity , and , which introduces atomic repulsion . The potential is modified in three stages during the DNEB refinement . During the first third of the DNEB steps and are used with modified distances . In the second third the distances are relaxed to physically meaningful values , and in the final third we switch to the full AMH potential . We also define a simpler potential function , often referred to as an elastic network model [72] to represent the system during some of the DNEB refinement . The two end points for the DNEB calculation are analyzed to identify pairs of atoms within a cutoff distance ( 10 Å ) that are found at the same separation within a given tolerance in both structures . If and are the distances between atoms and in the starting and finishing geometries , then we introduced a harmonic restraining potential for this pair if , where . For such pairs the restraining potential was then ( 5 ) where is the distance between the atoms involved in restraint , and was initially set equal to . The parameter was set to in the present calculations , where the DNEB spring constant was set to . has the appearance of an elastic network model [72] , which reflects the conserved interatomic distances in the two endpoints . Analyzing the conserved distances is also useful for diagnosing when crossings occur , so that corresponding changes can be made to the potential , as described below . The initial images in the DNEB interpolation were simply placed at regular intervals for a linear interpolation between the specified endpoints , after putting these two structures into optimal alignment [73] . All pairs of atoms corresponding to different restraints with no common atoms were then examined for all pairs of DNEB images . The crossing check was applied for the largest untested image separation of every remaining image . Only pairs of restraints where the separation of the midpoints between the restrained atoms in both images were below a cutoff value of 10 Å were considered . The midpoint separation in one of the two images was also required to change by at least 3 Å from the value in the nearest endpoint structure . For restrained pairs satisfying these criteria a crossing is diagnosed when the dot product between the vectors joining the midpoints between the constraint pairs in the two images is negative . Outer and inner atom pairs are then defined according to how far the midpoints move between the two images: the midpoint that moves the furthest is assumed to belong to the outer chain , which needs to move around the inner chain . To avoid unphysical crossings in the interpolation , we modify the potential and add repulsive terms through . If atomic contacts within the set of pairs are found to cross , using the above geometrical condition , then repulsive terms are added according to the four distances between the two pairs of atoms . For crossings of restrained distances , the repulsive contribution to the potential is ( 6 ) where is a step function , ( 10 Å ) is a cut-off for the repulsive terms , ( ) defines the magnitude of the repulsion , and is one of the four distances between pairs of atoms whose restrained contacts are found to cross . To enable chains to pass around one another when crossings are diagnosed , further changes were made to . For the restrained contact in the outer and inner chains was changed to and , respectively , for each crossing . Hence the outer chain shrinks while the inner chain expands . The first third of the DNEB iterations were run with the modified potential plus . For the middle third of the DNEB optimization the restraint distances were switched back to the value according to the schedule , with . The full AMH potential was then used for the last third of the DNEB iterations . To describe the global kinetics of the transition network , we calculated the rate coefficients associated with each transition state using transition state theory [74] ( TST ) with vibrational densities of states obtained from harmonic normal mode analysis . The most important features of the mechanism of folding to the knotted state are relatively insensitive to the values assigned to minimum-to-minimum rate coefficients , while the total rate coefficients that we report are order of magnitude estimates .
|
Proteins are chains , which must fold into a compact structure for the molecule to perform its biological function . There are a large number of ways the molecule can move into this final shape . Proteins have evolved sequences that perform this difficult task by having strong biases toward the final shape , while not getting stuck in different structures along the way . One way proteins can be trapped is by forming a knot in the chain . For the most part , proteins are remarkable in avoiding knotting . However , in order to function a few proteins form knots . We show how a model protein is able to knot itself , and estimate how fast this process occurs . Our goal is to treat a small and uncomplicated protein to estimate the fastest rate possible for the folding of a knotted protein . This rate is interesting when compared to the speed of folding of other proteins . We have visualized how the molecule changes shape to its functional position , and examined other paths the molecule may take .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"computational",
"biology/molecular",
"dynamics",
"biophysics/theory",
"and",
"simulation",
"biophysics/protein",
"folding"
] |
2010
|
The Energy Landscape, Folding Pathways and the Kinetics of a Knotted Protein
|
The clinical presentation of M . ulcerans disease and the safety and effectiveness of treatment may differ in elderly compared with younger populations related to relative immune defficiencies , co-morbidities and drug interactions . However , elderly populations with M . ulcerans disease have not been comprehensively studied . A retrospective analysis was performed on an observational cohort of all confirmed M . ulcerans cases managed at Barwon Health from 1/1/1998-31/12/2014 . The cohort included 327 patients; 131 ( 40 . 0% ) ≥65 years and 196 ( 60 . 0% ) <65 years of age . Patients ≥65 years had a shorter median duration of symptoms prior to diagnosis ( p<0 . 01 ) , a higher proportion with diabetes ( p<0 . 001 ) and immune suppression ( p<0 . 001 ) , and were more likely to have lesions that were multiple ( OR 4 . 67 , 95% CI 1 . 78–12 . 31 , p<0 . 001 ) and WHO category 3 ( OR 4 . 59 , 95% CI 1 . 98–10 . 59 , p<0 . 001 ) . Antibiotic complications occurred in 69 ( 24 . 3% ) treatment episodes at an increased incidence in those aged ≥65 years ( OR 5 . 29 , 95% CI 2 . 81–9 . 98 , p<0 . 001 ) . There were 4 ( 1 . 2% ) deaths , with significantly more in the age-group ≥65 years ( 4 compared with 0 deaths , p = 0 . 01 ) . The overall treatment success rate was 92 . 2% . For the age-group ≥65 years there was a reduced rate of treatment success overall ( OR 0 . 34 , 95% CI 0 . 14–0 . 80 , p = <0 . 01 ) and when surgery was used alone ( OR 0 . 21 , 95% CI 0 . 06–0 . 76 , p<0 . 01 ) . Patients ≥65 years were more likely to have a paradoxical reaction ( OR 2 . 06 , 95% CI 1 . 17–3 . 62 , p = 0 . 01 ) . Elderly patients comprise a significant proportion of M . ulcerans disease patients in Australian populations and present with more severe and advanced disease forms . Currently recommended treatments are associated with increased toxicity and reduced effectiveness in elderly populations . Increased efforts are required to diagnose M . ulcerans earlier in elderly populations , and research is urgently required to develop more effective and less toxic treatments for this age-group .
Mycobacterium ulcerans ( M . ulcerans ) is an infection that causes necrotizing lesions of skin and subcutaneous tissue . The majority of cases are reported from west and central Africa , but unlike Africa where the disease occurs mainly in children[1 , 2] , in south-eastern Victoria , Australia it occurs mainly in adults with a large proportion aged > 50 years . [3] Reported rates of disease in Australian populations are up to 7 times higher in those ≥55 years of age[4] . Current M . ulcerans treatment guidelines recommend combined antibiotics for 8 weeks with surgery as an adjunctive treatment[5 , 6] . The clinical presentation of M . ulcerans disease ( Buruli ulcer ) , as well as the safety and effectiveness of treatment , may differ in elderly compared with younger populations . It is known that immune function reduces with senescence , and as the immune system plays a vital role in the control of M . ulcerans[7 , 8] , this may lead to an increase in the incidence and severity of disease as well as reduced effectiveness of treatment . There may also be altered health-seeking behaviours in older people who may find accessing healthcare more difficult or neglect skin lesions , or for whom there is a potentially increased prevalence of alternative causes of ulceration ( eg venous disease ) resulting in misdiagnosis . These aforementioned issues could lead to delays in diagnosis with increased disease severity . Furthermore , increased rates of co-morbidities in elderly patients may adversely affect immune function , but may also lead to increased drug interactions and the potential for increased toxicity associated with antibiotic treatment[9] . In our practice we have observed significant numbers of elderly patients developing M . ulcerans disease . Our earlier published experience has suggested that populations older than 60 years of age may have had increased prevalence of multiple M . ulcerans lesions at presentation[3] , reduced rates of treatment success with surgical treatment[10] , and increased rates of antibiotic related paradoxical reactions[11] . However , populations aged ≥65 years with M . ulcerans have not been comprehensively studied . We therefore undertook to describe in an Australian cohort the proportion of patients aged ≥65 years affected by M . ulcerans and compare them with younger patients with respect to their clinical presentation , and the safety and effectiveness of treatment .
Data was collected prospectively using Epi-info 6 ( CDC , Atlanta ) and analysed retrospectively using STATA 12 ( StataCorp , Texas , USA ) . Outcome data were censored at the time of death , disease recurrence or after 12 months of follow-up from initiation of antibiotics . Categorical variables were compared using 2x2 tables and the Chi-squared test . Medians of non-parametric variables were compared using the Wilcoxon rank sum test . Odds ratios were calculated using the Mantel-Haenszel test . This study was approved by the Barwon Health Human Research and Ethics Committee . All previously gathered human medical data were analysed in a de-identified fashion .
There were some significant differences in baseline characteristics between the age-groups . ( Table 1 ) Patients in the age-group ≥65 years were less likely to be male ( OR 0 . 60 , 95% CI 0 . 38–0 . 94 , p = 0 . 02 ) , the median duration of symptoms prior to diagnosis was significantly shorter ( 35 compared to 42 days , p<0 . 01 ) , and there was a higher proportion of patients with diabetes ( p<0 . 001 ) and immune suppression ( p<0 . 001 ) . Patients in the age-group ≥65 years were more likely to have lesions that were multiple ( OR 3 . 49 , 95% CI 1 . 26–9 . 54 , p<0 . 01 ) and classified as WHO category 3 compared with category 1 and 2 combined ( OR 4 . 89 , 95% CI 1 . 95–12 . 25 , p<0 . 001 ) . They also had a higher proportion of oedematous compared to non-oedematous lesions ( 11 . 5% compared with 6 . 2% , p = 0 . 09 ) . ( Table 1 ) Two hundred and eighty ( 85 . 6% ) patients received antibiotic treatment for a median of 56 days ( IQR 49–83 days ) . 115 ( 87 . 8% ) of those ≥ 65 years received antibiotics and 165 ( 84 . 2% ) of those < 65 years received antibiotics ( p = 0 . 97 ) . There were 284 antibiotic treatment episodes in 280 patients ( 4 patients received a second antibiotic course—three due to disease recurrence , and 1 for a late paradoxical reaction ) . Initial antibiotic combinations used were rifampicin/ciprofloxacin in 162 ( 57 . 0% ) , rifampicin/clarithromycin in 95 ( 33 . 4% ) , rifampicin/moxifloxacin in 8 ( 2 . 8% ) , rifampicin/clarithromycin/ethambutol in 6 ( 2 . 1% ) , clarithromycin/ciprofloxacin in 3 ( 1 . 1% ) and other varied combinations in 10 ( 3 . 5% ) treatment episodes . Overall 69 ( 24 . 3% ) antibiotic treatment episodes were associated with a complication severe enough to require cessation of at least one antibiotic . There was an increased incidence of antibiotic complications in those aged ≥ 65 years compared with those aged <65 years ( OR 5 . 29 , 95% CI 2 . 81–9 . 98 , p<0 . 001 ) . Including antibiotics commenced as second-line treatment following cessation of one or more of the initial antibiotics due to complications , 276 ( 97 . 5% ) treatment episodes included rifampicin , 174 ( 61 . 5% ) ciprofloxacin , 127 ( 44 . 9% ) clarithromycin , 13 ( 4 . 6% ) ethambutol , 10 ( 3 . 5% ) moxifloxacin and 9 ( 3 . 2% ) amikacin . Rifampicin was associated with complications in 47 ( 17 . 0% ) treatment episodes in which it was used , and this was more common in those aged ≥ 65 years compared to < 65 years ( OR 4 . 87 , 95% CI 2 . 36–10 . 07 , p<0 . 001 ) . ( Table 2 , Fig 1 ) Ciprofloxacin was associated with complications in 32 ( 18 . 4% ) treatment episodes in which it was used , and this was more common in populations ≥ 65 years ( OR 2 . 92 , 95% CI 1 . 26–6 . 75 , p<0 . 01 ) . Clarithromycin was associated with complications in 24 ( 18 . 9% ) treatment episodes in which it was used , and this was increased in populations ≥ 65 years ( OR 3 . 38 , 95% CI 1 . 30–8 . 78 , p<0 . 01 ) . ( Table 2 , Fig 1 ) The specific complications associated with each antibiotic are listed in Table 3 . In 11 patients hospitalization was required to manage the antibiotic complication; 9/49 ( 18% ) in those ≥65 and 2/20 ( 10% ) in those <65 years ( OR 2 . 03 , 95% CI 0 . 39–10 . 55 , p = 0 . 39 ) . 210 ( 64 . 2% ) patients had surgery; More patients in the ≥ 65 years age-group had surgery compared to the < 65 years age-group [92 ( 70 . 2% ) compared to 118 ( 60 . 2% ) , ( p = 0 . 06 ) ] . 62 ( 29 . 5% ) had surgery alone and 148 ( 70 . 5% ) had surgery plus antibiotics . There was no difference in the proportions who had surgery alone between the age-groups ( p = 0 . 96 ) . Treatment outcomes for first M . ulcerans lesions could be determined for 323 ( 98 . 8% ) patients; 2 were lost to follow-up , 1 was transferred out , and 1 had an unclear outcome . At the time of submission , 300 ( 92 . 9% ) had their outcomes determined after 12 months of follow-up and 23 ( 7 . 1% ) patients after 9 months of follow-up . There were 4 ( 1 . 2% ) deaths ( Table 4 ) . The median age of those who died was 91 . 5 years ( IQR 71 . 5–94 . 5 years ) , with significantly more deaths in the age-group ≥ 65 years compared to the age-group < 65 years [4 ( 3 . 1% ) compared to 0 ( 0 . 0% ) , p = 0 . 01] . Only one of the deaths ( #2 ) was felt to be directly attributable to M . ulcerans infection as a result of skin sepsis and secondary decompensated cardiac failure . For the remaining 319 patients , the overall treatment success rate was 92 . 2% ( Table 5 ) ; there was a reduced rate of treatment success in the age-group ≥ 65 years compared to the age-group <65 years ( OR 0 . 34 , 95% CI 0 . 14–0 . 80 , p<0 . 01 ) . There was a significantly reduced treatment success rate for the age-group ≥ 65 years when surgery was used alone ( OR 0 . 21 , 95% CI 0 . 06–0 . 76 , p<0 . 01 ) . Rates of treatment success for surgery plus antibiotics or antibiotics alone were similar between the age-groups . ( Table 5 ) 67/275 ( 24 . 4% ) patients experienced antibiotic-associated paradoxical reactions; 36 ( 32 . 4% ) patients ≥ 65 years and 31 ( 18 . 9% ) patients < 65 years . Patients ≥ 65 years were significantly more likely to have a paradoxical reaction compared with those < 65 years ( OR 2 . 06 , 95% CI 1 . 17–3 . 62 , p = 0 . 01 ) . This was independent of the WHO category of the lesion ( 42% v 22% for category 1 , 56% v 44% for category 2 and 60% v 50% for category 3 when comparing age ≥65 to <65 years ) .
In describing a large cohort of patients aged ≥65 years , we have studied for the first time this unique population with M . ulcerans disease . Cohorts described from Africa , where most M . ulcerans cases are reported , mainly involve children with few numbers of patients aged ≥65 years[1 , 2] . Additionally , studies reported from Australia have focused on cohorts across all age-groups[3 , 13 , 14] . This study therefore provides important new information pertaining to the epidemiology , clinical characteristics , treatment and outcomes in elderly populations . Patients aged ≥65 years represent an important subgroup in our cohort with M . ulcerans disease comprising two out of every 5 patients . Previous reports suggest that they may have an increased incidence of disease[4] . Additionally , our study suggests that they have more advanced and severe disease at presentation with an increased rate of multiple , large and oedematous lesions . Early non-ulcerative lesions ( plaques or nodules ) were infrequently reported . This is not due to late presentation as in our study the time from reported symptom onset to presentation for care was reduced in elderly patients . Instead this may be related to reduced immunity in older populations that inhibits the control of M . ulcerans leading to larger and oedematous forms of disease and the dissemination of lesions to other sites . This would be similar to the effect of HIV induced immune suppression which is associated with more severe M . ulcerans disease with an increase in the size , number and proportion of advanced lesions[15] . The reduced immunity in elderly populations may relate to the increasing immune suppression associated with senescence , and the increased presence of immunosuppressive conditions such as diabetes and malignancy or the increased likelihood of receiving immunosuppressive medication . Our study demonstrates that treatment of M . ulcerans disease in elderly populations is associated with increased toxicity and reduced effectiveness . Nearly one-half ( 42% ) of patients aged ≥ 65 years had to cease an antibiotic due to complications at a rate 5 times higher than younger populations , and complications were more severe with nearly one-fifth ( 18% ) requiring hospitalization . However this cannot be avoided by treating without antibiotics as treatment with surgery alone resulted in a 79% increased failure rate in this age-group . Furthermore , there is a two-fold increase in antibiotic-associated paradoxical reactions which can cause significant morbidity and complicate treatment[11 , 16] . Therefore there is an urgent need to develop less toxic and more effective treatments for elderly populations . All of the most commonly used oral antibiotics active against M . ulcerans have significant drug interactions . Rifampicin induces , and ciprofloxacin and clarithromycin inhibit , the cytochrome P450 enzyme system[17 , 18] leading to interactions with many commonly used medications . Clarithromycin and fluoroquinolones can prolong the QT interval creating a potential for serious arrhythmias if combined with other medical conditions or medication who do the same . This makes treatment more difficult and increases the risk of toxicity in elderly populations who are frequently prescribed multiple other medications . In addition , the pharmacokinetics of the antibiotics may differ with increasing age potentially resulting in toxic levels with currently recommended doses . For example it has been shown that elderly patients have higher serum concentrations and a longer half-life for antibiotics due to either increased bioavailability ( ciprofloxacin ) and reduced renal function with age ( ciprofloxacin and clarithromycin ) [19 , 20] . We advocate that pharmacokinetic and pharmacodynamic studies of frequently used antibiotics be performed in elderly patients to explore the safety and effectiveness of current and lower doses of antibiotics , including intermittent dosing regimens ( e . g . thrice weekly ) . Further research should also be performed on the safety and effectiveness of shorter duration antibiotic regimens[21] . It would be worthwhile exploring the use of alternative antibiotics such as the new anti-tuberculous agent bedaqueline , which shows strong bactericidal activity against M . ulcerans in mouse models[22] , and avermectins which have shown promising in vivo activity against M . ulcerans[23] , as these may be equally effective but potentially less toxic in elderly populations . Elderly patients were also found to have increased rates of antibiotic-associated paradoxical reactions , independent of lesion size . These likely occur due to the reversal of mycolactone induced immune suppression and the increased antigenic stimulus provided by dying mycobacteria when antibiotics are administered[24 , 25] . The increased rate in elderly patients in theory could relate to an increased organism load secondary to their relatively weakened immune systems which provides a greater antigenic stimulus combined with a greater potential for rapid immune function improvements when the inhibitory effects of mycolactone toxin are removed with antibiotics[11] . Increased paradoxical reactions contribute to the increased toxicity associated with antibiotics in elderly patients , and research is required to try and understand the reasons for their increased incidence in this age-group and to try and minimise their impact . Our early experience is that pre-emptive corticosteroids commenced at the initiation of antibiotics may prevent paradoxical reactions in elderly patients with oedematous lesions[26] and this should be further studied . Finally it should be noted that M . ulcerans is not without mortality in elderly patients where sepsis secondary to skin ulceration , or complications of treatment , can contribute to death in patients with significant co-morbidities or frailty due to age . We acknowledge the limitation that this is an observational study and as treatments were not randomized between groups there may be unmeasured confounders that may have influenced the results . However the cohort is large , data is collected prospectively and rates of follow-up are very high supporting the validity of our findings . In conclusion , elderly patients comprise a significant proportion of M . ulcerans disease patients in Australian populations and present with more severe and advanced forms of disease . Currently recommended M . ulcerans treatments are associated with increased toxicity and reduced effectiveness in elderly populations . Increased efforts are required to diagnose M . ulcerans earlier in elderly populations , and research is urgently required to develop more effective and less toxic treatments for this age-group .
|
Mycobacterium ulcerans is an infection that can affect all age-groups . It causes necrosis of skin and soft-tissue often resulting in severe outcomes and long-term disability . However , due to the majority of infections worldwide occurring in children and young adults , there is a paucity of information available in elderly patients . It is important that elderly patients are not neglected as the clinical presentation and treatment outcomes may differ significantly from younger patients related to relative immune defficiencies , co-morbidities and increased potential for drug interactions . We specifically examined patients with M . ulcerans disease aged ≥ 65 years and showed that they comprise a significant proportion of patients affected in Australian populations . They present with more severe and advanced disease forms , and suffer from increased toxicity and reduced effectiveness of the currently recommended treatments . Therefore , our study demonstrates that increased efforts are required to diagnose M . ulcerans disease earlier in elderly populations , and that research is urgently required to develop more effective and less toxic treatments for this age-group .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[] |
2015
|
Mycobacterium ulcerans in the Elderly: More Severe Disease and Suboptimal Outcomes
|
Humans employ a high degree of redundancy in joint actuation , with different combinations of muscle and tendon action providing the same net joint torque . Both the resolution of these redundancies and the energetics of such systems depend on the dynamic properties of muscles and tendons , particularly their force-length relations . Current walking models that use stock parameters when simulating muscle-tendon dynamics tend to significantly overestimate metabolic consumption , perhaps because they do not adequately consider the role of elasticity . As an alternative , we posit that the muscle-tendon morphology of the human leg has evolved to maximize the metabolic efficiency of walking at self-selected speed . We use a data-driven approach to evaluate this hypothesis , utilizing kinematic , kinetic , electromyographic ( EMG ) , and metabolic data taken from five participants walking at self-selected speed . The kinematic and kinetic data are used to estimate muscle-tendon lengths , muscle moment arms , and joint moments while the EMG data are used to estimate muscle activations . For each subject we perform an optimization using prescribed skeletal kinematics , varying the parameters that govern the force-length curve of each tendon as well as the strength and optimal fiber length of each muscle while seeking to simultaneously minimize metabolic cost and maximize agreement with the estimated joint moments . We find that the metabolic cost of transport ( MCOT ) values of our participants may be correctly matched ( on average 0 . 36±0 . 02 predicted , 0 . 35±0 . 02 measured ) with acceptable joint torque fidelity through application of a single constraint to the muscle metabolic budget . The associated optimal muscle-tendon parameter sets allow us to estimate the forces and states of individual muscles , resolving redundancies in joint actuation and lending insight into the potential roles and control objectives of the muscles of the leg throughout the gait cycle .
Human walking relies on a complex interplay of several physiological systems , with each exhibiting some degree of redundancy . The nervous system directs muscle contraction while receiving input from many different neural pathways . Muscles work together to produce motion , but different combinations of muscle action can produce the same net torque at a given joint . Tendons provide the interface between muscle and bone , but the energy transferred to the skeleton can come from either the active muscle or the compliance of the tendon . Understanding how humans resolve these redundancies has been a long-standing problem in the fields of neuroscience and biomechanics [1 , 2] . Knowledge of how the neuromuscular system allocates load during a given task would provide insight into the control objectives that govern its actions . Potential objectives ( reviewed in [3] ) include joint trajectory planning or minimization of metabolic energy consumption , active muscle volume , or muscle fatigue . However without an adequate understanding of the roles of each component of the system , such control hypotheses remain mere speculation . The roles of individual muscles and tendons in producing motion depend on the neural drive to the muscles and on the force generation properties of both the muscles and the tendons . Several modes of experimental observations provide glimpses of these elements . Electromyography ( EMG ) can be used to quantify the neural drive to individual muscles , revealing which muscles contribute to a given movement and giving some measure of intensity [4–6] . However it is limited by signal variability , measurement artifacts , and an inexact mapping to physiology and muscle force . Ultrasound probes have recently been used to image the individual motions of some distal leg muscles and tendons in vivo [7–13] , but are practical only for small muscles and limited tasks . Motion capture can be combined with a knowledge of anatomy to infer the net movement of muscle-tendon units under much more general circumstances [14]; however breaking the resulting movement profiles into individual muscle and tendon contributions requires knowledge of often unavailable force generation parameters of muscles and tendons . These parameters are typically estimated through cadaver studies , but the scaling of the relevant quantities among different muscles and subjects ( not to mention the differences with living specimens ) is not well understood [14–16] and can have a significant impact on the resulting modeled dynamics [15 , 16] . Given the incomplete view afforded by current experimental measures , a unifying theoretical framework that combines the available data modes into a model of neuromuscular function is desirable . Two primary approaches have been taken to address this issue in walking: optimal control and optimal design . Human walking studies based on optimal control [17 , 18] model the morphology of leg muscle tendon units ( MTUs ) using literature-based estimates . They infer muscle activation through optimization , choosing control objectives such as metabolic energy minimization and/or motion tracking . They have been successful in predicting joint moments , joint torques , and ground reaction forces but often significantly overestimate the metabolic cost required for locomotion [17 , 19] . While it is clear that the neural control of the biological system is optimized in some way , it may be infeasible to determine the true objective function for this approach . Many different muscle activation combinations can produce similar muscle torque values , and several different unknown control objectives and neurological factors may contribute at once . Further the underlying uncertainty in the muscle-tendon morphology may result in the excess metabolic cost observed , as improper leveraging of tendon compliance would affect muscle force and state and therefore metabolic estimates . Human walking models based on optimal design utilize the efficiency gains that can be made through MTU parameter tuning . As Lichtwark and Wilson showed [20] , experimentally observed muscle-tendon strains may be predicted by maximizing the efficiency of isolated MTUs . This result likely stems from the well-documented ability of tendon to enable muscle to operate economically [21 , 22] . Krishnaswamy and Herr [23] further explored the potential of optimal design , estimating the torque breakdown of the muscles spanning the ankle during the stance phase of walking at self-selected speed . This work used EMG signals to estimate muscle activation during walking and developed an optimization framework based on the assumption that the morphology of the muscle-tendon units spanning the ankle has evolved to minimize the metabolic cost required for walking at self-selected speed . Its results indicated that one solution set is able to match both human metabolics and kinetics , demonstrating an efficient load-sharing amongst the plantar flexors that qualitatively matched available experimental data . In this work we further the study of [23] , modifying and extending it to permit investigation of the full leg . We collected kinematic , kinetic , electromyographic , and metabolic data from five subjects walking at self-selected speed and used them to perform optimizations with prescribed skeletal kinematics . We varied the parameters that determine where each muscle operates on its force-length curve as well as those that shape the force-length relation of each tendon , seeking to simultaneously minimize metabolic cost and maximize agreement with the observed joint moments . Each muscle in the model was modeled as Hill-type [19 , 24] and driven by activation estimates produced from EMG data . Overviews of the system model and procedure are shown in Figs 1 and 2 , respectively . We found that the correct metabolic consumption can be matched with reasonable fidelity in the modeled joint torque through application of a single constraint on the per-muscle metabolic budget . The resulting optimal parameter sets were used to compute muscle force and state , with fascicle length profiles being compared to available experimental measurements . Our results are organized as follows . First we provide our estimated muscle activation profiles , discussing their quality and implications . Second we display the results of our dual objective optimization problem and summarize the methodology used for choosing one optimal solution for each subject . Third we evaluate the optimal solutions , producing estimates of energetic variables and muscle state . We compare these results with available experimental measures , finding quantitative agreement with metabolic data and qualitative agreement with muscle fascicle length data .
Muscle activation provides a scaling factor for the active force generation capability of a given muscle at a given time . As described in Methods , we applied a hybrid approach similar to that of [23] to estimate muscle activation from surface EMG measurements of five participants during walking . A Bayesian algorithm first proposed by Sanger [25] was tuned ( as described in Methods ) and used to perform a hidden state estimation that effectively determined the neural excitation of each muscle . This method was chosen over more conventional bandpass filtering methods [4–6] because the slightly delayed timing of the profiles it produced more easily allows the production of the observed joint torques . We elaborate on this point further in the Discussion section . The estimated neural excitation was then passed to a shaping filter [26 , 27] that represents muscle activation dynamics . The resulting average profiles are plotted in Fig 3 . As can be seen from this plot , significant variations occurred ( both within and among subjects ) in the activation estimates obtained from the muscles spanning the hip . This lack of consistency was likely due to some combination of the relatively large depth of these muscles beneath the skin , motion artifacts , and the relative inaccessibility of the area . To prevent this from compromising the ensuing analysis , neural excitation profiles from the wire electrode experiments of [28] were used for the monoarticular muscles spanning the hip . These neural excitation profiles were passed to the activation dynamics from [26 , 27] for temporal consistency and subsequently used as model input along with the data-based activation estimates from all other muscles . Further details about this procedure are given in Methods and S1 Text . While there is no available ground truth to compare our activation estimates to , we note that the time dependence of the results ( when normalized to percent gait cycle ) were relatively invariant across trial and subject . The profiles follow the expected build up and decay time scales of muscle activation and , as can be seen in the following , allow the model to produce realistic kinetic and metabolic results . The estimated muscle activations a ( t ) as well as joint kinematics θjoint ( t ) derived from motion capture data were used to actuate the full leg model shown in Fig 1 according to the scheme shown in Fig 2 . The leg muscle-tendon model M was specified by a set m → of morphological parameters that describe the force generation characteristics of each MTU as well as two parameters that enable passive force generation by the iliofemoral , ischiofemoral , and pubofemoral ligaments as well as other connective tissue at the hip . These ligaments are known to prevent hip overextension and here allow for the recovery of elastic energy in the joint [18 , 29]; specifically they reduce the load on the iliacus muscle around toe-off . The contributions of each MTU to m → were its maximum isometric force Fmax , an overall scaling factor for tendon slack length lsl and optimal muscle length lopt , its tendon reference strain λref , and its tendon shape factor Ksh . The lumped hip flexor ligament ( HFL ) acted as a simple rotary spring , parameterized by spring constant KHFL and engagement angle θ0 , HFL . Each parameterization of the model generated kinetic ( τmod ( t ) ) and metabolic ( C ) output costs: M m → , a ( t ) , θ j o i n t ( t ) → τ m o d ( t ) , C . ( 1 ) The parameter vector m → was varied using a stochastic dual objective optimization scheme [30] that simultaneously minimized metabolic cost and the difference between the computed joint moments of the data and those produced by the model . The bounds were specified as described in the Materials and Methods section; they were chosen wide enough that they were not approached by the chosen optimal parameter sets . The solution spaces for this optimization are shown in Fig 4 . For each participant , the set of Pareto optimal solutions ( i . e . the set of solutions where one would have to compromise on one objective to improve on the other ) forms a rounded corner in the objective space . In the ideal case , this corner would be sharp and consist of one solution that optimizes both objectives . However this ideal does not typically occur in noisy ( realistic ) systems and does not here . To generate predictions based on our model , we chose one optimal parameter set along the Pareto Front for each subject . This was accomplished by evaluating the per-muscle metabolic consumption among all Pareto optimal solutions ( Fig 5 ) . Within the set of solutions , those where the metabolic cost was low and the kinetic fit was poor were seen to have uniformly low expenditure per muscle . Those that had the very best kinetic fits but relatively large metabolic costs were seen to be sinking large amounts of metabolic energy into a small number of muscles to produce incremental improvements in the kinetic fit . Such a phenomenon was likely enabled by noise in the data ( particularly in the EMG signals ) and the imperfect ability of our lumped , partial muscle set to match the force produced by the full set of the human body . The ramp up in metabolic energy seen in this subset of muscles is not physical as it would lead to either rapid fatigue or to the muscle being modeled as much larger than it actually is ( since muscle mass scales with Fmax ) . Hence we excluded solutions that displayed this behavior , choosing as optimal the remaining Pareto optimal solution with best kinetic fit . Mathematically this was achieved by setting a cutoff on the fractional expenditure of the vastus , as the metabolic cost of this muscle group was the largest and increased significantly as kinetic fit improved . The chosen fractional cutoff allowed us to apply one criterion to every participant to match the experimental metabolic cost , as shown in Figs 4 and 5 . Table 1 shows how our choice of optimal solution is able to quantitatively match the experimentally-observed metabolic cost of transport in four out of five subjects ( and on average ) while maintaining acceptable joint moment agreement . The lone subject where quantitative metabolic agreement was not reached displayed only a 6% error . Further details of the cutoff are included in Materials and Methods and in S3 Text while the optimal parameter set for each participant is given at the end of this document . To further evaluate the quality of our joint moment estimates we computed the fractional mean absolute error ( FMAE ) F M A E = 1 N 1 range ( τ e x p ) ∑ i = 1 N τ e x p , i - τ m o d , i ( 2 ) between the modeled joint moments τmod and the experimentally observed joint moments τobs . These quantities are computed only over the stance phase to facilitate comparison with those generated from another current EMG-driven analysis , [31] . The quantities quoted from [31] represent an average over four different activities ( walking , running , side stepping , and crossover ) but are generally close to those they generate for walking only ( except for the hip , which had lower errors for walking ) . Their analysis included two treatments; one that considered only one degree of freedom and one that considered multiple degrees of freedom ( as in our model ) . In general our moment fits compare favorably ( Table 2 ) . The metabolic expenditure of our model also aligns with previously published results . Across subjects , we found the average efficiency of positive muscle work to be 0 . 26 ± 0 . 02 , consistent with [23 , 32 , 33] . We also estimated the metabolic expenditure of the model during different portions of the gait cycle by cross referencing the simulation with the input force plate data . The results are compiled in Table 3 and show a breakdown that is very similar to that simulated in [34] . The optimal muscle-tendon parameter sets estimated by our optimization procedure provide a means to resolve the redundancy in joint actuation for each subject . When applied in conjunction with the computed kinematics and estimated muscle activations , individual muscle force and state may be estimated . Fig 6 shows the torque breakdown for each joint in terms of percentage of body weight times height and averaged over all subjects . Fig 7 shows the trajectories of the muscle fascicle length normalized by lopt and averaged over all subjects . Note that the spread of some of these profiles is not due to a difference in general shape but rather to an overall offset in length , as evidenced by their nearly constant standard deviations . The soleus , hamstring , and vastus fascicle lengths in this plot all agree quite well in both shape and offset with the predictions in [35] . The soleus force and length also agree qualitatively with the projections of [11] . Fig 8 shows the velocity trajectories for each muscle fascicle normalized by its maximal value vmax and again averaged over all participants . As can be seen from the small variation in the shapes of the profiles in these plots , muscle force and state followed similar trajectories across participants . These predictions were also seen to vary little along the Pareto front in the experimentally measured metabolic band of a given subject . In general , muscle fascicle state is extremely difficult to measure . The only currently available means to obtain these profiles is ultrasonography , which is only practical for the relatively short distal muscles of the leg . In Fig 9 the modeled fascicle trajectories of muscles from one subject are compared with the experimental profiles available from published ultrasound studies . In each case the muscle fascicle lengths lm are normalized by their length at heel strike , lmHS . To generate these plots we took the modeled muscle from the subject who most closely matched the average height and weight of the experimental study . Soleus and gastrocnemius profiles came from Ishikawa et al [8] , a gastrocnemius profile came from Fukunaga et al [7] , and the vastus lateralis profile came from Chleboun [10] . In the plantar flexors , long stretches of nearly isometric operation are observed in mid-stance in both the model and in vivo profiles . In the vastus , the fascicle trajectory is seen to somewhat track the flexion of the knee in both the model and the published data . However while the qualitative trends of each muscle are consistent , quantitative agreement is not observed . We believe that observed differences come from ( i ) the difference in walking speed between our study and the literature , ( ii ) natural variation in the kinematics of early stance ( which affects the initial muscle length for normalization ) , and ( iii ) uncertainty in the breakdown of what constitutes muscle and what constitutes tendon in ultrasound studies . We hope that future experimental methodologies , perhaps employing implantable sensors , will be able to further test our predicted fascicle trajectories .
Several observations may be made about both the methodology employed and the results obtained in this study . On the methodology side , we first address the steps taken to estimate the activations of the muscles in our model . As mentioned above , the conventional approach for estimating muscle excitation based on EMG data involves a bandpass filter [4–6] . We tried this approach with our optimization scheme ( on all subjects ) but found that it provided an inferior ability to match the observed joint moments compared to the implemented method . Digging further into this we found that two main differences existed between the approaches; ( i ) the chosen Bayesian method produces a profile that turns on and off more sharply than the signal produced by the bandpass filter and ( ii ) the signal produced by the Bayesian method consistently lags the bandpassed signal by about 50 ms . The sharpness of the profile does not affect the results significantly as it is mostly washed out by the ensuing activation dynamics and averaging . However the time dependence does matter; this lag enables the build up of muscle force in a manner consistent with the observed joint torques . Interestingly we found that [36] introduced a 40 ms lag to their EMG signal to “account for electromechanical delay between surface EMG and force production . ” This lag was employed by other studies [37 , 38] and fell within the 10−100 ms range given for these processes in the literature [39 , 40] . We found that adding a 40 ms lag as in [36] to the excitations produced by bandpassed filtered EMG signals gave performance nearly as good as those provided by the Sanger algorithm , with less variation among gait cycles . However this lagged bandpass method does not lend itself to a biophysical interpretation as clearly as the Bayesian model does . Despite our best efforts at EMG data collection and activation estimation , deficiencies in this part of our data set clearly exist . Surface EMG in general is prone to noise and artifacts , and Sanger’s algorithm ( tested only isometrically in his publication [25] ) does not remove them . While the effects of inconsistent artifacts was minimized by discarding the EMG data from clearly compromised gait cycles and the use of average trajectories in our model , they were likely not removed entirely . Better results may be obtained in future work through fine wire EMG measurements , which while more invasive are known to produce more reliable signals . Here generic fine wire EMG profiles reported in [28] were used to replace the noisy surface EMG measurements of the muscles spanning the hip . Interestingly our model actually displayed slightly better agreement with the observed hip torque profiles than with those of the other joints , but this is misleading as most of the modeled hip moment came from the hamstrings and the hip flexor ligament ( which were unaffected by the generic profiles ) . A more minor deficiency in the EMG pipeline was normalization by the maximal voluntary contraction ( MVC ) values . While we do not know of a better alternative , this approach did lead to the normalized excitations of our muscles occasionally exceeding one in fast walking trials . When that occurred we renormalized by the value in the fast walking trial and reprocessed , but the normalization constant could still have been too small in other cases . Fortunately the impact of this scaling is extremely minimal because the estimated activation directly multiplies the maximum isometric force Fmax , which is optimized . A small effect remains because the normalization occurs before the excitation undergoes the activation dynamics ( 5 ) , but that effect is largely irrelevant to our results . The system identification component of this study produced a methodology for estimating muscle-tendon parameters capable of matching the measured metabolic consumption while producing joint torque profiles that tracked observations reasonably well . One criterion based on the metabolic consumption of the vastus muscle group ( the largest in the model ) enabled the metabolic match for all participants . While the accuracies of our modeled joint torque profiles compare favorably with other current EMG driven modeling procedures [31] , they do tend to underestimate the required joint torques . This characteristic is likely due to the exclusion of some muscles as well as the lumping of some muscle groups . In particular the deficiency exhibited in ankle moment during late stance is likely due to the exclusion of smaller plantar flexor muscles which are known to engage during that time [28] . Similarly the lumping of the three hamstring muscles ( semimembranosus , semitendinosus , and biceps femoris long head ) may be responsible for some of the deficiencies observed in the knee flexion moments throughout the gait cycle . Further resolution of these and other muscle groups where the muscle activations are not quite concurrent and the muscle-tendon lines of action are not quite aligned would likely improve the predictive power of this approach , allowing more accurate determination of the roles of each muscle during a given task . One notable aspect of the model was the importance of the hip flexor ligament , which produced nearly all of the required hip flexion moment near toe off at no metabolic cost . Its linear form was chosen as in [29] for maximal simplicity , but did not agree with the nonlinear damped form used in previous work [18 , 41] . Given that it produced torque for free it may have suppressed the required action of the other hip flexors ( notably the iliacus ) near toe off and produced a better hip moment fit than would have otherwise been possible for the same metabolic cost . However the suppression of other hip flexors could not have been large as these muscles are not strong enough to produce the required torque alone and are not significantly stretched in this time frame . It is also known that the hip ligaments produce the flexion torque necessary to balance the upper body against gravity in a standing position , where the line of gravity passes posterior to the hip joint [42] . They enable people to stand erect and even carry extra weight without significant muscle work at the hip , allowing a low metabolic cost to be maintained [43] . Since the engagement angles we used are consistent with standing and damping in human connective tissue is believed to yield only a slight drift over walking time scales [16] , we believe that the contribution of the hip flexor ligament in our model is physiologically reasonable . The optimal muscle-tendon parameters found in this study play different roles in facilitating efficient locomotion . The maximal muscle isometric force ( Fmax ) must be large enough to meet the torque requirements at each joint , but small enough to keep muscle size and metabolic cost reasonable . The tendon slack lengths lsl and muscle optimal lengths lopt govern the timing of force production , acting in concert with muscle activation . The tendon shape parameters λref and Ksh define the elastic properties of the tendon and are tuned to ensure correct muscle operation and joint actuation . The resulting muscles and tendons together produce an interconnected system capable of producing the joint torque necessary for locomotion at a minimal metabolic cost . Evaluating the velocities at which the muscles in our model contract while activated can lend insight to the goals of their control . As was emphasized in [29 , 44] , muscles minimize metabolic consumption at low speeds ( i . e . when operating approximately isometrically ) . Further , A . V . Hill [45] demonstrated that skeletal muscle maximizes its efficiency while shortening at vCE ≈ −0 . 17vmax and its power output at vCE ≈ −0 . 30vmax . These three speeds ( vCE = 0 , −0 . 17 , −0 . 30vmax ) are indicated by the horizontal dashed lines in Fig 8 . Combining this with Fig 3 allows us to contextualize the modeled velocities of each muscle when activated . At the ankle , the tibialis anterior is seen to operate at low speeds when engaged . This minimizes metabolic cost and is consistent with isometric contractions , as modeled in [29 , 44] . Both the soleus and gastrocnemius are seen to operate approximately isometrically through their activation in mid-stance before rapidly increasing their contraction velocity in a power stroke toward the end of stance ( ≈ 60% GC ) . The required positive work of the plantar flexors is consistent with the need seen in [29 , 44] , but their rapid contraction at the end of stance does not strictly agree with the efficiency goal of the soleus and power goal of the gastrocnemius noted in [23] . This inconsistency amounts to a phase difference; the observation in [23] was based on muscle speed at toe off but our plantar flexor velocities reach approximately the same levels around 53% GC . At the knee , most muscles are seen to operate approximately isometrically . This agrees with [29 , 44] , which note that these muscles primarily serve to modulate the stiffness of the joint in an optimally economical fashion . It is worth noting that the vastus group is seen to re-engage at the end of swing near the the optimal efficiency regime ( extending the knee for heel strike ) before returning to the low speed regime . At the hip , most muscles are observed to contract at low speed when activated . Two exceptions are the iliacus and the adductor longus , both of which gravitate toward maximal efficiency as they flex the hip around toe off . The adductor magnus may also approach the maximal efficiency regime as it extends the hip around heel strike , but this is less clear . Endo et al [29 , 44] found that the muscles spanning the hip could not be modeled strictly isometrically , consistent with these observations . Several avenues exist for extending this work in future studies . One route would be to expand the model past the sagittal plane; as shown in [31] matching moments in three dimensions could have an impact on the optimized parameter sets . Another route would be to validate the model by testing under different walking conditions . If more reliable EMG measurements could be obtained , we could train the model based on level-ground walking at self-selected speed and then evaluate the ability of the optimal parameter set to match the observed joint torque profiles and metabolic costs under different conditions ( speeds , inclines , etc . ) In this case new experimental observations of the change in plantar flexor function across speed [12 , 13] could be used to evaluate the model . Another possibility is that our overarching hypothesis is incomplete- maybe the body , instead of adapting its morphology to maximize its efficiency at ( the presumably most common task of ) walking at self-selected speed , actually optimizes with multiple tasks in mind . One could imagine additionally collecting data from a subject running a self-selected speed and adding extra dimensions to the cost function wherein that task is optimized concurrently with walking at self-selected speed . The procedure followed here for self-selected speed walking could thus become part of a larger optimization scheme by including other tasks . While such experiments could lend considerable insight , they will undoubtedly prove challenging due to the difficulties associated with collecting EMG under the stated conditions . One alternative would be to simulate the neural control of the model . Recent forward dynamic models driven by reflexive feedback have shown the ability to walk stably across different terrains [19 , 46] , at different speeds [47 , 48] , and in three dimensions [48 , 49] . These models are built for a subject with “average” dimensions and could therefore be improved through data-driven customization for individual subjects and the inclusion of more realistic muscle-tendon geometries and morphologies . This would include scaling segment inertias , refining muscle-tendon lines of action and moment arms [14] , and optimization of muscle-tendon morphologies to simultaneously minimize metabolic cost and maximize agreement to experimentally observed kinematic and/or kinetic data . In this way a more realistic representation of an individual subject could be obtained , leading to further insights about the roles of individual muscles and tendons during gait and their variation amongst subjects . One could also imagine altering the neural control in this paradigm and using the framework to study movement disorders such as cerebral palsy .
The experiments of this study were conducted in compliance with the principles of the Declaration of Helsinki . The study was approved by the MIT Committee on the Use of Humans as Experimental Subjects ( Protocol 1101004266 ) . Prior to the experiments all participants provided written consent for data collection , analysis , and publication . Kinematic , kinetic , electromyographic , and metabolic data were collected at the Harvard University Skeletal Biology Lab . Five healthy adult males participated in the study , with their average height , mass , and age being 1 . 77 ± 5 m , 70 . 4 ± 6 . 3 kg , and 26 ± 2 years , respectively . The required data sets were collected in two phases . First the subjects were outfitted with a portable oxygen consumption mask attached to a Cosmed K4B2 VO2 system . This system employs a standard open-circuit gas analysis technique to estimate metabolic energy consumption based on measurements of oxygen inspired and expired [50] . Three of the subjects were asked to stand still for seven minutes while a basal measurement was recorded , with the rates of the other two participants taken to be the mean of these three closely grouped rates ( 1 . 38 ± 0 . 06 W/kg , 1 . 60 ± 0 . 07 W/kg , 1 . 63 ± 0 . 06 W/kg ) . Participants then walked barefoot on an instrumented treadmill for seven minutes at each of six speeds ( 0 . 75 m/s , 1 . 00 m/s , 1 . 25 m/s , 1 . 50 m/s , 1 . 75 m/s , and 2 . 00 m/s ) , allowing the variation of metabolic energy expenditure to be measured across speed . The results were quickly tabulated and used to estimate the walking speed where the metabolic cost of transport ( MCOT ) was minimal . Once the metabolic cost measurements were completed , the oxygen consumption mask and Cosmed system were removed and each participant was outfitted for the second phase . In this phase kinematic , kinetic , and electromyographic data were collected for two minutes of barefoot walking at each of seven speeds; the six listed above and the speed where the subject’s MCOT was found to be minimal . An infrared camera system ( 8 cameras , Qualisys Motion Capture Systems , Gothenburg , Sweden ) was used to track the motion of subjects as they walked in the capture volume . Reflective markers were placed at 43 ( bilateral ) locations on the participant’s body and their three dimensional trajectories were recorded at 500 Hz . The marker locations were chosen specifically to track joint motion , as prescribed by the Helen Hayes marker model . The ground reaction forces and contact centers of pressure were measured using a split-belt instrumented force plate treadmill ( Bertec Corporation , Columbus , OH ) . Electromyographic signals were collected using a surface system from Motion Lab Systems ( Baton Rouge , LA ) and electrode placements as dictated in [51] . Fourteen muscles ( tibialis anterior , soleus , medial gastrocnemius , vastus lateralis , biceps femoris shorthead , rectus femoris , semimembranosus , biceps femoris long head , illiacus , gluteus maximus ( lower ) , gluteus maximus ( upper ) , gluteus medius , adductor longus , and adductor magnus ) on one leg of each subject were recorded , with symmetry being assumed for the other leg . The signals were recorded at the surface using pre-gelled bipolar electrodes ( Electrode Store Model BS-24SAF , part number DDN-20 ) , sampled at 1000 Hz , and amplified 20 times by pre-amplifiers ( Motion Lab Systems , part number MA-411 ) . Prior to walking trials a maximum voluntary contraction ( MVC ) trial was conducted for each muscle group wherein the participant was asked to work that particular group as hard as possible . These MVC trials were used for normalization purposes in muscle excitation estimates . The processed data were used to build a dynamic model of the leg during walking ( Fig 1 ) . The model includes all muscles that make significant contributions to the components of ankle , knee , and hip torque perpendicular to the sagittal plane during walking . Several muscle groups were lumped together for simplicity; this was deemed appropriate if all muscles within a group had similar lines of action and were activated simultaneously during walking . These included the GAS group ( medial and lateral gastrocnemius ) , the VAS group ( vastus lateralis , medialis , and intermedius ) , the HAM group ( semimembranosus , semitendinosus , and biceps femoris long head ) , and the ILL group ( iliacus and psoas ) . Each muscle group had one effective tendon to represent the net compliance of the muscle’s interaction with the skeleton . In addition to the muscle-tendon units a passive ligament spanning the hip was included to allow for the recovery of elastic energy in that joint . This element represents the contributions of the iliofemoral , ischiofemoral , and pubofemoral ligaments as well as those of other connective tissue at the hip and was modeled using a nonlinear equation in [18]; for simplicity we took it to be a linear spring that only engages around the time the hip goes vertical and begins to extend [29]: τ H F L = - K H F L ( θ h i p - θ 0 , H F L ) . ( 6 ) Note that both the moment and the angle in this equation were defined with flexion being positive . Below we describe the muscle , tendon , and joint dynamics of the system . An overview of the applied optimization procedure is shown in Fig 2 . There were two categories of inputs for each muscle in our model: ( i ) a ( t ) , lmtc ( t ) , and ri ( t ) - all estimated from the data and ( ii ) muscle-tendon morphological parameters m → i which were identified via optimization . Morphological parameters which affect the Hill-type contraction dynamics of the modeled muscles include muscle maximum isometric force Fmax , length where active muscle force is maximal lopt , fiber composition FT , tendon slack length lsl , tendon shape factor Ksh , and tendon reference strain λref . We evaluated the consequences of varying all of these variables , finding differing levels of sensitivity for each . The variables to which the model was insensitive were fixed , leaving Fmax , Ksh , λref , and an overall scaling factor for lsl and lopt as the parameters to optimize . The last factor was chosen to ensure that the muscle operated in reasonable regions of its force-length space while preventing overfitting . The ratio lsl/lopt is known to vary significantly among muscles but not significantly in the same muscle among subjects [14] , justifying this choice . As mentioned above , our optimization is built on the hypothesis that the muscle-tendon morphologies of the muscles comprising the leg have evolved to minimize the metabolic cost of walking at the speed where the MCOT is minimal . This assumption is supported by the fundamental importance of bipedal walking as a means of transport and the tendency of humans to walk at or near this speed . For each subject we therefore sought a set of morphological parameters m → i that match the measured joint torques at the ankle , knee , and hip while consuming a minimal amount of metabolic energy . This amounts to a dual objective optimization problem , with metabolic and kinetic cost functions being simultaneously minimized . The specific forms of these two costs are summarized below . As mentioned above , optimal muscle-tendon parameters were identified using a dual objective optimization . Fifty morphological parameters were optimized for each subject; four morphological parameters for each of the twelve modeled muscles and two additional parameters describing the hip flexor ligament . The chosen muscle-tendon parameter bounds are shown in Table 4 . The maximum isometric force ( Fmax ) values for each muscle were constrained to fall in a window surrounding the scaled value from SIMM . The large width of this window helped to compensate for uncertainties in the EMG normalization ( via MVC values , as discussed above ) . The scaling factor for lslack and lopt was chosen to ensure that the muscle fascicle length stayed within reasonable physiological operating ranges . The bounds for Ksh and λref were taken from [16] . The bounds for the spring constant of the hip flexor ligament was chosen so that the ligament could provide anywhere from none to all of the required hip flexion moment near toe off . The engagement angle was chosen so that the tendon could turn on with the hip no more than 10° flexed , as its physiological role is to prevent overextension . The model was constructed using MATLAB and Simulink and integrated using the ode15s ( stiff/NDF ) solver . Computations were parallelized and carried out using the Mathworks Cloud Center , a computer cluster operated through Amazon Web Services . This resource proved useful as our runs were configured to test many solutions and took about 10 hours to run on a single CPU . The optimization algorithm employed was MATLAB’s gamultiobj , a controlled elitist genetic algorithm that is a variant of the NSGA-II algorithm [30] . This method was chosen because of the likely presence of local minima in both objectives and tuned as shown in Table 5 . Optimization tunings were chosen to ensure generations sufficiently large to force the optimizer to thoroughly search the space , eliminating the need for population seeding . Each trial solution was allowed to run for two gait cycles , with only the results of the second gait cycle being considered . This removed initial transient effects and allowed evaluation of the system over the full gait cycle .
|
Neuromuscular systems often employ redundancy in joint actuation , with different combinations of muscle and tendon action producing the same net joint torque . Both the resolution of these redundancies and the energetics of such systems depend strongly on the force-length relations of muscles and tendons . Many human walking models fail to properly account for elasticity by failing to scale muscle-tendon parameters for different individuals , relying instead on stock values taken from cadaver studies . This can result in inaccurate estimates of metabolic consumption as well as of the forces and states of individual muscles . Instead we estimate muscle-tendon parameters using a data-driven optimization procedure , testing the hypothesis that the human leg has evolved to maximize the metabolic efficiency of walking at self-selected speed . We find that the experimentally observed metabolic consumption can be matched with reasonable joint torque fidelity through the addition of a single constraint on the per-muscle metabolic budget . The associated muscle-tendon parameter sets were used to compute muscle force and state estimates , lending insight into potential roles and control objectives of the major muscles of the leg throughout the gait cycle .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"medicine",
"and",
"health",
"sciences",
"classical",
"mechanics",
"biological",
"locomotion",
"biomechanics",
"optimization",
"mathematics",
"pelvis",
"bioassays",
"and",
"physiological",
"analysis",
"muscle",
"electrophysiology",
"research",
"and",
"analysis",
"methods",
"musculoskeletal",
"mechanics",
"muscle",
"physiology",
"musculoskeletal",
"system",
"hip",
"joints",
"(anatomy)",
"connective",
"tissue",
"biological",
"tissue",
"electrophysiological",
"techniques",
"walking",
"torque",
"physics",
"anatomy",
"electromyography",
"tendons",
"physiology",
"biology",
"and",
"life",
"sciences",
"physical",
"sciences",
"motion"
] |
2016
|
Human Leg Model Predicts Muscle Forces, States, and Energetics during Walking
|
The world is continuously urbanising , resulting in clusters of densely populated urban areas and more sparsely populated rural areas . We propose a method for generating spatial fields with controllable levels of clustering of the population . We build a synthetic country , and use this method to generate versions of the country with different clustering levels . Combined with a metapopulation model for infectious disease spread , this allows us to in silico explore how urbanisation affects infectious disease spread . In a baseline scenario with no interventions , the underlying population clustering seems to have little effect on the final size and timing of the epidemic . Under within-country restrictions on non-commuting travel , the final size decreases with increased population clustering . The effect of travel restrictions on reducing the final size is larger with higher clustering . The reduction is larger in the more rural areas . Within-country travel restrictions delay the epidemic , and the delay is largest for lower clustering levels . We implemented three different vaccination strategies—uniform vaccination ( in space ) , preferentially vaccinating urban locations and preferentially vaccinating rural locations . The urban and uniform vaccination strategies were most effective in reducing the final size , while the rural vaccination strategy was clearly inferior . Visual inspection of some European countries shows that many countries already have high population clustering . In the future , they will likely become even more clustered . Hence , according to our model , within-country travel restrictions are likely to be less and less effective in delaying epidemics , while they will be more effective in decreasing final sizes . In addition , to minimise final sizes , it is important not to neglect urban locations when distributing vaccines . To our knowledge , this is the first study to systematically investigate the effect of urbanisation on infectious disease spread and in particular , to examine effectiveness of prevention measures as a function of urbanisation .
We are living in a world which is continuously urbanising . From a United Nations report [1] , we know that in 2014 , 54% of the population was living in urban areas . By 2050 , they estimate that 66% of the population will be living in urban areas . The number of “megacities” is also increasing [1] . Urbanisation involves clustering of people within a geographical area [2] . Migration from rural to urban areas is one of the key drivers of urbanisation , and results in spatial expansion of urban centers [3] . Though there are also other characteristics of urbanisation , like urban sprawl [2 , 3] and population growth , we here focus on this population clustering . By population clustering , we mean that large cities are often surrounded by other cities or suburbs with large population sizes . Rural areas also tend to appear in clusters ( i . e . positioned closely together ) . This has for instance been found to be the case for the Turku region in Finland , where regions of both high and low population densities are clustered [4] . Another example is Australia , where the majority of the population is clustered around the coastal belt [5] . From here on , the term urbanisation will be used to refer to the population clustering . We aim at studying the urbanisation phenomenon and its effect on infectious disease spread in a general setting . Our purpose is not to describe single outbreaks in specific populations , or finding the “best model” or strategy for a specific country , but rather study the phenomenon from a more generic and principled point of view . In this paper , we explore the effect of internal travel restrictions and vaccination on infectious disease spread , when the clustering of population is continuously varied . By internal travel restrictions , we will refer to restrictions on non-commuting travel within the country . If the disease dynamics is different for different levels of population clustering , this can have important implications for the effectiveness and design of interventions . In order to study this , we need a continuous series of countries where everything is fixed , apart from the urbanisation , which changes between the countries in a controllable and continuous manner . This is difficult in practice , and we therefore construct a fictional country for this purpose , however trying to maintain some realism . Our aim is not to develop a precise model for urbanisation in a country . We aim for a simple model , where urbanisation is controlled by one single tuning parameter , which captures key features of urbanisation . We use this model to generate a fictional country with a plausible population distributed in urban and rural areas . More specifically , we use a plausible population size distribution , plausible commuting patterns modelled by a gravity law and a plausible infectious disease model . We use a metapopulation infectious disease model , where an SEIR ( susceptible , exposed , infectious , recovered ) process [6] governs the disease dynamics in each location , and the different locations are coupled through individuals travelling between them . This framework allows us to investigate in silico how urbanisation affects various aspects of infectious disease spread . As a motivating example , we will study an influenza-like illness spreading in a single country , where we assume that the pathogen has already been imported to the country . We investigate how the infectious disease dynamics depends on the underlying population clustering in the country and focus on the effect of internal travel restrictions and three different vaccination strategies . The plausibility of the synthetic country is obtained by conserving the population size distribution of Norway and a gravity law fitted to data on commuting between Norwegian municipalities . We use Norwegian data to ensure reasonable population sizes , and plausible commuting for those population sizes , and because we have commuting data available on a relatively fine spatial scale . We are studying the urbanisation phenomenon generically , and we do not claim , nor is it our aim , that these results are directly applicable to Norway . This framework is a simulator , and it is not our purpose to model and provide results for a specific country . Our results allow a theoretical description of how urbanisation affects interventions to control epidemics . To our knowledge , this is the first modelling and simulation study which systematically investigates the effect of population clustering on the spread of infectious diseases . In particular , it is the first study to consider the effect of internal travel restrictions and vaccination in relation to urbanisation . This is a study of the urbanisation phenomenon represented in a very simplified and theoretical way , yet with important elements of realism . We focus on these two control strategies , because they are clearly affected by urbanisation . We do not consider international air travel restrictions , school closure or other sanitary measures , which were used for instance during the 2009 pandemic [7 , 8] . Travel restrictions have a long history , and date back at least to the 14th century , where people were prevented from leaving or entering specific communities during the plague epidemics [9] . Travel bans were also used in many cities and countries during the 1918-1919 influenza epidemic [9] . More recently , internal travel restrictions were used as a mitigation measure against influenza virus transmission during the 2009 pandemic in Mongolia , through interruption of provincial rail and road travel [10] , and during the recent Ebola outbreaks [11] . There has been some work on infectious disease spread and urbanisation focussing on the improved health conditions in urban areas compared to rural areas . For the developed countries , health has overall improved with increased urbanisation [12] . In low-income settings , health conditions are on average better in urban areas than in rural areas , but there are also significant challenges relating to inequities and heterogeneity in health among the urban population ( favelas , slums , etc . ) . High population density increases exposure to infectious diseases [12] . In Africa , the urban population has better nutritional status , fewer morbid events , increased vaccine coverage and better access to healthcare services compared to the rural population , and have reduced levels of malaria transmission and other severe diseases [13] . There is also one study , [14] , which develops a model for the effect of urbanisation on the transmission of infectious diseases , focussing on population growth and land use development . However , the infectious disease spread model is very simple . In addition , they do not consider the effect of interventions . The effectiveness of both vaccination and internal travel restrictions on mitigating an infectious disease have been studied in various settings , e . g . [9 , 10 , 15–20] . Germann et al . [15] study the spread of a hypothetical pandemic influenza , with a basic reproductive number in the range 1 . 6-2 . 4 , in the United States , and find that ( domestic ) travel restrictions do not have an effect on the final size of the epidemic , but might be able to slightly delay the time course . They also find that vaccination drastically reduces the final number of cases and delays the spread . Ferguson et al . [16] find that reducing long distance travel within the United States ( domestic air travel ) only slightly delays the influenza epidemic , for a hypothetical pandemic influenza strain with varying transmissibility . They consider vaccination in both the United Kingdom and the United States , and find that vaccination significantly reduces the final size of the epidemic . In accordance with these studies , the US Department of Health and Human Services also claims that vaccination is the most effective way of preventing the public health impact of ( pandemic ) influenza [21] . In a review study , Mateus et al . [17] find that domestic travel restrictions can delay the influenza epidemics by one week , and that extensive travel restrictions can reduce the final size of the epidemic . Both seasonal and pandemic influenza strains are considered . They also find that travel restrictions have minimal impact in urban centers with dense population and high mobility . Brownstein et al . [18] consider the influenza epidemic following the travel ban after 9/11 . They claim that the decrease in air traffic in the United States caused a delayed and prolonged influenza season ( however , note also the rebuttal in [22] ) . For a hypothetical influenza strain in Korea , it was found that 50% internal travel restrictions delayed the peak timing and had a slight reduction effect on the peak [19] . Some work has also been done on mathematical explanations and expressions for delay in epidemic spreading due to travel restrictions , for border control on international travel [23] , and for more general mobility networks [7 , 24] . During the 2009 influenza pandemic , uniform vaccination guidelines were given ( i . e . pro rata ) , for instance in Norway [25] , Ireland [26] and the United States [27 , 28] , and the default vaccination strategies are usually uniform [29] . However , the effectiveness of the different vaccination strategies and which strategy is best in terms of minimising attack rate , possibly depend on the population clustering of the underlying country . We thus simulate the epidemic with three different vaccination strategies—uniform vaccination , prioritising urban locations and prioritising rural locations . This allows us to compare the three vaccination strategies as a function of population clustering and provides us with a better understanding and possibilities for refined vaccination strategies . There are numerous examples of spatially targeted vaccination and antiviral strategies in the literature [29–39] . Some of these are based on dynamic optimisation strategies [29 , 31 , 33 , 39] , while others are based on prioritising locations with high prevalence ( i . e . geographically targeting hotspots ) [30 , 34] . In [38] , pro rata vaccination strategies are compared to vaccination strategies prioritising locations sequentially by population size ( among other vaccination strategies ) . This is similar to the vaccination strategy we investigate , targeting urban and/or rural locations . We investigate how domestic travel restrictions and vaccination affects the epidemic timing , spread and final size for the various levels of population clustering . For policy planners , the timing of peak dates and initial dates are important because they indicate how much time there is to implement interventions and preventive measures . If peak dates for spatially proximate regions are close in time , the efficiency of the health care organisation is challenged . This is extra problematic if there is in addition clustering of high peak incidences , since that would imply that spatially proximate locations have high disease activity at the same time . We further disentangle the effect of travel restrictions on final size of the epidemic by examining how the effect depends on how urban or rural the location is . We first describe the simulation set-up with details on the properties of the fictional country , the clustering algorithm , the disease dynamics model , the travel restrictions and vaccination strategies . We then use this tool to simulate the infectious disease spread for the various levels of clustering , investigate the effect of various amounts of internal travel restrictions , simulate and investigate the effect of the three vaccination strategies and finally a combination of travel restrictions and vaccination . We end with discussion and concluding remarks .
We construct a country , where the spatial clustering of population sizes is controlled by a design parameter . The country is a square , consisting of many small block units . The construction process consists of two steps . First , we generate the population sizes in the block units of the country by drawing a random sample from a reasonable population distribution . In the second step , we apply a clustering algorithm , generating different versions of the country with various levels of clustering . The clustering is done by rearranging the block units according to a mapping rule . The disease dynamics model is a metapopulation model which can be described as a network where every node represents a location , and every edge between locations represents people who travel between the locations and thus can spread the disease further . In every location there is a separate set of stochastic difference equations governing the local disease dynamics , but the processes are coupled through travellers . Similar disease dynamics models are used for instance in [45] and [46] for modelling the global spread of influenza-like illnesses . The following sections describe the two intervention measures examined in this work . They are applied both in separate simulations to isolate their effects , and in a combination strategy . The epidemics are compared by examination of final size , peak date , peak prevalence and the proportion of area that is not infected during the epidemic . The final size is defined as the total number who were infected during the epidemic . The peak date is the date with the highest number infected , and the peak prevalence is the proportion infected on the peak date . The area not infected is the proportion of block units where the prevalence was never larger than a threshold for seven consecutive days . The threshold is 1 . 0% , except from in the locations where the population size is less than 100 , then the threshold is one case . We compare quantitative properties to compare the effect sizes between different clustering levels in order to assess whether or not the clustering plays an important role . We are thus interested in whether there is a monotone relationship between such a quantity and κ , while the specific values are less interesting .
In order to compare the dynamics for the different levels of population clustering , the disease spread was simulated for the various clustering levels . It is intuitive that with a large enough amount of non-commuting travel , there will be sufficient mixing between all the block units in the country for the population structure not to play a role in the disease dynamics . The mean global prevalence curves for the different clustering levels are given in Fig 3 for the baseline scenario with no vaccination and no travel restrictions . The curves were visually very similar , but the higher clustering levels tended to have a higher and earlier peak . Note that the confidence bands were overlapping for most clustering levels . This indicates that with the baseline amount of non-commuting , the mixing was so high that the underlying population clustering seemed to have little effect on the disease dynamics . Slightly less area was infected for the higher clustering levels , and the final sizes were slightly smaller the more clustering ( Table 1 ) . The peak dates ( the day with the largest number of infected symptomatic individuals ) in the different block units are given in Fig 4 . The epidemic was able to spread throughout the whole country . We found spatial clustering in peak dates ( by visual inspection ) , and the spatial clustering was larger the more clustered the country in terms of population size . Three scenarios were considered: τ = 0 ( only commuting ) , τ = 1/100 ( 90% travel restrictions ) and τ = 1/1000 ( 99% travel restrictions ) . Table 2 shows the mean percentage of area not infected and the mean final size for τ = 0 . The final sizes of the epidemic decreased with increased clustering , and the area which did not experience the infection increased with increased clustering . The final size for the country with no clustering was 14% higher than for the most clustered ( κ = 3 . 0 ) . The global prevalence curves for the different clustering levels for τ = 0 are given in Fig 5a . The higher clustering levels experienced an earlier peak and a higher peak prevalence . Considering the confidence bands , these curves are significantly different—not between every clustering level , but the prevalence curves for the higher clustering levels are significantly different from the prevalence curves for the lower clustering levels . Interestingly , the higher peak prevalence did not imply a larger final size , as could be expected . Instead , the higher clustering levels had both higher peak prevalences and lower final sizes . In Fig 6 , we plotted the peak dates for τ = 0 . In addition , the initial date , peak prevalence and the probability of experiencing the epidemic in each location are given in S3 , S4 and S5 Figs . There was spatial clustering in both the initial dates , peak dates , peak incidences and probabilities of experiencing the epidemic , and the spatial clustering increased with increased ( population ) clustering levels . The more clustered the country , the fewer locations were infected on average ( Table 2 ) . In the peak dates plot in Fig 6 , we see that we had some locations which were not infected in any of the 100 simulations ( coloured white ) . For the higher κ , these non-infected locations were clustered , and the cluster sizes increased with κ . The epidemic did not spread throughout the whole country for the highest clustering levels , but seemed to be restricted to the highly populated area . The mean global prevalence curves for the settings with τ = 1/1000 and τ = 1/100 are given in Fig 5b and 5c , respectively . In the prevalence curves for τ = 1/1000 , we got similar results as for τ = 0 , with an earlier and sharper peak for the higher clustering levels . The same was found for τ = 1/100 , but the differences between the curves were smaller , as expected . For τ = 1/1000 , fewer locations were infected and the final size decreased with increased clustering , just as in the setting with τ = 0 ( cf . Table 3 ) . The same was found for τ = 1/100 in Table 4 , but the differences were smaller . The effect of travel restrictions on reducing the final size was larger with higher clustering . The τ = 0 , κ = 3 . 0 scenario had 21% reduced final size , while the τ = 0 , no clustering-scenario had a 9% reduction . For the 99% travel restrictions , the reductions were 19% for κ = 3 . 0 and 8% for the scenario without clustering . For the 90% travel restrictions , the corresponding reductions were 12% and 6% . The peak dates for τ = 1/1000 and τ = 1/100 are given in Figs 7 and 8 , respectively . Fig 7 shows that for τ = 1/1000 , there were some clusters of protected locations which did not experience the infection in any of the simulations . Comparing with the situation with τ = 0 in Fig 6 , we found that when adding the non-commuting travel , the infection was no longer trapped in the larger hubs for the higher levels of clustering , but was able to infect a larger area of the country . There were also some protected clusters of locations with τ = 1/100 ( as opposed to the baseline scenario ) . We have plotted the peak date for the mean global prevalence curve , peak prevalence , mean area not infected and mean final size for the various levels of clustering , for the baseline scenario and the travel restriction scenarios ( Fig 9 ) . The peak dates occurred earlier for increased levels of clustering , for all the travel ratios . In addition , the curves were very similar for the three travel restriction scenarios , while the peak dates occurred earlier for the baseline scenario . Hence , implementing travel restrictions delayed the epidemic peak . For the peak prevalence , we note that the higher levels of travel restrictions , the lower the peak . The decrease in peak prevalence was larger for the lower clustering levels . The decrease in peak prevalence for 99% travel restrictions compared to the baseline scenario was 38% for the highest clustering level and 48% for the lowest clustering level . There was almost no difference between the 99% travel ban scenario and the 100% travel ban scenario . The peak prevalence increased with increased clustering . The difference in peak prevalence with clustering was more prominent the more extensive the travel restrictions . The peak prevalence for κ = 3 . 0 was 20% higher than the peak prevalence for the “no clustering”-level for the complete travel ban scenario . For the mean area not infected , there was little difference between the 99% travel restrictions and the full travel ban setting . The more travel restrictions , the more area was protected . In addition , the amount of area which was protected increased with increased clustering , and the effect of clustering was stronger the more travel restrictions . For the final sizes , we found that the more travel restrictions , the lower the final size . In addition , as we have seen , for the travel restriction scenarios , the final size was lower for higher clustering levels . The more extensive the travel restrictions , the larger the difference between the various clustering levels . In the setting with only regular commuting , there were some clusters of protected locations for the highest clustering levels , and the epidemic seemed to be restricted to the hubs . It might therefore be a more effective use of vaccines to allocate all the resources to the most urban locations , since the more rural locations are more protected from the epidemic . However , a different strategy would be to preferentially allocate resources to exactly these rural locations , since the vaccination is more likely to successfully eliminate the risk in these locations . The global prevalence curves in the uniform ( pro rata ) vaccination setting are given in Fig 12a . The peak timing of the epidemic was similar for the different clustering levels , but there was a higher peak for the higher clustering levels ( however note that the confidence bands are overlapping ) . The mean area not infected and final sizes are given in Table 5 . Uniform vaccination reduced the final size substantially for all the clustering levels , compared to the baseline scenario and the travel restriction settings ( cf . Tables 1–4 ) . The reduction was slightly larger for the lower levels of clustering ( i . e . a 65-66% reduction in final size for the no clustering , κ = 0 . 1 and κ = 0 . 2 versions , compared to a 63-64% reduction for the κ ≥ 1 versions ) . The global prevalence curves for the urban vaccination strategy are given in Fig 12b . The prevalence curves were very similar to the uniform vaccination setting . The mean area not infected and mean final size are given in Table 6 . More area was infected in this setting compared to the uniform vaccination setting . The final sizes were slightly smaller in the urban vaccination setting compared to the uniform vaccination setting , for all clustering levels except κ = 3 . 0 . The effectiveness of this vaccination strategy compared to the uniform vaccination strategy did not seem to depend on the underlying population clustering of the country . The reduction in final size for the country without clustering was 67% while the reduction under the κ = 3 . 0 scenario was 63% . The global prevalence curves for the rural vaccination strategy are given in Fig 12c . Again , the peak timing seemed to be quite similar for all the clustering levels , with a higher peak for the higher clustering levels . The mean area not infected and mean final size are given in Table 7 . With the rural vaccination strategy , less area was infected , but the final size was a lot larger than for the urban and uniform vaccination strategies , due to both a higher peak prevalence and a longer epidemic . There was only a 42% reduction in final size for the κ = 3 . 0 version , and 44% for the least clustered version . Comparing the final sizes for the rural vaccination strategy and the urban vaccination strategy yields that the difference was larger for the lowest clustering levels . We have plotted the peak date for the mean global prevalence curve , peak prevalence , mean area not infected and mean final size for the various levels of clustering , for the baseline scenario and the different vaccination strategies . The plot is given in Fig 13 . The peak dates were slightly later for the lower clustering levels than for the higher clustering levels , for the baseline and the rural vaccination strategy . The peak date was similar for the urban and uniform vaccination strategy , while it occurred later for the rural vaccination strategy . All vaccination strategies reduced the peak . The reduction was larger for the uniform and urban vaccination strategies ( which were very similar ) , than for the rural vaccination strategy . The peak prevalence increased slightly with increased clustering under all vaccination schemes . The area infected seemed robust to the population clustering . For the final sizes , we clearly see a reduction with all the vaccination strategies , and the uniform and urban vaccination strategies were the most effective ( and very similar ) , while the rural was a lot less effective . In addition , the final size was quite robust to the underlying clustering , but there was a slightly larger final size for the higher clustering levels . In S1 Text , we repeated the vaccination simulations , where we assumed that the vaccinated non-immune individuals were 80% less infectious ( and not 20% as in the results presented here ) . The qualitative results were the same , and the final sizes were only slightly smaller for this more optimistic scenario . The urban vaccination strategy was the most effective in reducing final size for all clustering levels , and the uniform vaccination strategy performed similarly . The impact of the vaccination strategies were quite robust to the underlying population clustering . Since the travel restriction intervention resulted in more protected rural area the higher the clustering , we investigated whether the performance of the vaccination strategies in combination with travel restriction differed from the performance without any travel restrictions . The peak dates , peak prevalences , areas not infected and final sizes for the different strategies and clustering levels are given in Fig 14 . The urban and the uniform vaccination strategies were very similar , also when travel restrictions were included . The peak dates were slightly delayed ( only with a couple of days ) with increased clustering for the urban and the uniform vaccination strategies . Under the rural vaccination strategy , the peak dates occurred earlier for the higher clustering levels than for the lower clustering levels . For the area not infected , the different vaccination strategies were much more similar when travel restrictions were included , and the protected area increased compared to vaccination only or travel restrictions only . The combination strategy further reduced the final size and the peak prevalence for all vaccination strategies , but the qualitative relationship with population clustering was similar to the vaccination only setting . The urban vaccination strategy combined with travel restrictions reduced the final size with 87% for the lowest clustering level , and with 78% for the highest clustering level . For the uniform vaccination strategy , the corresponding reductions were 85% and 76% . Though they were very similar , the difference between the final size for the urban vaccination strategy and the uniform vaccination strategy was larger for the higher clustering levels . Hence for all clustering levels , there was a ( small and non-significant ) benefit in using the urban vaccination strategy instead of the uniform vaccination strategy , but the benefit was ( slightly ) larger for higher clustering levels .
In our main analysis , we have chosen to model commuting with a gravity law . The gravity law is a popular choice for modelling influenza/infectious disease spread [52] , and has been found to capture well spatio-temporal influenza [53] and measles [54] dynamics . An alternative to the gravity law is the radiation law [51] , which is parameter free , and therefore claimed to be more universal . Some analytical inconsistencies of the gravity law are also pointed out in [51] . The radiation law does not directly depend on distance , but rather on the population density between locations . The fact that the radiation law is parameter free is attractive in this setting , since we are modelling commuting in a fictional country and thus obviously do not have commuting data available for that country . However , the gravity law has been shown to have a better fit [55] , especially for finer scales , which is the case in our setting . For further information on the gravity law , the radiation law , extensions of these and other mobility models , we refer to the thorough review in [56] . We have chosen to use the gravity law in our main analysis since it had a better fit to our commuting data ( R2 = 0 . 82 versus R2 = 0 . 67 for the radiation law ) , and because it is commonly used in spatial models of infectious disease transmission . Regarding the shape of the gravity law , we have chosen to use a power function of distance , as in for instance [57] , [58] , [59] , [53] and [54] . Some commuting features that are not well captured by the gravity model are pointed out in [52] , and they suggest some extensions to the gravity law for solving these issues . However , for reasons of simplicity , we have chosen to work with the more standard model . In S1 Text , we perform sensitivity analysis on the parameters of the gravity law , the shape of the distance function , and we redo the analysis using a radiation law . The qualitative patterns are robust to the choice of commuting model . However , the different effect sizes varied with the shape of the commuting law . Comments on discrepancies in the effect sizes under the different scenarios are provided below . Our study is subject to limitations . The scaling from Norwegian municipalities to block units should be handled with care . The data that we used to fit the population size distribution and gravity law were on a different scale than the block units used in the fictional country , and we can not assess how well these models generalise to the finer block unit scale . In addition , different ways of dividing the population into administrative units could yield different population distributions , which could affect the results . We believe that it is of key importance to model the commuting and the population sizes for the same scale and population , and since commuting data are not available on a finer scale ( due to for instance privacy regulations ) , we have chosen to use the finest scale available for the commuting data . The block units can then be interpreted as a discretisation of administrative units . Census data are often collected for administrative units , but mobile phone data could be used as an information source for commuting on a square gridded fine scale resolution , and would have been an interesting alternative . The fine scale resolution of the fictional country is necessary for the clustering algorithm to provide smooth transitions between the different clustering levels . The fact that we consider the country in isolation and ignore bordering countries can result in edge effects , since the locations on the border might in reality have a connection to the neighbouring countries . We address some of these limitations in our sensitivity analysis in S1 Text . To handle the fact that different ways of defining administrative regions could affect the population distribution and in turn the results , we also perform the analysis on a country based on population data from the United Kingdom . The population size distribution for Norway was quite heterogeneous compared to other European countries examined , while United Kingdom had a more homogeneous population size distribution . Here , a gravity law fitted to commuting data for the United Kingdom is also used . We also perform sensitivity analysis on the parameters and shape of the gravity law . Though the qualitative patterns were not so sensitive to the choice of commuting model , the effect sizes were not robust to the shape of the commuting law . According to the general model developed here , population clustering is an important determinant for the effect of travel restrictions . For high levels of clustering , internal travel restrictions decrease the final size of the epidemic , and this is most prominent in the rural areas . There is no large effect on the peak date for the high clustering levels . For the lower clustering levels , there is less of a benefit in terms of final size reduction . However , for lower clustering levels , the peak date is delayed when implementing internal travel restrictions . That means more time to plan and implement interventions and preventive measures . Internal travel restrictions reduce the peak prevalence for all the clustering levels , reducing the stress on the health care systems . In addition , we found that the higher clustering levels have larger spatial clustering in peak dates . The more spatial clustering in peak dates , the more stress on the health care systems . Whether it is more attractive to delay an epidemic or decrease the final size , depends on the specific influenza strain . If there is high morbidity , such that the patients require more and/or longer health care , it might be of importance to delay the epidemic , to prepare the health care system for the peak period . In addition , if there exists vaccines or other interventions , a delay in the epidemic can be of key importance for the effectiveness of the vaccination programme , since there is more time to distribute ( and develop/improve ) the vaccine or implement other interventions . The impact of vaccines depends on how early they are introduced ( see for instance [25 , 33 , 62 , 63] ) . For the European countries we considered ( by visual inspection ) , the minimum clustering level was around κ = 0 . 5 , and the highest around κ = 3 . 0 , and they will likely become even more clustered in the future . This means that according to our model , internal travel restrictions are likely to be less and less effective in delaying epidemics , while they will be more effective in decreasing final sizes ( especially in rural areas ) . In addition , proximate regions will , to an even higher extent than today , experience their peak simultaneously . Hence , it will be even more important to be able to predict the peak timing , in order to prepare the health care system for the peak period . In addition , in order to minimise the final sizes of the epidemic , it is important not to neglect the urban locations for vaccination , and thus specific vaccination sentiment campaigns might target urban locations .
|
We study the interplay between urbanisation and infectious disease spread . As part of the worldwide urbanisation process , people are continuously moving to urban areas , and the cities are growing in size . This causes clusters of areas with high population density and clusters of areas with low population density , which is what we call population clustering . By simulating infectious disease spread in a synthetic country where we vary this population clustering , we explore the consequences of urbanisation on infectious disease spread . Our qualitative results have direct implications for infectious disease control guidelines and policies . We find that implementing internal travel restrictions have greater impact on the final number ill in the most urbanised countries than in the less urbanised countries . The effect is largest in the more rural parts of the country . According to our model , travel restrictions are more effective in delaying the epidemic in the less urbanised countries than in the more urbanised countries . We investigate vaccination strategies , where locations are targeted depending on how urban or rural they are . We find that it is important to vaccinate the urban locations—if the most urban locations are not covered by the vaccine , the final number ill will be a lot larger .
|
[
"Abstract",
"Introduction",
"Models",
"Results",
"Discussion"
] |
[
"medicine",
"and",
"health",
"sciences",
"influenza",
"immunology",
"geographical",
"locations",
"social",
"sciences",
"human",
"mobility",
"preventive",
"medicine",
"rural",
"areas",
"norway",
"vaccination",
"and",
"immunization",
"human",
"geography",
"public",
"and",
"occupational",
"health",
"infectious",
"diseases",
"geography",
"epidemiology",
"people",
"and",
"places",
"urban",
"areas",
"disease",
"dynamics",
"earth",
"sciences",
"geographic",
"areas",
"biology",
"and",
"life",
"sciences",
"viral",
"diseases",
"europe"
] |
2019
|
A theoretical single-parameter model for urbanisation to study infectious disease spread and interventions
|
Multivariate statistical techniques such as principal components analysis ( PCA ) and multidimensional scaling ( MDS ) have been widely used to summarize the structure of human genetic variation , often in easily visualized two-dimensional maps . Many recent studies have reported similarity between geographic maps of population locations and MDS or PCA maps of genetic variation inferred from single-nucleotide polymorphisms ( SNPs ) . However , this similarity has been evident primarily in a qualitative sense; and , because different multivariate techniques and marker sets have been used in different studies , it has not been possible to formally compare genetic variation datasets in terms of their levels of similarity with geography . In this study , using genome-wide SNP data from 128 populations worldwide , we perform a systematic analysis to quantitatively evaluate the similarity of genes and geography in different geographic regions . For each of a series of regions , we apply a Procrustes analysis approach to find an optimal transformation that maximizes the similarity between PCA maps of genetic variation and geographic maps of population locations . We consider examples in Europe , Sub-Saharan Africa , Asia , East Asia , and Central/South Asia , as well as in a worldwide sample , finding that significant similarity between genes and geography exists in general at different geographic levels . The similarity is highest in our examples for Asia and , once highly distinctive populations have been removed , Sub-Saharan Africa . Our results provide a quantitative assessment of the geographic structure of human genetic variation worldwide , supporting the view that geography plays a strong role in giving rise to human population structure .
The geographic structure of human genetic variation has long been of interest for its implications for studying human evolutionary history [1] , [2] , [3] , [4] , [5] . In recent years , the expansion of population-genetic datasets has contributed to an increase in geographic investigations of human genetic variation , often on the basis of classic multivariate statistical techniques such as PCA and MDS [6] , [7] , [8] , [9] , [10] . In PCA , samples are projected onto a series of orthogonal axes ( principal components or PCs ) that are constructed from a linear combination of genotypic values across genetic markers , such that each PC sequentially maximizes the variance among samples projected on it [11] , [12] . Classic MDS analyzes a genetic distance matrix between pairs of samples and places the samples in a low-dimensional space in such a way that pairwise Euclidean distances among samples in the low-dimensional space approximate their relative genetic distances [13] . The population structure of genetic variation is often summarized in easily visualized two-dimensional statistical maps obtained from the first two components of PCA or MDS . Especially for large-scale single-nucleotide polymorphism ( SNP ) data , PCA and MDS are popular because of their computational efficiency and high level of resolution in decomposing the complex structure of human genetic variation [12] , [14] . Generally , results produced by PCA and MDS are very similar to each other [15] . Several recent studies have reported detectable similarity between statistical maps of genetic variation and geographic maps of population locations . Such observations are particularly prominent within Europe , where striking similarity between genes and geography is observed both at a continental level [9] , [16] , [17] and in more localized studies such as in Finland [18] , [19] , Iceland [20] , and Sweden [21] . Analogous but visually less striking observations have also been reported in studies of other geographic regions , including in worldwide samples [6] , [7] , [8] , [10] , [22] , [23] and in samples from Asia [23] , [24] , [25] , Africa [26] , [27] , China [28] , [29] , and Japan [30] . However , this similarity of genes and geography is in many cases reported in a qualitative sense and has not been assessed systematically across different studies , so that it has been difficult to compare levels of agreement between genes and geography in different regions . Further , different studies have used different sets of genetic markers and different statistical techniques ( e . g . PCA and MDS ) , further complicating comparisons across datasets . Even for studies that used PCA , several versions of this technique have been employed in different studies . For example , some studies have performed PCA on genotypic matrices [9] , [10] , [12] , [20] , whereas others have applied PCA on pairwise genetic distance matrices [7] , [22] , [23] . A formal comparison of genes and geography in different regions using a single technique and a common marker set can provide a systematic basis for evaluating the role of geography in explaining the genetic similarity of individuals or populations in different locations . We have previously developed a Procrustes analysis approach to quantify the similarity between statistical maps of genetic variation and geographic maps [15] . This approach identifies data transformations that minimize the sum of squared Euclidean distances between two sets of coordinates while preserving relative pairwise distances among points within each set . The statistical significance of the similarity between genetic coordinates and geographic coordinates is then examined using a permutation test . In this study , we apply the Procrustes approach together with PCA to systematically study the geographic structure of human genetic variation across different geographic regions . By compiling data from a variety of published sources [9] , [23] , [26] , [31] , [32] , we have assembled genome-wide SNP data and geographic coordinates for 149 populations worldwide . Based on a common set of autosomal SNP markers shared by datasets collected from different studies , we evaluate the similarity between genes and geography in examples from Europe , Sub-Saharan Africa , Asia , East Asia , and Central/South Asia , as well as in a worldwide sample . We compare the level of similarity across the various datasets , finding that all show a high level of similarity , and that the highest similarity score appears in Asia . We also examine the dependence of the similarity on the choice of populations included in the analysis and on the number of markers studied . Our results provide information about the importance of geography in human evolutionary history , and can facilitate statistical methods for inferring the ancestral origin of human individuals from their genotypes .
Our worldwide example was based on 938 unrelated individuals from 53 worldwide populations ( Figure 1A ) , taken from the study of Li et al . [7] . None of these individuals was found to have 5% missing data or to appear as a PCA outlier . A PCA plot finds that as in previous studies [7] , [8] , [10] , samples from the same geographic region ( indicated by colors in Figure 1 ) generally cluster together , and that different clusters align on the PCA plot in a way that qualitatively resembles the geographic map of sampling locations . The first two PCs of our PCA explain 6 . 22% and 4 . 72% of the total genetic variation , respectively . These values are considerably less than the values reported by Li et al . [7] in their Figure S3B , which were 52 . 3% for PC1 and 27 . 8% for PC2 . The difference can be attributed primarily to the different versions of PCA used in the analyses . We applied PCA on the genotypic matrix for individuals and loci , whereas Li et al . applied PCA on an matrix recording levels of identity-by-state for pairs of individuals [7] . Although the two approaches provide visually similar PCA plots , the values and the interpretation of the proportions of variance explained by each PC differ , as they are based on quite distinct computations . Using Procrustes analysis , we identified an optimal alignment of the genetic coordinates to the ( Gall-Peters-projected ) geographic coordinates that involved a rotation of the PCA plot by 31 . 91 counterclockwise . The genetic coordinates were then superimposed on the geographic map by applying the optimal transformation , thereby highlighting the similarity between genes and geography ( Figure 1 ) . This qualitative resemblance is demonstrated by the Procrustes similarity score of , which is highly significant in 100 , 000 permutations ( ) . Applying the leave-one-out approach with populations excluded individually , the similarity score between genes and geography ranges from 0 . 697 to 0 . 715 , with mean 0 . 705 and standard deviation 0 . 003 ( Table S4 ) . Some populations , such as Native American and Oceanian populations , align in Figure 1B distantly from their geographic locations . In most but not all cases , excluding one of these populations leads to an increase in the Procrustes similarity score . Visually striking similarity betweeen PCA plots of genetic variation and a geographic map of Europe has been reported by several studies [9] , [16] , [17] . Our analysis was based on nearly the same sample studied by Novembre et al . [9] , containing 1 , 385 individuals from 37 populations widely spread across Europe ( Figure 2A ) . After excluding five individuals with 5% missing data and two PCA outliers , our final analysis examined 1 , 378 individuals . Our PCA plot is very similar to the plot of Novembre et al . [9] , with a close correspondence of genes and geography ( Figure 2B ) . One difference is that in the PCA plot of Novembre et al . [9] , individuals are more widely spread along PC2 than in our plot . As we applied PCA in the same way as Novembre et al . [9] , the difference arises primarily because they employed coordinates given directly by the eigenvectors in PCA , such that PC1 and PC2 were scaled to have the same variance ( J . Novembre , personal communication ) . To simplify the standardization of analyses across datasets , we chose not to scale the PC axes in our analyses , so that the relative amounts of genetic variation explained by each PC are reflected in the PCA plot ( see Materials and Methods ) . Our PC1 and PC2 explain 0 . 30% and 0 . 16% of the total genetic variation respectively , in close agreement with the values of 0 . 30% and 0 . 15% reported by Novembre et al . [9] . We used Procrustes analysis to superimpose the PCA plot on the geographic map , rotating the PCA coordinates 72 . 66 clockwise ( Figure 2 ) . The rotated genetic coordinates of the European samples are spread over a larger distance along longitudinal lines than along latitudinal lines , although the geographic locations of the samples are distributed in the opposite way . This observation reflects the result that the genetic differentiation among Europeans is larger in a north-south direction than in an east-west direction [34] . The Procrustes similarity between the genetic coordinates and the geographic coordinates is ( ) . Excluding populations from the analysis individually , the Procrustes similarity between genes and geography ranges from 0 . 764 for the analysis without the United Kingdom to 0 . 810 for the analysis without Italy , with a mean of 0 . 780 across populations and a standard deviation of 0 . 007 ( Table S5 ) . Populations that have a relatively large effect on the similarity score are mostly those with large sample sizes ( e . g . , Italy , Portugal , Spain and United Kingdom ) . The Russian population is an exception; its sample size is small ( ) , but the genetic coordinates of the Russian sample align poorly with the geographic coordinates [9] ( Figure 2 ) . Thus , this population has a relatively large effect on the similarity with geography ( when excluding Russians , Table S5 ) . Excluding Russians has minimal effect on the PCA coordinates for the remaining samples , however , as reflected in the high similarity score between the PCA coordinates before and after excluding the Russian sample ( , Table S5 ) . Reducing the sizes of large samples also has a relatively small impact; when repeating our analyses on a subset of the data in which 50 individuals are selected randomly from populations with larger samples , changes slightly to 0 . 777 , and both and the proportions of variance explained by PC1 and PC2 undergo slight increases ( Figure S1 ) . Sub-Saharan Africa is the location of the origin of modern humans and has the highest genetic variation among all continents [7] , [22] , [35] , [36] , [37] . Previous studies have found that when isolated hunter-gatherer populations are included in the analysis , PCA plots of genetic variation in Sub-Saharan Africa display low qualitative similarity to the geographic map of sampling locations [7] , [22] , [38] . Bryc et al . recently studied 12 populations in West Africa , and revealed a high similarity between a SNP-based PCA map and the corresponding geographic map , when Mbororo Fulani , a nomadic pastoralist population , was excluded from the analysis [26] . By integrating SNP data from multiple sources [23] , [26] , [31] , we investigated Sub-Saharan African populations in a broader region than in the analysis of Bryc et al . [26] . We first excluded four hunter-gatherer populations ( ! Kung , San , Biaka Pygmy , and Mbuti Pygmy ) and Mbororo Fulani . After further excluding six individuals with 5% missing data and two PCA outliers , our analyses examined 348 individuals from 23 populations in Sub-Saharan Africa ( Figure 3A ) . Applying PCA on this combined Sub-Saharan African dataset , we found that PC1 accounts for 1 . 34% of the total genetic variation , largely separating populations from west to east . PC2 accounts for 0 . 69% of the total genetic variation and largely separates populations from north to south ( Figure 3B ) . Generally , populations along the west coast of Africa cluster closely with each other , while interior populations form relatively isolated clusters . Bantu-speaking populations tend to cluster with each other , and can be divided into three groups according to their geographic locations: two populations in the west ( Fang and Kongo ) , two in the east ( Kenyan Bantus from the HGDP and Luhya ) , and five in the south ( Southern African Bantus from the HGDP , Nguni , Pedi , Sotho/Tswana , and Xhosa ) . Despite the large geographic separation among these three groups , their genetic separation in the PCA plot is relatively small ( Figure 3B ) . In particular , Luhya and Kenyan Bantus from the HGDP align between the western Bantu populations and the eastern non-Bantu populations such as Alur and Hema . The Maasai sample , consisting of 30 unrelated individuals randomly selected from the HapMap Phase III [31] , [33] , forms a cluster distant from the other populations along PC1 ( and PC3 , results not shown ) . Procrustes analysis identifies a rotation angle of 16 . 11 counterclockwise for the genetic coordinates ( Figure 3B ) , and the similarity score between genes and geography is ( ) . Among all populations , Maasai has the largest impact on both the PCA and Procrustes analysis ( Table S6 ) ; as shown in Figure S2 , when analyzed without Maasai , the other 22 populations align more closely with geography , and the Procrustes similarity score increases to 0 . 832 ( ) . Excluding any of the populations in South Africa leads to a decrease of the similarity between genes and geography , and the lowest similarity is obtained when excluding the combined Sotho/Tswana sample ( , Table S6 ) . This result suggests that the genetic map of Sub-Saharan Africans might look more similar to the geographic map if additional populations from the undersampled southern region of Africa were included . When hunter-gatherer populations ( ! Kung , San , Biaka Pygmy , and Mbuti Pygmy ) and Mbororo Fulani were included in the analysis , they appeared as isolated clusters on the PCA plots and greatly reduced the similarity between PCA maps and geographic maps ( Figure S3 , Table S7 ) . The similarity score decreased from 0 . 790 to 0 . 548 after including all five of these populations in the analysis . This value , however , is still statistically significant , with a -value of ; further , if we disregard the hunter-gatherer populations and Mbororo Fulani in Figure S3B and only examine the relative locations of the original 23 populations , we can still find a clear resemblance between genetic and geographic coordinates . Compared to the other 23 populations , the four hunter-gatherer populations appear as isolated groups at the south , and Mbororo Fulani appears at the north . These observations are clearer in plots with only one among the five outlier populations included at a time ( Figure S3C–S3G ) , each of which also produces significant similarity scores between genetic and geographic coordinates ( Figure S4 , Table S7 ) . Our Asian example included 760 individuals from 44 populations distributed widely across Asia ( Figure 4A ) . Previous studies based on largely overlapping datasets have reported correlations between genetic and geographic distances across Eurasia [22] , [23] . Our dataset combined data from these studies as well as from Li et al . [7] and Simonson et al . [32] , and after excluding 11 PCA outliers , our final dataset for Asia contains 749 individuals . In our PCA plot ( Figure 4B ) , PC1 largely separates populations on different sides of the Himalayas , accounting for genetic variation in an east-west direction . PC2 , on the other hand , distinguishes northern and southern populations . PC1 accounts for 5 . 42% of the total genetic variation , a much larger value than the 0 . 85% captured by PC2 , reflecting large genetic distances between populations separated by the Himalayas . Interestingly , populations around the Himalayas form a ring shape on the PCA plot , with the Nepalese population from the Himalaya region aligning in the middle . As noted by Xing et al . [23] , the Nepalese samples were collected from different subgroups that have different levels of ancestry shared with Central/South Asians and East Asians , and the dispersion of the Nepalese sample is therefore not unexpected . Tibetans , on the northern side of the Himalayas , do not spread over a large area in the plot and are well clustered with other East Asian populations . One interesting result concerns the Uygur and Kyrgyzstani populations , both of which lie along ancient trade routes between Europe and East Asia . Compared to the Uygur population , which lies farther to the east , the Kyrgyzstani population clusters closer to East Asian populations , especially to the Yakut and Buryat populations , supporting a view that the Kyrgyzstani group has a proportion of its ancestry in Siberia [39] . A third population sampled from near the Uygur and Kyrgyzstani populations is the Xibo population , which clusters clearly with East Asians from northeastern China . This pattern matches the expectation given documentation that this Xibo group moved in 1764 from northeastern China to Xinjiang province [40] , [41] . The PCA map of genetic variation in Asia is rotated 5 . 05 counterclockwise in the Procrustes superposition on the geographic map ( Figure 4B ) . Despite the discontinuity caused by the Himalayas , most populations align in a way that is highly concordant with their geographic locations . This observation is confirmed by a Procrustes similarity score of ( ) . Among all populations , the tribal population Irula , which appears south of India as an isolated cluster in Figure 4B , has the largest impact among all populations on the Procrustes similarity with geography ( Table S8 ) . When excluding Irula , the PCA map aligns more closely with geography , with the Procrustes similarity increasing to 0 . 871 ( , Figure S5 ) . This exclusion generates increased separation on the PCA map for some populations . For example , in Figure S5 , Iban from Sarawak is more clearly distinguished from other Southeast Asian populations . Overall , the similarity score between genes and geography in Asia is robust to the exclusion of any one population , with the lowest Procrustes similarity score of occurring when the Buryat population is excluded ( Table S8 ) . To further examine populations on either side of the Himalaya Moutains , we performed additional analyses of East Asia and Central/South Asia . We first considered the East Asian populations in our Asian example . This dataset consists of 341 individuals from 23 populations . After excluding seven PCA outliers , our analyses were based on 334 individuals from 23 East Asian populations ( Figure 5A ) . Individuals in this East Asian dataset generally align along a curve on the PCA plot . PC1 explains 1 . 58% of the total genetic variation and largely accounts for a north-south genetic gradient; PC2 explains 0 . 98% of the genetic variation and mainly separates two Siberian populations ( Buryat and Yakut ) and three Southeast Asian populations ( Cambodians , Iban , and Thai ) from the other East Asian populations ( Figure 5B ) . The Tibetan population is also separated by PC2 , but on the opposite side to the Siberians and Southeast Asians . Overall , PC1 largely matches geography in the north-south direction , and PC2 shows only a partial similarity to the east-west direction . The imperfect match between PCA coordinates and geography is reflected by a relatively low Procrustes similarity score of , which , however , is still statistically significant with . The optimal transformation rotates the PCA map 67 . 27 counterclockwise prior to superposition on the geographic map ( Figure 5B ) . Interestingly , excluding populations one at a time , we found that the PCA coordinates were reflected over PC1 when Procrustes-transformed to match the geographic coordinates if either the Iban , Tibetan , or Yakut population was excluded ( Figure S6 ) . Such abrupt changes of the Procrustes transformation are consistent with the fact that PC2 matches less closely with geography; a reflection over PC1 has a small effect on the similarity score . The Procrustes similarity score with geography can be substantially increased by excluding Japanese ( , ) ; other than the Japanese population , Iban , Thai , and Yakut have the largest effect on the similarity scores both with geography and with the original PCA ( Table S9 ) . Our last example focused on Central/South Asia , using an initial sample of 372 individuals from 18 populations . Ten individuals were excluded as PCA outliers , leaving 362 individuals from 18 populations for the final analysis ( Figure 6A ) . The first two components of the PCA anlaysis account for 1 . 59% and 1 . 31% of the total genetic variation , respectively . Overall , the PCA pattern for the separate anlaysis of Central/South Asian populations is similar to the pattern for the same set of populations in our analysis of all of Asia ( Figure 4 ) . After rotating the PCA coordinates 11 . 78 counterclockwise , we obtained a Procrustes similarity score of 0 . 737 ( ) when comparing PCA coordinates to geography ( Figure 6B ) . Most populations from Pakistan cluster closely on the first two PCs except for the Hazara population , which clusters with the Uygur population and aligns distantly from its sampling location . When excluding Hazara , the Procrustes similarity score to geography increases from 0 . 737 to , larger than for any other exclusion ( Table S10 ) . Excluding Irula has the second largest effect on the similarity score to geography , but more interestingly , this exclusion has the largest effect on the PCA coordinates ( smallest value for in Table S10 ) . A closer examination of the PCA results reveals that when Irula is excluded , the Kalash population in Pakistan is separated from the other Pakistani populations and appears as an isolated group in the north ( results not shown ) . This result accords with the identification of this isolated group as distinct in previous studies [8] , [36] . We have found that significant similarity between genes and geography exists in general at different geographic levels ( Table 2 ) . The highest similarity score was found in the data from Asia , followed by Sub-Saharan Africa when five outlier populations were excluded , and by Europe . Five of the six analyses had -values smaller than , and only the data from East Asia had a nonzero -value in 100 , 000 permutations . When comparing the permutation distributions of the similarity score ( Figure 7 ) , however , a difference in the significance levels is evident for the five examples with . The worldwide and Asian datasets have similarity scores considerably exceeding the similarity scores from all 100 , 000 permutations ( Figure 7A and 7D ) . By contrast , although the European , Sub-Saharan African , and Central/South Asian datasets have similarity scores higher than that of the worldwide dataset , their similarity scores are closer to the corresponding permutation distributions ( Figure 7B , 7C , and 7F ) , indicating relatively high -values compared to the worldwide data . To examine the robustness of our results to the number of SNPs analyzed , we repeated our analyses with subsets of randomly selected loci . We found that our Procrustes similarity scores between genes and geography are quite robust as long as enough SNPs ( 10 , 000 ) are used ( Figure 8 ) . Indeed , for the worldwide and Asian datasets , 1 , 000 SNPs are sufficient to obtain a similarity score between genes and geography close to the score obtained using all 32 , 991 SNPs . For the African , East Asian , and Central/South Asian datasets , the number of SNPs needed increases to 4 , 000 . Interestingly , many more SNPs are required for the European dataset to reach a high similarity score between genes and geography . Although the increase of the similarity score for the European dataset becomes slow when the number of SNPs exceeds 10 , 000 , it continues even when the number of SNPs is as high as 30 , 000 ( Figure 8 ) . If we use the same 197 , 146 SNPs as used by Novembre et al . [9] , the similarity score between genes and geography for the European example would become 0 . 799 , slightly higher than the value for our Sub-Saharan African example based on 32 , 991 SNPs . This larger number of SNPs required might reflect a relatively homogeneous population structure in Europe that requires more genetic markers to characterize subtle differentiation . To explore the relationship between genetic differentiation and the number of SNPs required to produce convergence in the Procrustes similarity , we computed across populations , a measurement of population differentiation , for all of our datasets , on the basis of the 32 , 991 autosomal SNP markers . We found for the European dataset , much smaller than the values of 9 . 704% and 4 . 706% for the worldwide and Asian datasets . The values of for the Sub-Saharan Africans ( without outlier populations ) , the East Asians , and the Central/South Asians are 1 . 334% , 1 . 874% and 2 . 140% , respectively . As expected , datasets that have less population differentiation , as indicated by smaller values , need more markers to reveal geographic structure in the PCA plot , consistent with a previous finding that the dataset size required for the population structure to be evident in PCA is inversely related to [12] . Further , we found and the sum of the proportions of variance explained by PC1 and PC2 to be positively correlated ( Pearson correlation , Figure 9 ) . This strong linear correlation is not surprising because of the connection between and the proportions of variance: can be computed as the proportion of the variance in an allelic indicator variable contributed by between-population differences [42] . It has been shown under a two-population model that the proportion of the total variance explained by PC1 is approximately equal to [43] . Here , we have observed a qualitatively similar relationship .
Both simulation-based and theoretical studies have shown that under spatial models in which migration and gene flow occur in a homogeneous manner over short distances , a similarity between PCA maps of genetic variation and geography is predicted [43] , [44] . In this study , we have systematically assessed this similarity in different geographic regions using a shared set of autosomal SNPs and a shared statistical approach . We have found that although they generally explain a relatively small proportion of the total genetic variation , the first two principal components in PCA often produce a map that resembles the geographic distribution of sampling locations . Our results quantitatively demonstrate the general existence in different geographic regions of a considerable similarity between genes and geography , supporting the view that geography , in the form of incremental migration and gene flow primarily with nearby neighbors , plays a strong role in producing human population structure . One particularly interesting observation concerns our analysis of the Asian dataset . Asia contains the Himalaya region , a strong geographic barrier to gene flow that has generated noticeable genetic differentiation between populations on opposite sides [45] . Such barrier effects can produce a distortion of PCA maps from those expected under homogeneous isolation-by-distance models [43] , [44] , leading to a decrease in the similarity to geography . However , although the concordance of a PCA plot with geography is perhaps best known for Europe — which does not have a barrier of comparable importance to the Himalayas — we obtained the unexpected result that in spite of the Himalaya barrier , the Procrustes similarity score was actually highest in Asia . When further examining the population structure on separate sides of the Himalayas , we found lower similarity scores between genes and geography in our East Asian and Central/South Asian samples . Especially for the East Asian sample , our results indicate weaker correlation between genes and geography in the east-west direction . To make the similarity scores between genes and geography commensurable for different datasets , we performed our analyses with the same markers and the same statistical approach . However , one aspect of the analysis that is not homogeneous across datasets is the nature of the geographic coordinates . For example , while most of the analyses employed population sampling locations , for the European dataset , coordinates did not necessarily represent sampling locations . Sampling locations may also vary in the extent to which they represent long-term locations where groups have resided . One example that highlights this issue is the Xibo population , which was sampled in northwestern China , but which clusters genetically with populations in northeastern China ( Figure 5 ) . This group is known to have migrated westward from near Shenyang in northeastern China about 250 years ago [40] , [41] , and if we were to use the coordinates of Shenyang ( N , E ) for Xibo rather than the sampling location , would increase from 0 . 640 to 0 . 654 for the East Asian dataset , from 0 . 849 to 0 . 859 for the Asian dataset , and from 0 . 705 to 0 . 709 for the worldwide dataset . Additional limitations apply to our geographic analysis . In all of the datasets , population-level rather than individual-level coordinates were used , so that all individuals from the same population were assigned to a single geographic location . This approach can potentially obscure substructure within populations . For example , although both the northern and southern Han Chinese groups from the HGDP dataset were assigned to the same location , they can be genetically distinguished from each other , with the northern group clustering closer to the northern populations in China ( Figure 5 ) . Use of individual-level coordinates might lead to higher values of the similarity score . Another concern is that the choice of a map projection ( including the projection that consists of using unprojected latitudes and longitudes as a rectangular coordinate system ) can have different effects in geographic regions at different distances from the equator , as the level of distortion of the surface of the earth varies with the choice of projection . This issue is expected to be of greatest concern in analyses at high latitudes or in datasets with a wide range of latitudes . We note that theoretical work and simulation studies have found that results from the PCA approach can be sensitive to the sample size distribution over geographic space [43] , [44] , [46] . In most of our analyses excluding one population at a time , patterns in PC1 and PC2 did not differ greatly from analyses in which all populations were included . However , exclusions of genetically distinctive populations , populations that were geographically distant from the center of a dataset , or populations with large sample sizes sometimes had sizeable effects on . In some analyses , particularly in considering the Luhya and Maasai populations from the HapMap , we therefore included only a subset of available individuals in order to reduce the influence of the large sample sizes for these populations . More generally , an analysis of the role of the geographic distribution of the sample can be performed by analysis of subsamples of a full dataset with different levels of geographic unevenness . A previous analysis of population structure inference using STRUCTURE for a variety of samples with different geographic distributions did not find a particularly strong role for the geographic dispersion of the sample [47] , but the issue has not yet been systematically investigated with PCA . Through a combination of PCA and Procrustes analysis , we have investigated genes and geography using the same standardized approach in different regions . The general observation of a concordance of genes and geography in different regions and at different geographic levels can provide a foundation for refinement of methods for inferring local geographic origin of human individuals from their genotypes [e . g . 9] , [19] , [48] . In addition , our computations illustrate the use of Procrustes analysis in assisting the interpretation of PCA , such as in comparing PCA maps to different types of spatial maps and in assessing the impact of certain populations or individuals on PCA results . Similar applications of PCA and Procrustes approaches can be used to evaluate evolutionary models by comparing PCA maps obtained from observed data to those obtained from simulated data generated by these models . With the incorporation of the Procrustes similarity score for quantifying patterns in PCA , results from PCA can potentially find new uses in additional applications in population-genetic studies .
We examined genome-wide SNP datasets previously reported in several studies [9] , [23] , [26] , [31] , [32] . The data of Pemberton et al . [31] merged unrelated samples from earlier datasets obtained from the HGDP [7] and HapMap Phase III [33] , [49] . Some of the data of Xing et al . [23] were previously reported in an earlier paper of Xing et al . [22] . Because the datasets were genotyped on different genotyping platforms , including Illumina 650 K [31] , Illumina Human 1 M [31] , Affymetrix 500 K [9] , [26] , Affymetrix NspI 250 K [23] , and Affymetrix 6 . 0 [23] , [31] , [32] , we identified a shared set of 32 , 991 autosomal SNPs included in all five datasets [9] , [23] , [26] , [31] , [32] . This number was smaller than the maximum possible set of overlapping SNPs shared among these genotyping platforms , because some SNPs were excluded during the quality control procedures of the studies that originally published the data [9] , [23] , [26] , [31] , [32] . At 6 , 549 among these 32 , 991 markers , the datasets from Novembre et al . [9] and Bryc et al . [26] had genotypes given for opposite strands when compared to the datasets of Xing et al . [23] , Pemberton et al . [31] , and Simonson et al . [32] . In these instances , we converted the genotypes from Novembre et al . [9] and Bryc et al . [26] to the opposite strand , so that genotypes were consistent across datasets from different sources . In total , we obtained genotype data on 32 , 991 autosomal SNPs for 4 , 257 samples from 149 populations worldwide , with dense sampling from Asia , Europe , and Sub-Saharan Africa . In our final dataset , the physical distance between pairs of nearby SNPs has mean 84 kb ( median 45 kb ) . We next created six datasets at different geographic scales , including a worldwide sample , continental samples for Europe , Sub-Saharan Africa , and Asia , and subcontinental samples from East Asia and Central/South Asia ( Figure S7 , Table 1 ) . For the worldwide example , we included 938 unrelated individuals from 53 populations in the HGDP [7] , [31] . For the European sample , we used a set of individuals that was nearly identical to that analyzed by Novembre et al . [9] , containing 1 , 385 individuals from 37 populations defined by ancestral origins . We did not include two French individuals ( sample ID 31645 and 32480 ) that were included by Novembre et al . [9] but that are not found in the release we obtained of the POPRES dataset in the NCBI dbGaP database [50] , [51] . For Sub-Saharan Africa , we integrated data on African populations from three sources [23] , [26] , [31] , including 30 unrelated Luhya ( LWK ) individuals and 30 unrelated Maasai ( MKK ) individuals , both randomly selected from the HapMap Phase III [31] . Because some populations in Sub-Saharan Africa are known to be genetically distinctive when compared to most other Sub-Saharan Africans [7] , [8] , [23] , [26] , [36] , [37] , we created two datasets for Sub-Saharan Africa , one including and the other excluding these distinctive populations ( ! Kung , San , Biaka Pygmy , Mbuti Pygmy , and Mbororo Fulani ) . When excluding all five of these populations , we have 356 individuals from 23 Sub-Saharan African populations . Including them , we have 422 individuals from 28 groups . Note that both Pygmy populations that we examined are from the HGDP [7] , [31] , and we did not include the Mbuti Pygmy data from Xing et al . [23] . Further , we also did not include the Luhya individuals from Xing et al . [23]; these individuals are a subset of those of the HapMap [31] , [33] . As in Xing et al . [23] , we analyzed three Sotho samples and five Tswana samples together as a single population , labeled as “Sotho/Tswana . ” Our sample from Asia has 760 individuals from 44 populations with sampling locations distributed widely across Asia . These data include 27 populations from the HGDP dataset [7] , [31] , 16 populations from Xing et al . [23] , and one population ( Tibetan ) from Simonson et al . [32] . For populations studied by both Pemberton et al . [31] and Xing et al . [23] ( Cambodian , Han Chinese , and Japanese ) , we only included the HGDP samples from Pemberton et al . [31] . Samples for East Asia and Central/South Asia are subsets of the Asian sample . The East Asian sample consists of 341 individuals from 23 populations: 18 populations from the HGDP dataset [7] , [31] , 4 populations from Xing et al . [23] , and the Tibetan population from Simonson et al . [32] . The Central/South Asian sample has 372 individuals from 18 populations in total , including 9 populations each from the HGDP dataset [7] , [31] and the Xing et al . dataset [23] . We applied two additional processing steps on each dataset to remove samples with high missing data rates and samples that appear to be outliers . First , we removed individuals with more than 5% missing data in the 32 , 991 SNPs . Next , in each analysis , we used an iterative PCA approach to identify and remove outlier individuals , as outliers can potentially distort PCA maps of genetic variation [52] . After applying PCA on a dataset , individuals greater than 10 standard deviations from the mean PC position on at least one of the top 10 PCs were considered outliers and were removed from the dataset . This procedure was repeated iteratively until no more outliers were detected . For all datasets , only a small proportion of samples were identified as outliers and removed by this procedure ( Table 1 ) . The data processing procedures are illustrated in Figures S7 , S8 , S9 , and are summarized in Table 1 . Individuals that were identified as PCA outliers are listed in Table S11 . We assigned all individuals from the same population to a single geographic location , as listed in Tables S1 , S2 , S3 . For the HGDP samples [31] , we used previously reported coordinates as the geographic locations for all populations ( Table 1 in [45] ) . The geographic locations for the European dataset were reported in Table S3 of Novembre et al . [9] , and represent countries of origin . The geographic coordinates for the African populations from Bryc et al . [26] are sampling locations , and we used the values reported by Tishkoff et al . [37] in their Table S1 . Geographic coordinates for populations from Xing et al . [23] were kindly provided by J . Xing . For the Tibetan samples , we used the sampling location reported by Simonson et al . [32] . For the two HapMap populations included in this study ( Luhya and Maasai ) , we used the sampling locations reported by HapMap [33] . We used longitude and latitude measured in degrees as our geographic coordinates for all datasets except the worldwide dataset . Latitudes in the southern hemisphere and longitudes in the western hemisphere were denoted by negative values . For the worldwide dataset , we shifted the Americas by adding to longitudes smaller than . We then used the Gall-Peters projection , an equal-area projection that preserves distance along the N parallel , to obtain rectangular coordinates as our geographic coordinates . For other datasets , we used unprojected longitude-latitude coordinates . We coded the genotype data for each dataset by an matrix , in which counts the number of copies of a reference allele at locus of individual , is the number of individuals , and is the number of loci . For autosomal SNPs , is 0 , 1 , 2 , or missing . We first ignored missing data and estimated the reference allele frequency among nonmissing genotypes , or . Following the smartpca program [12] , we standardized the nonmissing entries in by ( 1 ) where is a matrix with the same dimensions as . If a locus was monomorphic in a dataset ( or 1 ) , eq . 1 is undefined , and we set all entries in the column of for this locus to zero . Entries representing missing data were set to zero in as well . We performed PCA by applying the function eigen in R ( www . r-project . org ) to the matrix [43] . The coordinates of the individuals on the th PC are given by , where is the th eigenvalue of , sorted in decreasing order , and is the corresponding eigenvector . The proportion of variance explained by the th PC is calculated as , where is the total number of eigenvectors of . This quantity measures the variation among individuals along the th PC direction , relative to the total variance in the standardized genotypic matrix . In our examples , , and because has rank after standardization ( eq . 1 ) . We note that some studies have used the eigenvectors directly as PCs , so that all PCs have equal variance . We follow an alternative convention [43] , [53] , reporting PCs using , so that the proportions of variance explained by each PC are reflected on the PCA plot . In PCA plots superimposed on geographic maps , because horizontal and vertical axes are plotted on different scales , PC1 and PC2 can appear to not be perpendicular . We applied Procrustes analysis [13] , [15] to compare the individual-level coordinates of the first two components ( PC1 and PC2 ) in the PCA performed on the SNP data to the geographic coordinates . Procrustes analysis minimizes the sum of squared Euclidean distances between two sets of points ( two “maps” ) by transforming one set of points to optimally match the other set , while preserving the relative pairwise distances among all points within maps . Possible transformations include translation , scaling , rotation , and reflection . The similarity between two maps is then quantified by a Procrustes similarity statistic , in which is the minimum sum of squared Euclidean distances between the two maps across all possible transformations . , which is given by equation 6 in Wang et al . [15] , has been scaled to have minimum 0 and maximum 1 . The similarity statistic therefore also ranges from 0 to 1 . In our analyses , we fixed the geographic coordinates and Procrustes-transformed the PCA coordinates in order to superimpose the PCA maps on the geographic maps . In addition to , we also report the rotation angle of the PCA map as given by the Procrustes analysis , measured in degrees counterclockwise . To test the statistical significance of , we used a permutation test . In each permutation , we randomly permuted the population geographic locations , assigning all individuals from the same population to a single geographic location in the permuted dataset . We then applied Procrustes analysis to compute the similarity score between the PCA coordinates and the randomly permuted geographic coordinates . We calculated the -value as , representing the probability of observing a similarity statistic higher than under the null hypothesis that no geographic pattern exists in the population structure . For each dataset , we employed 100 , 000 permutations for the permutation test . We investigated the effect of each population on our PCA and Procrustes analysis using a leave-one-out approach . For each dataset , we excluded one population at a time and repeated PCA to obtain a new set of genetic coordinates ( for each population excluded , this PCA started from the same final set of individuals after exclusions owing to missing data and PCA outliers , and we did not repeat the search for outliers ) . We then performed two Procrustes analyses . In the first one , we compared the new PCA coordinates and the original PCA coordinates obtained before removing any population . This comparison was based on the common set of individuals included in both analyses , and its similarity score was denoted . In the second Procrustes analysis , we computed the similarity between the new set of PCA coordinates and the corresponding geographic coordinates , denoting the similarity score by . To investigate the effect of the number of markers on our results , we created a series of marker lists by randomly selecting loci from the 32 , 991 total loci . These marker lists were selected independently of each other and had . We then repeated PCA and Procrustes analysis for each geographic region using genotypes at the loci in each of our marker lists . For Sub-Saharan Africa , we used the dataset that excludes hunter-gatherer populations and the Mbororo Fulani . Given , the analyses for different geographic regions are based on the same set of markers , so that their results are comparable . We calculated in each dataset using Weir and Cockerham's estimator ( eq . 10 in [54] ) based on all 32 , 991 loci .
|
The spatial pattern of human genetic variation provides a basis for investigating the history of human migrations . Statistical techniques such as principal components analysis ( PCA ) and multidimensional scaling ( MDS ) have been used to summarize spatial patterns of genetic variation , typically by placing individuals on a two-dimensional map in such a way that pairwise Euclidean distances between individuals on the map approximately reflect corresponding genetic relationships . Although similarity between these statistical maps of genetic variation and the geographic maps of sampling locations is often observed , it has not been assessed systematically across different parts of the world . In this study , we combine genome-wide SNP data from more than 100 populations worldwide to perform a formal comparison between genes and geography in different regions . By examining a worldwide sample and samples from Europe , Sub-Saharan Africa , Asia , East Asia , and Central/South Asia , we find that significant similarity between genes and geography exists in general in different geographic regions and at different geographic levels . Surprisingly , the highest similarity is found in Asia , even though the geographic barrier of the Himalaya Mountains has created a discontinuity on the PCA map of genetic variation .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"genetic",
"polymorphism",
"genetics",
"population",
"genetics",
"biology",
"genetics",
"and",
"genomics"
] |
2012
|
A Quantitative Comparison of the Similarity between Genes and Geography in Worldwide Human Populations
|
The microbial conversion of solid cellulosic biomass to liquid biofuels may provide a renewable energy source for transportation fuels . Endophytes represent a promising group of organisms , as they are a mostly untapped reservoir of metabolic diversity . They are often able to degrade cellulose , and they can produce an extraordinary diversity of metabolites . The filamentous fungal endophyte Ascocoryne sarcoides was shown to produce potential-biofuel metabolites when grown on a cellulose-based medium; however , the genetic pathways needed for this production are unknown and the lack of genetic tools makes traditional reverse genetics difficult . We present the genomic characterization of A . sarcoides and use transcriptomic and metabolomic data to describe the genes involved in cellulose degradation and to provide hypotheses for the biofuel production pathways . In total , almost 80 biosynthetic clusters were identified , including several previously found only in plants . Additionally , many transcriptionally active regions outside of genes showed condition-specific expression , offering more evidence for the role of long non-coding RNA in gene regulation . This is one of the highest quality fungal genomes and , to our knowledge , the only thoroughly annotated and transcriptionally profiled fungal endophyte genome currently available . The analyses and datasets contribute to the study of cellulose degradation and biofuel production and provide the genomic foundation for the study of a model endophyte system .
Global climate change and decreasing fuel reserves are driving a push towards biologically derived fuels from plant wastes . The optimal biofuel for immediate implementation is one that functions within the context of current infrastructure , in particular with existing engines and distribution systems . This would require chemical similarity to gasoline , which is a mixture of hydrocarbons with an average chain length of eight [1] . Fungi have been recognized as producers of eight carbon ( C8 ) volatiles for nearly 80 years and are a major global carbon recycler [2] , [3]; however , despite the interest in these compounds , the genes responsible for their production remain largely undefined . One such producer of C8 volatiles is the endophyte Ascocoryne sarcoides ( NRRL 50072 ) . Originally identified as Gliocladium roseum , this organism was shown to produce a series of molecules of potential interest as biofuels when grown on a cellulose-based medium [4] . The taxonomy was later revised to A . sarcoides and its production profile of Volatile Organic Compounds ( VOCs ) was amended to remove branched-chain alkanes . However , this follow-up work also confirmed the production of straight-chain alkanes from C6 to C9 , as well as branched-chain alcohols varying in length from C3 ( 2-methyl-1-propanol ) to C7 ( 5-methyl-1-hexanol ) ( Table S1 ) [5]–[7] . Understanding and optimizing biological production of such molecules is an area of active research ( reviewed in [8] ) . Bacteria have been shown to produce alkenes through “head-to-head” condensation of fatty acids; however , products with fewer than 23 carbons , like those from A . sarcoides , are not known to be synthesized by this mechanism [9] , [10] . Odd-chain alkanes and alkenes of chain lengths 13–19 have been observed in bacteria as products of the decarbonylation of aldehydes and the decarboxylation of fatty acids , respectively [11] , [12] . However , currently there are no known eukaryotic homologs for these enzymes . C8 alcohols and ketones have been identified as the products of linoleic acid breakdown; however , the genes responsible for the downstream reductions that generate C8 alkenes and alkanes are still unknown [13]–[17] . In order to gain a better perspective on these pathways and the cellulolytic machinery used by an endophyte , we coupled genome sequencing and short and long RNA-seq with metabolomic profiling of A . sarcoides . Generation of metabolic pathway predictions in organisms for which genetic tools have not yet been developed remains a difficult problem . Techniques such as gene expression analyses and metabolomics profiling have the advantage that genetic tractability is not required . In a pioneering study , Askenazi et al , showed that gene expression could be linked to specific metabolite production [18] . The authors profiled the level of lovastatin production in engineered strains of the fungus Aspergillus terreus and showed that strains with similar transcriptional profiles also had similar amounts of lovastatin production [18] . Furthermore , extensive metabolic network analyses have demonstrated the ability to link the transcription of individual genes to metabolites [19] , [20] . Metabolite-transcriptional coupling has since been validated extensively for the monitoring of different stress responses [21]–[23] . We used RNA-seq based gene expression measurements to accurately map gene structures and to generate candidate gene lists for novel metabolic pathways . In particular , we used gene expression and the co-occurrence of a compound across multiple experimental perturbations to generate candidate genes and pathways for the production of C8 volatiles and several other alkanes and alkenes that currently have no known eukaryotic pathway . In addition , we extensively mapped and annotated the A . sarcoides cellulose breakdown machinery using RNA-seq expression analysis after growth on different carbon substrates . Together with the high quality genome assembly and annotation , these data provide the most complete genomic characterization of any fungal endophyte to date . The analyses and datasets contribute to the development of biofuels from microbial metabolites and the related study of cellulose degradation and may be a reservoir of information for studying the plant-endophyte relationship .
The A . sarcoides NRRL 50072 genome was sequenced resulting in approximately 38-fold coverage of the estimated 34 Mb genome [24] . Reads were assembled into 16 scaffolds incorporating 99 . 5% of the total genomic base pairs . The genome size and overall GC content ( 45% ) is within the average range for other Leotiomycetes fungi [25] . We predicted 10 , 831 genes resulting in 100% recovery of annotated Core Eukaryotic Genes Mapping Approach ( CEGMA ) genes which is a benchmark for a high quality genome assembly ( Text S1 ) [26] . Roughly 70% of the gene models had at least one match to one of the 42 available fully sequenced fungal genomes . Approximately 22% of the gene models are seemingly species-specific and did not match to anything currently in GenBank [27]; the remaining 8% were homologous to genes outside of the fungal kingdom . Eighty-seven percent of the gene models were validated with long-read transcriptome profiling ( Text S1 ) and 75% of the potential exon-exon junctions were confirmed ( see Figure 1A and 1B ) . Although a subset of the unvalidated gene models and exon junctions may be spurious , the majority are most likely true genes that are silent under these specific conditions [28] , [29] . We subjected A . sarcoides to seven different growth conditions to assay diversity in both transcription and compound production ( Table S2 ) . Volatile metabolite production was analyzed by gas chromatography mass spectrometry ( GC/MS ) for six of these seven conditions ( no GC/MS dataset was obtainable on the day 9 potato dextrose harvest; Table S1 and S2 ) . We monitored A . sarcoides cultures for production of volatiles and selected this subset of six conditions for RNA-seq analysis , which provided differential compound production profiles . Under these six conditions , A . sarcoides produced 48 identifiable volatile metabolites including 18 alcohols and 7 alkanes/alkenes including heptane , octane , and nonane . All volatile metabolites were scored with a binary scale to indicate their presence or absence in each culture headspace . We chose this digitized scoring because different analyses required variation in culture and headspace volumes and our method of detection of VOCs ( Solid Phase Micro Extraction ( SPME ) , see Materials and Methods ) is sensitive to such variation [30] . The large number of functionally diverse metabolites in the headspaces also precluded the use of external or internal standards to determine the absolute amount detected for each compound across all conditions . Coupled transcriptional profiles for the six conditions obtained via RNA-seq resulted in more than 200 million reads alignable to the reference genome or exon junctions ( Table S3 ) and greater than 99% similarity between the two technical replicates ( Figures S1 and S2 ) . Six diverse sampling conditions were chosen for the RNA-seq analysis in lieu of replicates in order to more thoroughly explore the transcriptional landscape of A . sarcoides and more completely map gene structure throughout the genome . The genome and transcriptome data can be accessed at http://asco . gersteinlab . org . In addition to the 10 , 831 gene models predicted , we identified a number of RNA-seq reads which map outside of the gene models . A subset of these reads formed well-defined regions on the reference genome . 602 of these regions are at least 1 kb away from any annotated genes and are designated as transcriptionally active regions ( TARs ) ( Figure 1C , Figures S3 and for sensitivity analysis and examples ) . These TARs were seemingly devoid of open reading frames and in some cases were quite long ( up to 3 . 7 kb in length ) . Forty percent of these TARs illustrated condition-specificity ( standard deviation greater than 1; see Figure 1D–1F ) as has previously been observed in S . cerevisiae and H . sapiens [31] . The importance of these polyadenylated non-coding RNAs in regulating gene expression has only recently been discovered [31] , [32] and their exact role remains an active area of research . Given the emphasis on cellulose breakdown and utilization for the development of alternative fuels , we were interested in exploring and annotating the cellulolytic capabilities of A . sarcoides . We analyzed the transcription profiles of A . sarcoides for growth on three different carbon sources: cellulose ( CELL ) , cellobiose ( CB ) , and potato dextrose ( PD4 ) . While cellulose and cellobiose share the same β ( 1–4 ) linkage between monomer units , potato dextrose contains predominantly glucose-monomer . Differential gene expression between the potato dextrose and the two more complex substrates ( CELL and CB ) provides information on the pathways and mechanisms of cellulose breakdown; whereas , differences between the CELL and CB provides information on the genes necessary to utilize a soluble versus an insoluble polymer . Such differences are particularly useful as they can inform methods aimed at increasing cellulose breakdown efficiency . We first examined the differential expression across these three conditions ( Table S15 ) [29] , [33] . There were 1 , 435 genes that were expressed under all three conditions ( Figure 2A ) . A smaller number , 142 genes , were only expressed during growth on cellulose or cellobiose , including the endo- and exo-cellulases , as expected based on their role in cellulose utilization . 398 and 380 genes were exclusively expressed on cellobiose and cellulose , respectively , reflecting the significant differences in the machinery necessary to utilize a soluble disaccharide versus an insoluble polymer and in the resulting downstream changes in the cellular state . We focused on the subset of genes with homologs in the CAZY database , a manually curated repository for carbohydrate metabolism ( see Text S1 ) [34] . In total , 52% ( 89 of 169 ) of glycosyl-hydrolase homologs ( GH ) , 45% ( 25 of 56 ) of glycosyl-transferases ( GT ) , 50% ( 3 of 6 ) of carbohydrate-binding module genes ( CBM ) , 41% ( 9 of 22 ) of carbohydrate esterases ( CE ) , and 0% ( 0 of 1 ) of polylyase ( PL ) were differentially expressed across the three conditions ( Figure 2B; Table S4 ) . The most highly expressed gene in the cellulose condition was AS6577 , which is homologous to the gene encoding the protein swollenin . Swollenin was first identified in the cellulolytic model organism , Trichoderma reesei . Heterologous expression in yeast and Aspergillus niger showed that swollenin mediates disruption of plant cell walls without releasing monomeric sugars [35] . Supplementation of a cellulase mixture with swollenin increased saccharification rates suggesting this protein may play an important role in efficient cellulose breakdown [36] . While A . sarcoides growth on a lignin-containing medium was not analyzed , we identified the full pathways for 5-carbon sugar utilization e . g . arabinose and xylose , sugars which comprise 10–25% of carbohydrates resulting from hemi-cellulolysis [37] . We further validated the presence of these pathways by demonstrating A . sarcoides growth on media with either xylose or arabinose as the sole carbon source ( Materials and Methods ) . The genes responsible for both cellulose degradation and the production of secondary metabolites are non-randomly distributed in a number of sequenced genomes , such that they are clustered into regions of higher than average gene density [38] , [39] . Therefore , we searched for clusters in A . sarcoides as a strategy to identify genes involved in these processes . We generated a simulated set of scaffolds where the number of genes was kept constant but the placement was randomized to identify regions of the genome with higher than expected gene density . We identified 77 clusters ranging in length from 10–72 kb ( p< . 05 , Text S1 ) . Twenty-six clusters contained genes or domains known to be involved in secondary metabolism , particularly oxidoreductases and permeases . We noted five gene-clusters that were involved in the production of secondary metabolites usually restricted to plants , including two clusters containing genes homologous to those involved in the synthesis of patatin ( Table S5 ) . Patatin is a plant storage glycoprotein implicated in plant-fungal communication [40] . Expression of this protein in Arabidopsis negatively affects resistance to Botrytis cinerea and Pseudomonas syringae , but it increases resistance to the cucumber mosaic virus [40] . Interestingly , all genes in this cluster were transcriptionally silent under the conditions we tested . Given their known functional role in mediating plant-fungal interactions , it is possible they are strictly regulated by interactions with the plant host . The classes of genes most frequently involved in secondary metabolite production are Polyketide synthases ( PKS ) and Non-ribosomal peptide synthetases ( NRPS ) . We identified 19 PKS and NRPS clusters through fungal-specific Hidden Markov Models of beta ketoacyl synthase ( KS ) and acyltransferase ( AT ) domains and an additional 8 gene clusters and 11 gene models composed solely of enoyl reductase and/or dehydratase accessory domains ( Text S1 , Tables S6 and S7 ) . The identified PKS genes ranged in size from a few kb to the 13 kb and 13-exon hybrid PKS/NRPS AS8071 , which is by far the largest predicted gene model in A . sarcoides . Examination of the 3 kb region upstream and downstream of each PKS element also revealed a number of major facilitator superfamily transporters and permeases which may confer resistance to both PKS-derived and exogenous toxins [41] . However , comprehensive searches of previously identified PKS clusters [42] , laeA element identification to delineate possible cluster boundaries [43] , and use of domain to structure software [44] failed to yield any predictions for possible biosynthetic products . Intriguingly , one PKS , AS1082 was first found to contain a beta ketoacyl synthase domain , but subsequent searches revealed that it contained two distinct KS domains and an acyl carrier protein domain . However , no acyl transferase domain , which typically functions in substrate loading , was identified . While separately encoded acyl transferase enzymes that act in trans have been found in bacteria , only trans-acting enoyl reductase domains have yet been characterized in fungi [45] . A more direct method to investigate the A . sarcoides genes responsible for production of the novel metabolites is the use of association analysis . As mentioned above , the concordance of gene expression and metabolite production can be used to guide prediction of genes involved in metabolic pathways [18] . A complication in the application of these methods for novel metabolic pathways , as opposed to those generated either via PKS or as part of conserved metabolism , is that we know neither the genes that are involved nor the pathway structure ( i . e . the reactant-product pairings that result in the downstream compound ) . For example , we do not know the genes responsible for the production of octane , nor do we definitively know the starting compound or what intermediates may have been subsequently generated . Thus , we need a series of analyses that simultaneously infer the potential genes and the pathway trajectory as defined by the chemical elements ( Figure 3A–3D; Figures S5 , S6 , S7 ) . It was previously shown that by examining the “correlation” and “anti-correlation” of sets of genes across a wide spread of phylogenetic space , the importance , ordering , operons , and additional members of the pathway can be discerned [46]–[50] . Furthermore , genes belonging to the same pathway or complex often show both coordinated regulation and conservation [50] . By substituting the phylogenetic profiles from these previous studies with our compound profiles generated from compound presence or absence across all conditions , the resulting character matrix can be used to determine the relatedness of these compounds ( Figure 3B and 3F , Figures S7 and S8 ) . On the basis of these relationships , compounds can be then grouped into pathways . To apply this correlation analysis , each metabolite produced by A . sarcoides under each of the six growth conditions was assigned a “1” if it was detected in the particular condition and “0” if it was not detected , as depicted in the schema in Figure 3A–3B and 3E–3F ) . To further inform the metabolite analysis , we also used a recent meta-analysis that profiled the production of 10 Ascocoryne isolates under varying growth conditions resulting in 20 different GC/MS profiles [5] . Compounds that consistently co-occurred across the genus are more likely to be in the same pathway and were given more weight than those showing inconsistent behavior ( Figure S7 ) . We then grouped sets of compounds that co-occurred into single or related sets of pathways ( Figure 3C , compounds A and C ) and those that rarely or never co-occurred into different pathways ( Figure 3C , compound B ) . To identify possible metabolite-gene linkages , we then computed the correlation between the compound profile and expression of each gene under the different conditions . Correlations between compounds and expression were used instead of strictly quantitative changes in gene expression because this more effectively integrated the expression analysis with the binary compound production data . To ensure the correlations were significant , we computed a p-value for the compound co-expression scores ( See Text S1 , Tables S8 , S9 , S10 , S11 , S12 , S13 ) . For a set of compounds with the same compound profile , there may be many genes with correlated expression , not only those involved in the compound production . Therefore , retrosynthesis was used to disambiguate which of the significantly correlated genes were most likely involved in the production of those compounds ( Figure S8 ) . As one example of this method identifying candidate genes , we identified 60 genes with homology to putative alcohol dehydrogenases ( EC 1 . 1 . 1 . 1 ) , which have a wide range of specificities and annotation quality . However , only three of the identified alcohol dehydrogenases were significantly co-expressed with any compound production profile . In particular , AS5307 was co-expressed with the compound profile that had a predominance of branched medium chain alcohols , including 3-methyl butanol , 3-methyl-3-buten-1-ol , and 2-methyl-1-propanol . We predict that these three dehydrogenases , from amongst the 60 , play a key role in the production of the observed medium-chain alcohol metabolites . Co-expression has been used to assign functions to genes with known homologs as well as to genes without primary sequence or domain level annotations [51] . All genes co-expressed with a particular compound profile were examined as shown in Figure 4 , where each line represents a single gene . A subset of the genes was homologous to well-known secondary metabolite pathway elements , but some had no known function ( Figures S9 , S10 , S11 ) . In the latter cases , gene co-expression was used to infer additional pathway elements as well as associated regulators and transporters . Below , we provide an example set of predictions for a C8 product pathway . The full set of predicted pathway schemas and potential enzymes are provided in the supplement . An R package containing the code and documentation for RNA-seq processing and the association analysis is provided in Text S2 . Given the average chain length is about eight for hydrocarbons in gasoline , the production of molecules with similar lengths represents an obvious starting point for next generation biofuels that will be compatible with pre-existing infrastructure [52] , [53] . We identified candidate pathway elements for the production of reduced C8 volatiles in A . sarcoides and assigned correlated genes to each step of the reconstructed C8 pathway ( Figure 3 ) . As an example , lipoxygenases ( EC 1 . 13 . 12 . 12 ) are known to be involved in the formation of C8 alcohols and ketones in fungi via the catabolism of linoleic acid [3] , [54] . There are five lipoxygenases in the A . sarcoides genome , and two of these are correlated with C8 production ( AS2804 and AS3405 , Figure 3H , II ) . The most strongly correlated lipoxygenase , AS2804 is homologous to the Aspergillus gene ppoC ( Figure 3H , II ) ( p< . 01 ) . Recently , Brodhun et al showed that expression of ppoC is sufficient to catalyze the breakdown of linoleic acid into a wide range of compounds including: 1-octen-3-ol , 2-octen-1-ol , 2-octenal , and 3-octanone in a crude E . coli lysate [17] . All of these compounds were observed as products of A . sarcoides with the exception of 2-octenal ( Table S1 ) . The original hypotheses for the production of these C8 volatiles from linoleic acid involved two enzymes , a lipoxygenase to form a peroxidated intermediate , and a lyase ( EC 4 . 1 . 2 . - ) to catalyze its breakdown into smaller , volatile products . However , an active lyase has yet to be successfully purified in fungi [14]–[16] , [55] , and recent work argues against the need for this activity [17] . We identified one lyase , AS9537; however , its expression did not correlate with the production of C8 volatiles ( Figure 3G , III ) , arguing against the dual-enzyme hypothesis for C8 production and supporting the more central role for the lipooxygenase ( AS2804 ) . In addition to the oxygenated C8 volatiles observed by Brodhun et al . from Aspergillus , A . sarcoides produces the reduced compounds 1 , 3-octadiene; 1 , 3-trans-5-cis-octatriene; 1 , 5-octadien-3-ol; 1-octene; and 3-octanol suggesting that downstream processing of linoleic acid breakdown products has occurred . One potential route to these compounds is shown in Figure 3G , whereby 1-octen-3-one is further reduced to 1-octen-3-ol by FabG ( EC 1 . 1 . 1 . 100 ) , a 3-oxoacyl-[acyl-carrier protein] reductase ( Figure 3G , IV ) . In total , A . sarcoides has 10 genes with strong homology to FabG ( Table S14 ) , however , only the three co-expressed with C8 production are shown in Figure 3H , IV ( AS1593 , AS4820 , and AS5565 , with AS5565 exhibiting the largest expression change ) . The nearest sequenced relatives of A . sarcoides , Botryotinia fuckeliana , and Sclerotinia sclerotiorum , have only two and three FabG genes , respectively ( Figures S13 and S14 ) . Since the reduced C8 compounds have not previously been found outside the Ascocoryne genus and the expression of some FabG genes do correlate with these compound production profiles , it is possible that at least some of these additional FabG genes may participate in the reduction of eight carbon volatile compounds . In addition to the FabG homologs , 317 oxidoreductases particularly aldo-keto reductases , were identified in A . sarcoides . Of these 11 were correlated with C8 production ( Table S14 ) . Oxidreductases are able to reduce various functional groups , such as ketones and alcohols , and are expected to participate in the biosynthesis of the C8 reduced products and other volatiles ( Figure 3G , 3H; IV and VI ) . In addition , of all sequenced fungal genomes , only A . fumigatus ( 626 ) and T . reesei ( 494 ) have a commensurate number . Both B . fuckeliana and S . sclerotiorum have less than 200 oxidoreductases , which is approximately the median number for sequenced fungi . The above average number of oxidoreductases found within A . sarcoides suggests a large reducing capability and extensive secondary metabolism potential .
The unknown pathway for the production of potential biofuel compounds in A . sarcoides is part of a more general trend . Microorganisms produce an extraordinary diversity of natural products that have the potential to be used in numerous applications from medicines to biofuels to commodity chemicals [37] , [56] , [57] . However , identifying the genes responsible for their production remains a major hurdle for organisms that are not genetically tractable . Despite promising developments in pathway prediction algorithms , a substantial gap remains between metabolic capabilities and genetic characterization [58]–[61] . As an example , Metacyc , a repository of metabolic pathways , contains 8 , 869 compounds linked to 1 , 908 known pathways , but this represents less than 1% of the compounds estimated to be produced by micro-organisms [62] , [63] . An integrated omics approach could provide a relatively simple means of exploring the biosynthetic potential of uncharacterized non-model organisms . By examining changes in the A . sarcoides transcriptome across a diverse array of conditions , we were able to explore a wide fraction of genes and refine gene and exon boundaries to improve the genome annotation quality . Additionally , with co-expression patterns we generated hypotheses for the genes involved in undefined metabolic pathways and regulatory mechanisms . Through TAR building we identified a number of long , highly expressed regions seemingly devoid of open reading frames that may have a regulatory role . The recovery of 100% of all CEGMA [26] genes suggests a high quality genome assembly , and the number of scaffolds is on par with the number of expected chromosomes in Ascomycete fungi [64] . We used an expanded version of association analysis to generate hypotheses for products from unknown pathways . Such methods are flexible enough to integrate coupled transcriptome and metabolomics data and will take on increasing importance as the throughput of both transcriptome and metabolomics continues to increase . The means to leverage these datasets will be key to our understanding of novel metabolite production particularly for genetically intractable organisms . From its plant mediators to its oxidoreductases and its cellulases , A . sarcoides's gene complement represents several avenues for further research and its diverse array of enzymatic capabilities will contribute to the study of cellulose degradation and secondary metabolite production .
Isolate NRRL 50072 was obtained under a material transfer agreement from Montana State University ( GA Strobel , Bozeman , MT ) . Genomic DNA was isolated using the Plant DNeasy MaxiPrep kit ( Qiagen ) according to the manufacturer's instructions with the following modifications: mycelia were grown in potato dextrose broth for approximately 3 weeks at 25°C , shaking at 150 rpm and were harvested via filtration . The filtrate ( 1 g ) was homogenized by mortar and pestle under liquid nitrogen before the addition of 80 µL RNase ( 100 mg/mL ) , 80 µL proteinase K ( 10 mg/mL ) and lysis buffer P1 ( Qiagen ) . Homogenized material was heated for 10 min at 65°C and then processed through the remainder of the Qiagen protocol . Please see Table S2 for detailed growth and inoculation conditions for CB , PD4 , and PD14 as distinguished by the short code referred to in both the text and figure legends . For the remaining 3 conditions ( OAC , AMM , and CELL ) , media were prepared and inoculated with 50 mg filtered culture ( 1× PD ) as reported in Griffin et al . , 2010 [5] . Carbon starvation ( OAC ) was prepared as a minimal medium base with sodium acetate ( 50 mM ) as the sole carbon source . Nitrogen starvation ( AMM ) was prepared as a minimal medium base with no ammonium chloride and with 83 . 3 mM glucose . Cellulose substrate ( CELL ) was prepared as a minimal medium base with cellulose ( 15 g/L ) as the sole carbon source . All were titrated to a pH to 6 . 0 with NaOH . Vials were incubated for 2 days at 23°C before GC/MS analysis and RNA extraction . For each of these conditions , seven vials were inoculated , with three subjected to GC/MS analysis while the remaining four vials were concurrently used for RNA harvesting . Total RNA was isolated using the Ambion RiboPure Kit ( California , USA ) , and then poly-A purified and prepared for sequencing as in Nagalakshmi et al . , 2008 [28] . Sample PD9 was selected for RNA preparation and long-read transcriptomics , which was used to confirm gene models . RNA was extracted from a 9-day old 1 L PDB culture grown at 23°C , 150 rpm ( Table S2 ) . Extraction performed as in Nagalakshmi et al . , 2008 [28] . All conditions were as specified under the RNA-seq preparation . GC/MS was carried out in parallel with cultures harvested for RNA seq with the exception of PD9 , which was not profiled . Control samples for each media condition were prepared for use in GC/MS analysis with the same methods as described in the RNA seq conditions section above , but without the addition of inoculums . Analysis of culture headspaces was performed on a gas-chromatograph coupled to a time-of-flight mass spectrometer ( GCT Premier , Waters ) . Automated culture sampling was mediated by a CTC CombiPAL Autosampler ( Leap Technologies ) and all cultures were sampled with a 50/30 µm divinylbenzene/carboxen/polydimethylsiloxane StableFlex Fiber ( Supelco ) . GC injection and column parameters , GC temperature program and MS data acquisition parameters were as described previously [5] . Parameters for SPME headspace sampling were as follows . OAC , CELL , and AMM vial cultures were analyzed via automated sampling with a pre-extraction SPME fiber conditioning ( 7 min , 250°C ) , 35 min headspace extraction at 30°C , and a splitless GC injection ( 30 sec , 240°C , 0 . 75 mm ID injection liner ) . Manual headspace sampling of CB , PD4 , and PD14 flask cultures used the following sampling parameters: pre-extraction SPME fiber conditioning ( 12 min , 250°C ) , 30 min headspace extraction ( room temperature , approximately 20–25°C ) , and splitless GC injection ( 30 seconds , 240°C , 0 . 75 mm ID injection liner ) . Data were analyzed with the MassLynx Software Suite™ ( Waters ) . Chromatographic peaks were identified with a combination of spectral search comparisons with the Wiley Registry™ of Mass Spectral Data , 8th Edition , elemental composition analysis and the comparison of retention times and spectra with pure standards for compounds where noted ( Sigma-Aldrich ) . Compounds identified during the analysis of control media samples , including contaminants resulting from the SPME fiber and Wax capillary column , as well as media derived compounds , were excluded from the final compound report for each condition . See Table S1 for the full compound profiles . Growth assays were performed in 96 well plates in 200 µL media containing trace metals as in Griffin et al . , 2010 [65] , 0 . 67 g/L Yeast Nitrogen Base ( Difco ) and supplemented with 100 mM of either glucose , xylose , arabinose , mannose , cellobiose , or sodium acetate , titrated to a pH of 6 with KOH . Wells containing no added carbon source served as the control . The cultures were inoculated by adding 5 µL of 5×107 spores/mL in Phosphate Buffered Saline ( Gibco ) , and the cultures were grown for 5 days at 23°C . Growth was determined by visual inspection . Initial assembly with single end shot gun titanium reads with Roche's GS DeNove Assembler ( Newbler ) resulted in 137 scaffolds with an N50 of 2 . 8 Mb [24] . Following addition of a paired end 3 kb-insert sequencing run , these were assembled into 16 scaffolds encompassing 99 . 5% of the total sequenced base pairs . Called genes were first aligned to the GenBank non-redundant database using blastx ( v2 . 2 . 24 ) [27] , [66] . A hit was defined as a match when overlap with the length of the query protein was greater than 60% and E-value<1e-10 . We extracted the subset of genes found in the CAZY database , a repository for manually curated carbohydrate machinery , and performed a similar procedure [67] . Domains were identified using the hmmsearch function from HMMer [68] with both a set of fungal-specific protein domains [69] and the entire PFAM database [70] . A domain was considered a match if the E-value was greater than 1 and the length of the match was at least 15 . Pathway predictions and enzyme classification was completed through KEGG/KAAS [71] , [72] . GO predictions were made by first mapping the set of A . sarcoides genes to their corresponding Aspergillus nidulans homolog [73] , [74] . Please see Text S1for a full description of the gene cluster and PKS/NRPS identification strategies . In the case of the Illumina runs , mapping was done via building bowtie indices for splice junction libraries , and the genome respectively using default parameters ( tolerated up to 2 mismatches and screened for quality scores ) [75] . Splice junction libraries were generated as described in Habeggar et al . , with 30 bp exon ends [76] . The bowtie reads were converted to mapped read format ( MRF ) and mrfQuantifier was used to compute a variation of reads per kilobase of exon per million mapped sequence reads ( RPKM ) for each gene using RSEQtools [76] . Briefly , we computed RPKM as the number of nucleotides that map per kilobase of exon model per million mapped nucleotides for each gene rather than the read count . It is computed by summing the total number of reads that cover each base pair of an annotation feature of interest ( in this case of exons ) and normalizing by the total length of the feature . For the conditions denoted by CELL , OAC , and AMM , technical replicates were performed yielding one lane per replicate . In the case of PD4 , there were two technical replicates performed 2 months apart . A comparison of the RPKM of the genes between lane replicates showed greater than 99% agreement ( Figure S2 ) , although the correlation was slightly less between the two AMM replicates than between any other conditions . The 454 long reads ( average size 410 bp ) were mapped against the gene models and the genome using BLAT with default parameters [77] . In all cases , only reads that unambiguously mapped to a single location were used for the downstream analysis . For each gene , we calculated the RPKM score as described above . To estimate depth of coverage , the percentage of genes that were detectable using subsamples of reads was computed where detectable was defined as having at least 1 , 2 , 5 , or 10 reads , respectively , overlapping the gene ( Figure S12 ) . A database of transcriptionally active regions ( TARs ) was constructed from those RNA-seq long reads that map uniquely to the genome via BLAT [77] . The TAR database was built by employing the minrun/maxgap segmenting module [76] . Gene coverage values were calculated for a range of minrun/maxgap parameters to assess their impact on observed gene coverage . Included in the coverage analysis were TAR file sets with maximum read gaps between zero and five and minimum read run from 30 to 40 ( See Text S1for a full description; Figures S3 and S4 ) .
|
A renewable source of energy is a pressing global need . The biological conversion of lignocellulose to biofuels by microorganisms presents a promising avenue , but few organisms have been studied thoroughly enough to develop the genetic tools necessary for rigorous experimentation . The filamentous-fungal endophyte A . sarcoides produces metabolites when grown on a cellulose-based medium that include eight-carbon volatile organic compounds , which are potential biofuel targets . Here we use broadly applicable methods including genomics , transcriptomics , and metabolomics to explore the biofuel production of A . sarcoides . These data were used to assemble the genome into 16 scaffolds , to thoroughly annotate the cellulose-degradation machinery , and to make predictions for the production pathway for the eight-carbon volatiles . Extremely high expression of the gene swollenin when grown on cellulose highlights the importance of accessory proteins in addition to the enzymes that catalyze the breakdown of the polymers . Correlation of the production of the eight-carbon biofuel-like metabolites with the expression of lipoxygenase pathway genes suggests the catabolism of linoleic acid as the mechanism of eight-carbon compound production . This is the first fungal genome to be sequenced in the family Helotiaceae , and A . sarcoides was isolated as an endophyte , making this work also potentially useful in fungal systematics and the study of plant–fungus relationships .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"genome",
"analysis",
"tools",
"genomics",
"biology",
"computational",
"biology",
"microbiology",
"genetics",
"and",
"genomics"
] |
2012
|
Genomic Analysis of the Hydrocarbon-Producing, Cellulolytic, Endophytic Fungus Ascocoryne sarcoides
|
Precise neuronal networks underlie normal brain function and require distinct classes of synaptic connections . Although it has been shown that certain individual proteins can localize to different classes of synapses , the biochemical composition of specific synapse types is not known . Here , we have used a combination of genetically engineered mice , affinity purification , and mass spectrometry to profile proteins at parallel fiber/Purkinje cell synapses . We identify approximately 60 candidate postsynaptic proteins that can be classified into 11 functional categories . Proteins involved in phospholipid metabolism and signaling , such as the protein kinase MRCKγ , are major unrecognized components of this synapse type . We demonstrate that MRCKγ can modulate maturation of dendritic spines in cultured cortical neurons , and that it is localized specifically to parallel fiber/Purkinje cell synapses in vivo . Our data identify a novel synapse-specific signaling pathway , and provide an approach for detailed investigations of the biochemical complexity of central nervous system synapse types .
Each of the thousands of cell types present in the nervous system receives multiple classes of inputs that are spatially segregated and functionally distinct . The chemoaffinity hypothesis stated that “the establishment and maintenance of synaptic associations were conceived to be regulated by highly specific cytochemical affinities… . ” [1] . Support for this idea has come from studies of specific synaptic proteins [2 , 3] . For example , different sets of neurotransmitter receptors are found at different synapse types [3] , even at excitatory synapses made on the same neuron [4] . Precise subcellular targeting of synapses is also dependent on the recognition of specific molecules such as adhesion proteins [5] . In addition to these direct-recognition mechanisms , guidepost cells seem to target synapse formation to precise locations: their role has been demonstrated in both invertebrates [6] and vertebrates [7] . Synaptic physiology is also regulated by mechanisms that are synapse type-dependent , since similar stimulation patterns can have opposite effects on plasticity of different synapses [8] . Therefore , the formation and function of each type of synapse is controlled by a complex activation of signaling pathways through specific proteins . Since the visualization of synapses by electron microscopy , attempts have been made at biochemically purifying them and at identifying their chemical composition , especially for the postsynaptic densities characteristic of excitatory synapses [9 , 10] . The use of mass spectrometry ( MS ) to identify proteins in complex mixtures has greatly improved our ability to unravel the protein composition of organelles . Using this technique , over 1 , 000 different postsynaptic proteins have been identified in “bulk” postsynaptic density preparations or in affinity-purified receptor complexes [11–16] . These proteins have a wide range of functions: receptors to neurotransmitters , scaffold proteins , kinases , enzymes , etc . Recently , combining comparative genomics and proteomics , Emes and collaborators [17] have shown that increased behavioral complexity correlates with a phylogenetic expansion of synaptic proteins that are involved in upstream signaling pathways , such as receptors and adhesion molecules . Microarray analysis also showed a very variable regional expression pattern for these upstream synaptic proteins [17] , in accordance with previously obtained results for neurotransmitter receptors [3] . The complexity of the synaptic proteome illustrated by these data highlights the need for studies aimed at systematically identifying the protein composition of individual synapse types , and understanding their mechanistic diversity . To address this issue , we have developed synaptic protein profiling as an approach to isolate and biochemically characterize specific types of central nervous system ( CNS ) synapses . We chose to analyze first the parallel fiber to Purkinje cell ( PF/PC ) synapse in the cerebellum , because of its unique physiological properties and its involvement in neurological disease [18 , 19] . We engineered mice to tag and purify specifically PF/PC synapses and , using MS , we have identified 65 proteins located at the PF/PC synapse . This dataset provides clues to PF/PC synapse-specific signaling pathways , as illustrated by our functional analysis of one of these proteins , MRCKγ . Our results provide an important example of the biochemical complexity of an individual synapse type , and reveal a new mechanism for the regulation of synaptic function .
To enable purification of PF/PC synapses , we developed a transgenic line that expresses an affinity tag only at the PF/PC synapse . We generated a fusion between the glutamate receptor delta2 , GluRδ2 ( National Center for Biotechnology Information [NCBI]# EDK98768 ) , which is specifically localized at the PF/PC postsynaptic density [20] , and Venus , a variant of the green fluorescent protein ( GFP ) . The resulting fusion protein , VGluRδ2 , is properly processed and transported to the cell surface ( Figure S1 ) . To express the fusion specifically in cerebellar Purkinje cells , the VGluRδ2 cDNA was then incorporated into a Pcp2 bacterial artificial chromosome ( BAC ) by homologous recombination , and the resulting Pcp2/VGluRδ2 BAC construct was used to generate transgenic mice ( Figures 1A and S1 ) . Expression of the fusion polypeptide was detected in the cerebellar extracts of Pcp2/VGluRδ2 transgenic mice ( Figure 1B ) , and coimmunoprecipitation experiments demonstrated proper assembly of the VGluRδ2 fusion with the endogenous GluRδ2 receptor subunits ( Figure 1C ) . As shown in Figure 1D , the localization of VGluRδ2 in the molecular layer and somata of PCs agrees with the synaptic localization of the GluRδ2 receptor . In contrast , the enhanced GFP ( eGFP ) control protein expressed using the same BAC vector ( Pcp2/eGFP; http://www . gensat . org ) is detected throughout the cell , including marked labeling of both Purkinje cell dendrites and axons ( Figure 1D ) . Prior to the affinity-purification step , we sought to produce cerebellar extracts enriched for synaptic structures relative to trafficking complexes , and to maximize the recovery of VGluRδ2-tagged postsynaptic densities ( PSDs ) . This was performed by fractionation of a solubilized crude synaptosome fraction ( S3 ) on a gel-filtration column ( Figure 2A ) . As shown in Figure 2B and 2C , this resulted in an enrichment of postsynaptic and mitochondrial proteins , and a relative depletion of endoplasmic reticulum ( ER ) components and presynaptic proteins in the high molecular weight fractions . These excitatory synaptic fractions contain essentially all of the PSD95 scaffolding protein . They also contain VGluRδ2 , which was distributed amongst the different fractions in the same manner as wild-type GluRδ2 ( Figure 2C ) . This was also observed using a standard synaptosome purification ( Figure S2 ) and shows that the fusion receptor VGluRδ2 is targeted to the synapse similarly to the wild-type GluRδ2 . To separate PF/PC PSDs from other cerebellar synapses , we performed affinity purification from the pooled excitatory synaptic fractions ( red rectangle , Figure 2C ) using an anti-eGFP antibody . Electron microscopy of the affinity purified material showed electron-dense structures that were reminiscent of PSDs [21] on the surface of the beads used for purification of VGluRδ2 extracts ( Figures 3C and S3 ) . These structures were absent from beads used to immunopurify extracts from Pcp2/eGFP control cerebella . Using western blots , we could show that more than 50% of the target protein was immunopurified from the input extract for either the control eGFP or the VGluRδ2 extracts ( Figure 3A and unpublished data ) . Western blotting also demonstrated copurification of several PF/PC synaptic components with VGluRδ2 , including the GluRδ2 and GluR2/3 receptors , and the scaffolding proteins PSD93 and Homer ( Figure 3B ) . Markers of inhibitory synapses ( GABA ( A ) receptor α6 , GABA ( A ) receptor β , and gephyrin ) , presynaptic structures ( synapsin I and synaptophysin ) , or of mitochondria ( Cox ) did not copurify , demonstrating the specificity of this approach ( Figure 3B ) . As expected , none of these markers copurified with soluble eGFP in extracts prepared from Pcp2/eGFP control mice ( Figure 3B ) . Taken together , these results demonstrate that the combination of cell-specific genetic targeting , molecular tagging of specific CNS synapses , biochemical fractionation , and affinity purification can be used to enrich for a specific type of PSD from crude brain extracts . To systematically identify components of the PF/PC PSDs , we analyzed the protein content of pooled PF/PC PSD preparations using single- and two-stage MS [22] . A first sample , prepared by pooling three experiments using ten Pcp2/VGluRδ2 cerebella each , enabled us to identify 12 components present at the PF/PC synapse but not present in the control sample prepared in parallel from Pcp2/eGFP cerebella ( Table S1 ) . To increase the number of PF/PC PSD components identified , we performed a second analysis on a sample prepared with a total of 50 cerebella ( Figure 3D ) . A total of 65 proteins were identified: 37 proteins were detected with high confidence ( Figures S4 and S5; Tables S1 and S3 ) , and 28 were observed at lower levels and identified with less confidence ( Figure S6; Tables S2 and S4 ) . This analysis confirmed the presence of the PF/PC synapse proteins GluRδ2 [23] , Homer-3 [24] , PSD93 [23] , delphilin [23] , Shank1 , and Shank2 [25] , and the absence of proteins located at other excitatory ( NMDA receptor subunits , GABA ( A ) receptor α6 ) or inhibitory synapses ( GABA ( A ) receptor α6 , GABA ( A ) receptor β , and gephyrin ) in the cerebellum . Forty of the identified proteins in our affinity-purified PSDs have been previously detected in preparations of synaptic proteins ( [16] and Tables S1 and S2 ) . The 65 proteins we have identified can be grouped into 11 different functional categories ( Figure 3E; Tables S1 and S2 ) . These categories have been previously annotated in studies of the postsynaptic density [13] , with the exception of a class of proteins that we have called “phospholipid metabolism and signaling . ” In the “scaffolds and adaptors” category , several members of the Shank family ( 1 and 2 ) and the PSD family ( PSD93 and PSD95 ) were detected at the PF/PC synapse , illustrating redundancy for scaffold proteins , certainly due to their importance for synaptic function . Other functional categories include proteins important for synapse formation and physiology , such as regulators of small GTPases and protein kinases . Interestingly , eight of the proteins identified in our study can regulate or be regulated by phospholipid metabolism ( Iptr1 , synaptojanin 1 and 2 , phospholipase B , ABCA12 , and MRCKγ ) , or contain phospholipid-binding domains ( Plekha7 , annexin A6 , and MRCKγ ) , and were thus grouped into a previously unrecognized category “phospholipid metabolism and signaling . ” This suggests that phospholipid regulation is a major feature of the PF/PC synapse . Another important category present at synapses groups receptors and ion channels: several glutamate receptor subunits and several G protein–coupled receptors ( GABA-B and BAI receptors ) were detected in our analysis of the PF/PC PSD . Interestingly , the extracellular domain of BAI receptors contains thrombospondin repeats , which can mediate cell adhesion [26] . Several other proteins identified at the PF/PC synapse in our study are involved in cell adhesion and interaction with the extracellular matrix: receptor protein tyrosine phosphatases [27] , delta-catenin-2 [28] , Neph1 [29] and laminins [30] . These diverse potential recognition proteins could form together a “code” defining the PF/PC synapse . To provide additional evidence for the synaptic localization of the novel components that we have identified , we performed immunofluorescence studies on cerebellar sections from wild-type mice . Localization in the molecular layer of the cerebellum , which contains the PF/PC synapses , was evident for MRCKγ , Gm941 , BAIAP2 , RPTPm , Neph1 , and delta2-catenin ( Figure 4 ) . Delta2-catenin and Gm941 could also be detected in some cerebellar interneurons . We also examined the expression of candidates reported in in situ hybridization databases ( http://www . stjudebgem . org; http://www . brain-map . org; and http://www . genepaint . org ) . Interpretable data were available for 42 candidates , and all but two were expressed in Purkinje cells , with a majority showing little detectable expression in the granule cell layer ( Tables S1 and S2 ) . These expression data provide additional confirmation that the majority of the proteins identified in our study are bona fide components of the PF/PC synapse . Within our “phospholipid metabolism and signaling” category , we identified the kinase MRCK gamma ( MRCKγ , NCBI# Q80UW5 ) , which has not previously been localized to synapses . Since MRCK family members have been shown to regulate cytoskeleton reorganization and cell morphology [31 , 32] , we sought to test the role of MRCKγ in dendritic spine morphogenesis in primary cortical cultures . Comparative analysis of dendritic protrusions was carried out for cultures transfected either with GFP alone , or in combination with full-length MRCKγ ( MRCKγFL ) or a MRCKγ construct lacking the kinase domain ( MRCKγDN ) ( Figure 5A ) . Protrusion density is not significantly affected by overexpression of either form of the kinase ( GFP: 9 . 5 ± 0 . 7 protrusions per 20 μm; MRCKgDN: 8 . 0 ± 0 . 6; MRCKgFL: 9 . 1 ± 0 . 6; one-way ANOVA , p = 0 . 29 ) . However , the mean length of dendritic protrusions in neurons overexpressing MRCKγFL decreased when compared to control neurons , whereas the length of protrusions in MRCKγDN-transfected neurons increased ( GFP: 1 . 79 ± 0 . 07 μm; MRCKgDN: 2 . 11 ± 0 . 08; MRCKgFL: 1 . 48 ± 0 . 06; p < 0 . 05 for all comparisons , Dunn multiple comparison test ) . The effect of MRCKγDN implies that it can interfere with endogenously expressed MRCK kinases . Indeed , after data mining of previously published results , we found that MRCKβ has been identified in “bulk” PSD preparations from mouse brain , and thus could be present in cortical neurons [16] . Since mean spine length decreases with maturation [33] , our data demonstrate that MRCK family members , through their kinase function , increase maturation of dendritic spines in primary CNS neurons . Given the presence of MRCKγ in our PSD preparations , and its ability to modulate dendritic spine morphogenesis , we were next interested in its subcellular localization in Purkinje cells ( Figure 5B ) . High-resolution confocal immunofluorescence using an antibody against MRCKγ clearly demonstrated its presence in Purkinje cell dendritic spines , which have been shown to contain GluRδ2 [4] . Moreover , colabeling with markers of specific synapses on Purkinje cells showed that MRCKγ is extensively colocalized with VGluT1 , a marker for PF/PC synapses . MRCKγ is not present in structures labeled by VGluT2 or GAD65/67 , which are present at climbing fiber and inhibitory Purkinje cell synapses , respectively . Taken together , our data support a specific role for MRCKγ in the maturation and plasticity of PF/PC synapses , and confirm the importance of synapse-specific protein profiling for the discovery of signaling pathways that modulate the development and function of specific CNS synapse types .
We have demonstrated that the biochemical components of a specific synapse type from a particular neuronal population can be identified using a combination of genetically engineered mice , affinity purification , and MS . Using our approach , we have prepared a fraction enriched in PF/PC PSDs and identified 65 proteins classified in 11 different functional categories . This dataset provides information on signaling pathways specifically tethered to this synapse , as exemplified by our functional analysis of MRCKγ . It also provides information on the variety of proteins that can be part of the code defining the PF/PC synapse . Approximately 700 different proteins have been identified in PSD preparations from whole brain [16] . However , it has been estimated that , given the mass of a single PSD , the copy number of scaffold proteins in a PSD , and an average size of 100 kDa for each synaptic protein , only about 100 different proteins can be expected to be found at one particular type of PSD [34] . The number of proteins we find in our study is consistent with that estimate . Although our analysis may not have revealed all PF/PC postsynaptic proteins , the successful identification of AMPA receptor subunits in our preparations suggests that any proteins not detected in our sample must be present at low stochiometries in the PSD . Synaptic protein profiling can reveal novel sets of proteins that allow formulation of specific hypotheses regarding synaptic function . For example , we discovered MRCKγ at PF/PC synapses: this kinase is part of a family that has never been described at synapses . This result was striking since MRCK proteins can respond to small GTPases signaling and have been shown to modulate actin cytoskeleton and cell morphology in nonneuronal systems [31] . These characteristics immediately suggest a role for these kinases in spine morphogenesis , which we have now shown for MRCKγ using transfection of cultured cortical neurons . Taken together , these data also have implications for the study of neurodevelopmental diseases . Deficiencies in spine length and spine morphology in Purkinje cells have been found in models of mental retardation and Angelman syndrome [35 , 36] . Given the link between another small GTPase-dependent kinase , PAK3 , and mental retardation [37] , our results suggest that MRCKγ could participate in the signaling pathways involved in mental retardation and autism spectrum disorders . Another interesting finding of our study is the presence of a high proportion of proteins involved in phospholipid metabolism and signaling at the PF/PC PSD . A major regulator of the physiology of the PF/PC synapse is the metabotropic glutamate receptor 1 ( mGluR1 ) which induces phosphatidylinositol-4 , 5-P2 ( PIP2 ) hydrolysis through activation of phospholipase C [8 , 18] . Our results show the presence of MRCKγ and Itpr1 in affinity-purified PF/PC PSDs: these proteins can respond to , respectively , DAG and IP3 , which are the metabolites of PIP2 hydrolysis . This further supports the importance of mGluR1 signaling at the PF/PC synapse and extends the number of regulatory pathways potentially activated by mGluR1 . Also included in the “phospholipid signaling and metabolism” category in our data are synaptojanin-1 and −2 , two PIP2-metabolizing enzymes . These enzymes are best known for their regulation of vesicle recycling at synapses , but have also been found by other biochemical studies at PSDs [16] . Phospholipid metabolism is known to be critical for the function of the presynaptic side of the synapse , especially vesicle recycling [38] . It also plays a role in defining the boundaries of the apical pole and the localization of tight junctions in epithelial cells [39] . Our results suggest that phospholipid signaling also participates in regulating the structure and stability of PSDs . Given the fact that lithium is used as a treatment for schizophrenia and bipolar disorders , and that it might act by modulating phospholipids' metabolism [40] , our results may be particularly relevant for studies of a variety of human neurological disorders . Indeed , it has been suggested that synaptojanin-1 is involved in the cognitive defects observed in Down syndrome [41] , and that PIP2 metabolism may be linked to synaptic dysfunctions in Alzheimer disease [42] . The results presented here provide clues to the nature of the “synaptic code” and the types of molecules that may be critical in definition of specific synapse types . As expected from previous studies [2] , we find proteins with classical adhesion domains such as Neph1 and the receptor tyrosine phosphatase RPTPmu . SYG1 , the Caenorhabditis elegans homolog of Neph1 , has been shown to define synapse location in vivo [6] , and may play a similar role for the PF/PC synapse . Receptor tyrosine phosphatases play important roles in axon guidance , and have also been shown to control synapse formation [43] . We also find proteins at the PF/PC synapse with as yet unknown functions in synaptogenesis , such as the BAI receptors or GABA-B receptors . In this regard , it is interesting to note that the GABA-B receptor 1 contains a CCP module in its extracellular domain . This module is also found in proteins of the complement cascade , which have recently been shown to be involved in synapse development [44] . These proteins , and the majority of the remaining proteins identified in this study , are specifically expressed in Purkinje cells within the cerebellum ( see Results ) . Since cerebellar granule cells also receive excitatory inputs from mossy fibers , we can conclude that , even within the cerebellum , the synaptic codes for specific synapse types must be quite distinct . This supports the results of expression analysis of proteins identified in bulk synapse preparations showing that receptors and other upstream signaling molecules have a highly variable expression pattern in the vertebrate brain [17] . Taken together , these data indicate that very different sets of molecules must define different excitatory synapse types . Although our approach employed the expression of a fusion of GluRδ2 with EGFP in a specific cell type , this basic approach can readily be adapted to characterize a wide variety of synapse types , given the wide range of affinity tags that are now available and the hundreds of BAC vectors that can be used to target expression to specific neurons ( http://www . gensat . org ) . We anticipate that these additional studies of the biochemical diversity of synapses will be critical for understanding the development and function of specific CNS circuits and their dysfunction in disease [45 , 46] .
All experiments using animals were performed according to protocols approved by the Institutional Animal Care and Use Committee at the Rockefeller University . Both the Pcp2/eGFP and the Pcp2/VGluRδ2 transgenics were bred on the FVB background , and littermates were used as wild-type controls . Ten cerebella from adult mice were used for the preparation of a crude synaptosome fraction P2 as presented in Figure 2A ( based on previously published protocols [47] ) . The solution used as a homogenization and resuspension buffer contained 0 . 32 M sucrose , 5 mM HEPES , 0 . 1 mM EDTA ( pH 7 . 3 ) , and a protease inhibitor cocktail ( Sigma ) . P2 was then solubilized 30 min at 4 °C using a final concentration of 0 . 5% Triton X-100 . The cleared solubilized fraction was separated by gravity flow on a gel-filtration column ( Sephacryl S1000 Superfine; GE Healthcare ) prepared using a solution containing 2 mM CaCl2 , 132 mM NaCl , 3 mM KCl , 2 mM MgSO4 , 1 . 2 mM NaH2PO4 , 10 mM HEPES , and 0 . 5% Triton X-100 ( pH 7 . 4 ) . 2-ml fractions were collected , and aliquots were used for protein dosage using the BCA Protein assay kit ( Pierce Biotechnology ) . Calibration of the gel-filtration column was performed using the gel-filtration HMW calibration kit ( GE Healthcare ) . Pooled fractions from the column were used for affinity purification of tagged PSDs . Dynabeads M-270 epoxy beads ( Dynal ) were conjugated using 15 μg of affinity-purified goat anti-GFP antibody per milligram of beads [22]; 6 mg of beads were used for affinity purification of pooled synaptic fractions from ten cerebella during 1 h at 4 °C . Beads were then washed in 2 mM CaCl2 , 300 mM NaCl , 3 mM KCl , 2 mM MgSO4 , 1 . 2 mM NaH2PO4 , 10 mM HEPES , and 0 . 5% Triton X-100 . Purified complexes were finally eluted in 0 . 5 N NH4OH , 0 . 5 mM EDTA for 20 min , dried , and then resuspended in the desired volume of protein electrophoresis sample buffer . Biochemical preparations and affinity purifications were performed in parallel for each genotype , starting with ten cerebella each . For MS analysis , samples from several successive experiments were pooled . Following immunopurification , the isolated proteins were resolved by 1-D SDS-PAGE and stained with Coomassie Blue ( GelCode Blue; Pierce ) . As proteins stain with varied efficiency , for each sample ( from the 30- and 50-mice preparations ) , the complete gel was subjected to mass spectrometric analysis . Each entire gel lane was cut into 66 × 1 mm sections . The 1-mm sections were combined in approximately 30 samples , and proteins were digested with 12 . 5 ng/μl sequencing-grade modified trypsin ( Promega ) . The resulting peptides were extracted on reverse-phase resin ( Poros 20 R2; PerSeptive Biosystems ) and eluted with 50% ( v/v ) methanol , 20% ( v/v ) acetonitrile , and 0 . 1% ( v/v ) trifluoroacetic acid containing 2 , 5-dihydroxybenzoic acid ( 2 , 5-DHB; 1:3 v/v saturated matrix solution in elution solution ) . Samples were subjected to matrix-assisted laser desorption/ionization ( MALDI ) quadrupole/time-of-flight ( QqTOF ) MS and MALDI ion trap ( MALDI-IT ) tandem MS ( MS/MS ) analyses using an in-house–built MALDI interface coupled to a Qq-TOF instrument ( QqTOF Centaur; Sciex ) and an ion trap ( LCQDECAXPPLUS; Finnigan ) as described [22 , 48 , 49] . XProteo computer algorithm ( http://www . xproteo . com ) was used to search the peptide fingerprint data and collision-induced dissociation ( CID ) MS/MS data in the NCBI database ( see Text S1 ) . Due to the limited amount of samples , all MALDI-IT CID MS/MS spectra were carefully acquired and interpreted manually . A MS/MS hypothesis-driven approach on isolates from control Pcp2/eGFP transgenic mice was used to probe for the specificity of the proteins copurified with VGluRδ2 ( Text S1 ) . The MRCKγ cDNA was amplified from cerebellar cDNA . The MRCKγDN construct was obtained by deleting the sequence encoding for amino acids 1–426 , and by replacing it with ATG . MRCKγ and eGFP cDNAs were subcloned in the bidirectional Tet-responsive vector pBI ( Clontech ) . Primary neuronal cultures were prepared from E15 mouse embryos ( Swiss strain ) . Cortices were dissected and triturated using a fire-polished Pasteur pipette and 0 . 05% trypsin . Neurons were plated on poly-d-lysine and laminin-coated coverslips at a density of 1 . 5 × 105 cells/cm2 and cultured in neurobasal medium supplemented with 2% B27 supplement , 0 . 5 mM glutamine , and antibiotics . Transfections were performed at DIV7 with a 1:1 ratio of a tTA-expressing plasmid and the bidirectional vector containing GFP ( with or without the kinases ) using Lipofectamine 2000 according to the manufacturer's instructions ( Invitrogen ) . Dendrites of transfected neurons were imaged using a confocal microscope and a 63× objective with a 5× zoom . Quantifications of protrusion density and length were performed using the NeuronJ plugin and the ImageJ software on several dendrites per neuron ( at least five different cells per transfection condition , four independent experiments ) . A total of 1 , 029 , 861 , and 950 protrusions were counted and measured for the GFP , DN , and FL transfections , respectively . Statistical analysis was performed using the GraphPad Prism software .
|
The brain is composed of many different types of neurons that form very specific connections: synapses are formed with specific cellular partners and on precise subcellular domains . It has been proposed that different combinations of molecules encode the specificity of neuronal connections , implying the existence of a “molecular synaptic code . ” To test this hypothesis , we describe a new experimental strategy that allows systematic identification of the protein composition for individual synapse types . We start with mice that are genetically engineered to facilitate the purification of one type of synapse from a given neuronal population in the central nervous system , the parallel fiber/Purkinje cell synapse . The purification is performed using a combination of biochemical fractionation and affinity purification . Subsequent mass spectrometry allows us to identify approximately 60 different proteins present in the resulting sample . We have further analyzed some of the 60 proteins and show that MRCKγ , a newly identified kinase , is localized in the dendritic spines where the parallel fiber/Purkinje cell synapses are formed and that it can modulate the morphogenesis of dendritic spines . The use of this experimental strategy opens up the ability to provide insights into the underlying “molecular code” for the diverse types of synapses in the brain .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"neuroscience"
] |
2009
|
Proteomic Studies of a Single CNS Synapse Type: The Parallel Fiber/Purkinje Cell Synapse
|
Uncertainty of fear conditioning is crucial for the acquisition and extinction of fear memory . Fear memory acquired through partial pairings of a conditioned stimulus ( CS ) and an unconditioned stimulus ( US ) is more resistant to extinction than that acquired through full pairings; this effect is known as the partial reinforcement extinction effect ( PREE ) . Although the PREE has been explained by psychological theories , the neural mechanisms underlying the PREE remain largely unclear . Here , we developed a neural circuit model based on three distinct types of neurons ( fear , persistent and extinction neurons ) in the amygdala and medial prefrontal cortex ( mPFC ) . In the model , the fear , persistent and extinction neurons encode predictions of net severity , of unconditioned stimulus ( US ) intensity , and of net safety , respectively . Our simulation successfully reproduces the PREE . We revealed that unpredictability of the US during extinction was represented by the combined responses of the three types of neurons , which are critical for the PREE . In addition , we extended the model to include amygdala subregions and the mPFC to address a recent finding that the ventral mPFC ( vmPFC ) is required for consolidating extinction memory but not for memory retrieval . Furthermore , model simulations led us to propose a novel procedure to enhance extinction learning through re-conditioning with a stronger US; strengthened fear memory up-regulates the extinction neuron , which , in turn , further inhibits the fear neuron during re-extinction . Thus , our models increased the understanding of the functional roles of the amygdala and vmPFC in the processing of uncertainty in fear conditioning and extinction .
The associative memories acquired through both appetitive and aversive conditioning with uncertainty have been shown to exhibit substantial resistance to extinction , known as the “partial reinforcement extinction effect ( PREE ) ” [1–3] . For example , the fear memory acquired through a fear conditioning procedure in which the CS is probabilistically paired with the US ( partial reinforcement ) is more resistant to extinction than the fear memory acquired after full pairings of the CS and US ( full reinforcement ) ( Fig 1A–1C ) . This initially sounds paradoxical because one may assume that the causal relationship between the CS and US is strong and weak in the cases of full and partial reinforcement , respectively [4] . Although psychological theories [5–7] and computational models [8–12] for how uncertainty induces paradoxical PREE have been proposed , the neural underpinnings of the PREE remain largely unclear . The neural substrates implicated in fear conditioning and extinction are the amygdala and mPFC , respectively . The amygdala is a major region for the acquisition and expression of fear memory [13–15] . In contrast , the ventral subdivision of the mPFC ( vmPFC ) , called the infralimbic cortex ( IL ) in rodents and the ventral mPFC in primates [16] , plays an important role in the extinction of fear memory [17–19] . Although both the amygdala and mPFC function during partial reinforcement fear conditioning [20–25] , their roles in the PREE have rarely been examined [26] . Thus , how the amygdala and mPFC are coordinated for the PREE remains elusive . Recently , the electrophysiological properties of neurons in the amygdala and mPFC have been extensively investigated; interestingly , three different types of neurons have been identified and defined as the following basic properties: “fear neurons” , which exhibit CS-evoked activity ( spike firing ) after fear conditioning and abolished activity after subsequent extinction; extinction-resistant “persistent neurons” , which also exhibit CS-evoked activity after fear conditioning but are resistant to subsequent extinction and display sustained activity; and “extinction neurons” , which are silent after fear conditioning but display CS-evoked activity after subsequent extinction . Neural populations that match the definitions of these three-types of neurons are not localized to specific regions; instead , they are redundantly distributed over the amygdala and mPFC: fear neurons have been found in the basal nuclei of the amygdala ( BA ) [27 , 28] , lateral nuclei of the amygdala ( LA ) [28–31] and central nuclei of the amygdala ( CEA ) [32]; persistent neurons have been found in the BA [27 , 28] and LA [28 , 30 , 31]; and extinction neurons have been found in the BA [27 , 28] , the group of intercalated cells ( ITC ) [32] and the vmPFC [33–35] . The following questions arise: How do these three neural populations interact ? Furthermore , how do their interactions process both CS and uncertainly generated US inputs during partial reinforcement fear conditioning and generate an extinction-resistant fear response as output ? The vmPFC is widely considered to be a primary assembly of extinction neurons because it inhibits the amygdala through activating the GABAergic ITC [36 , 37] . In fact , activation of the vmPFC led to the suppression of CS-evoked fear memory [38 , 39] . Nevertheless , a recent optogenetic study showed that the vmPFC is necessary for the formation but not the expression of the extinction memory , suggesting that inhibitory sources other than the vmPFC could also suppress the fear memory [40] . Thus , the functional role of the vmPFC remains controversial . Based on these neural findings , this study sought a possible explanation of the PREE by hypothesizing that a combination of fear , persistent and extinction neurons plays an important role in the PREE . To test this hypothesis , we first developed a mathematical model of a neural circuit based on three neural units with the basic properties of the fear , persistent and extinction neurons . We then presented how uncertainly generated US inputs were processed in the neural circuit model , with a particular eye to the PREE . Finally , an extension of the model provided a plausible explanation for the controversial role of the vmPFC in the formation of extinction memory .
The model mainly consisted of fear and persistent neurons in the amygdala and extinction neurons that were considered to be within the vmPFC ( Fig 1D ) . Note that extinction neurons were also found in the amygdala and that those neurons receive synaptic inputs from the vmPFC [28 , 37] . Here , we simply addressed populations of fear , persistent and extinction neurons as single representative units: fear , persistent and extinction neural units . Thus , the activity of each neural unit represents the averaged firing rate of each neural population . In the model , these neural units composed two kinds of networks for neural activity and learning signals . In the neural activity-regulating network , all units were activated by excitatory synaptic input from the CS , and the extinction neural unit inhibited the fear neural unit ( black line in Fig 1D ) . Behavioral fear responses were simply represented by the activity of the fear neural unit , reflecting the fact that the firing rate of fear neurons is well correlated with the freezing response of animals [41] . The activities of the fear neural unit ( F ) , persistent neural unit ( P ) , and extinction neural unit ( E ) at trial t were described by F ( t ) =wF ( t ) CS ( t ) −wF , EE ( t ) , ( 1 ) P ( t ) =wP ( t ) CS ( t ) , ( 2 ) E ( t ) =wE ( t ) CS ( t ) , ( 3 ) where CS denotes the CS input , which was 1 when the CS was provided and 0 otherwise; wF , E denotes the synaptic weight with which the extinction neural unit inhibits the fear neural unit; and wF , wP and wE indicate the synaptic weights of the CS-related inputs ( black lines in Fig 1D ) . In the learning signal-regulating network , the learning signals inducing synaptic plasticity of wF , wP and wM were computed in a neural activity-dependent manner ( blue line in Fig 1D ) . These weights were updated on a trial-by-trial basis after each CS-US presentation by the following synaptic plasticity rules: ΔwF=αFCS ( t ) [US ( t ) −F ( t ) ]+ , ( 4 ) ΔwP=αPCS ( t ) [US ( t ) −P ( t ) ]+ , ( 5 ) ΔwE=αECS ( t ) [F ( t ) {P ( t ) −US ( t ) −E ( t ) }]+ , ( 6 ) where αF , αP and αE denote the learning rates; US is the intensity of US input; and [x]+ is a rectified linear function: [x]+ is 0 and x when x<0 and x≧0 , respectively . The brackets ( []+ ) in the equations represent the learning signals that regulate the gain of synaptic plasticity [42] ( blue lines in Fig 1D ) . In this scheme of synaptic plasticity , the activity of the fear , persistent and extinction neural units can be interpreted to represent ‘prediction of severity ( threat ) ’ , ‘prediction of US intensity’ and ‘prediction of safety ( no presentation of US; we subsequently refer to this case as ‘no-US’ ) ’ , respectively . The fear neural unit receives two inputs ( eq ( 1 ) ) : the CS input , which is a cue signal for the subsequent US , and the inhibitory input from the extinction neural unit , which predicts the degree of safety . This suggests that the fear neural unit predicts the net severity . In eq ( 4 ) , the synaptic weight of the CS input to the fear neural unit , wF , is modulated , according to the Rescorla-Wagner learning rule [43 , 44] , based on the prediction error of the net severity . The persistent neural unit responds to the sole CS input ( eq ( 2 ) ) . The synaptic weight of the CS input to the persistent neural unit , wP , is also modified according to Rescorla-Wagner learning ( eq ( 5 ) ) , which allows the persistent neural unit to predict the US intensity ( P = US ) . To reflect that actual extinction neurons come to respond to the CS after extinction training [28] , the extinction neural unit was assumed to encode safety ( no-US ) in our model . In eq ( 6 ) , the synaptic weight of the CS input to the extinction neural unit , wE , is modified by the prediction error of the safety , where P−US represents the actual safety ( no-US ) , e . g . , after acquisition of fear memory , P−US = 0 when US , and P−US = learned US intensity when no-US . In the model , the CS-US pair was applied in a sequential training manner during fear conditioning and extinction . During fear conditioning , the CS and US were repeatedly paired in the full reinforcement case ( US = 1 ) ( Fig 2A ) , whereas the US was probabilistically paired with the CS in the partial reinforcement case ( P ( US = 1|CS = 1 ) <1 ) ( Fig 2B ) . During extinction , the CS was repeatedly applied without being paired with the US ( US = 0 ) . The basic model was extended by additionally introducing two factors: another extinction neural unit and multiple timescales of synaptic plasticity . These two factors were implied by a recent optogenetic study [40] . Silencing of the vmPFC ( i . e . , IL in rodents ) had no effect on CS-evoked behavioral responses during extinction , suggesting another inhibitory source of fear memory other than the vmPFC . In addition , silencing of the vmPFC during extinction impaired the retrieval of extinction memory , suggesting that formation and consolidation of extinction memory are regulated by fast and slow timescales of synaptic plasticity . The extended model took into account a heterogeneous collection of subregions in the amygdala as well as in the vmPFC [45] , in which the LA and CEA were represented as the persistent and fear neural units , respectively , whereas the ITC and the vmPFC were represented as extinction neural units ( Fig 1E ) . All subregions receive CS-related synaptic inputs through the thalamus and cortex . The ITC received CS-related input and input from the vmPFC , and it inhibited the CEA . Behavioral fear response was represented by the activity of the CEA . The activity of the CEA ( F ) , LA ( P ) , vmPFC ( E1 ) and ITC ( E2 ) are described by F ( t ) =wF ( t ) CS ( t ) −wF , E2E2 ( t ) , ( 7 ) P ( t ) =wP ( t ) CS ( t ) , ( 8 ) E1 ( t ) =wE1 ( t ) CS ( t ) , ( 9 ) E2 ( t ) =wE2 ( t ) CS ( t ) +wE2 , E1E1 ( t ) , ( 10 ) where wi ( i = {F , P , E1 , E2} ) indicates the activity-dependent modifiable synaptic weight of CS-related input to i ( black lines in Fig 1E ) and wi , j indicates the constant synaptic weight of input from j to i . The synaptic plasticity rules were the same as those in the basic model ( eqs ( 4–6 ) ) , except that two different timescale dynamics were introduced to wE2 to represent early- and late-phase plasticity [46 , 47]: ΔwF=αFCS ( t ) [US ( t ) −F ( t ) ]+ , ( 11 ) ΔwP=αPCS ( t ) [US ( t ) −P ( t ) ]+ , ( 12 ) ΔwE1=αE1CS ( t ) [F ( t ) {P ( t ) −US ( t ) −E1 ( t ) }]+ , ( 13 ) ΔwE2=αE2CS ( t ) [F ( t ) {P ( t ) −US ( t ) −E2 ( t ) }]+−βE2 ( wE2 ( t ) −wE2∞ ( t ) ) , ( 14 ) where αi ( i = {F , P , E1 , E2} ) denotes the learning rate , and βE2 denotes the relaxation rate , of wE2 ( t ) to wE2∞ ( t ) . The first terms correspond to early-phase plasticity , which depend on the learning signals ( blue lines in Fig 1E ) indicated by brackets ( []+ ) . The second term in eq ( 14 ) represents late-phase plasticity , where wE2∞ indicates the weight capacity , the dynamics of which are described by ΔwE2∞=αE2∞CS ( t ) E1 ( t ) −βE2∞ ( wE2∞ ( t ) −wE2 ( t ) ) , ( 15 ) where αE2∞ and βE2∞ indicate the learning rate and relaxation rate , respectively . Here , the learning signal to wE2∞ is E1 ( t ) . According to eqs ( 14 ) and ( 15 ) , wE2 is consolidated to wE2∞ , and wE2∞ is also relaxed to wE2 depending on the activity of the vmPFC , E1 ( t ) . What are the molecular substrates of early- and late-phase plasticity ? Early-phase long-term potentiation ( LTP ) is regulated by Ca2+ signaling-regulated phosphorylation of AMPA-R on endosomes , which induces the exocytosis and membrane accumulation of AMPA-R [48] . In contrast , late-phase LTP is thought to be regulated by gene expression with slow dynamics , in which proteins are newly synthesized in the soma , actively transported to spines and inserted into the postsynaptic density ( PSD ) [49 , 50] . Thus , wE2 and wE2∞ in eqs ( 14 and 15 ) correspond to the total number of AMPA-Rs on membrane and the size of PSD , i . e . , AMPA-R capacity , respectively . Then , each term in eqs ( 14 and 15 ) can be interpreted as the following biological processes: The first term in eq ( 14 ) represents the early-phase LTP , i . e . , an increase in the total AMPA-Rs , regulated by the learning signal . The first term in eq ( 15 ) represents an increase in the AMPA-R capacity , regulated by the vmPFC . The second term in eq ( 14 ) represents spontaneous cycling ( i . e . , exocytosis and endocytosis ) of the total AMPA-Rs , converging to the AMPA-R capacity . The second term in eq ( 15 ) represents spontaneous cycling of the AMPA-R capacity ( i . e . , synthesis and degradation of proteins in the PSD ) depending on the total AMPA-Rs .
To determine the basic dynamics of the fear , persistent and extinction neurons , we started with the basic model ( Fig 1D ) . In the simulation of fear conditioning with full reinforcement and subsequent extinction ( Fig 2A ) , the CS-evoked activities of the fear , persistent and extinction neural units were consistent with the respective properties of fear , persistent and extinction neurons in the amygdala ( Fig 2C ) [26 , 28 , 32] . In addition , the activity of the fear neural unit well represented the behavioral freezing rate as observed in fear conditioning and extinction [28] . Thus , the basic model reproduced the behaviors of the fear , persistent and extinction neurons . Next , we performed simulations in the partial reinforcement case ( Fig 2B ) . During the fear conditioning with partial pairing of the US , the activity of the fear neural unit increased when the US was presented and decreased when it was not ( no-US ) , but the overall activity tended to increase; in contrast , the activities of the persistent and extinction neural units increased only when US and no-US were presented , respectively ( Fig 2D ) . During the subsequent extinction phase , we observed the PREE ( Fig 2D ) : the activity of the fear neural unit slowly decreased with residual activity , in contrast to what was observed in the full reinforcement case ( Fig 2C ) . Consistently , residual neural firing has been observed in the amygdala after the extinction training of partially reinforced fear memory [26] . We also found that the synaptic weight from the extinction neural unit to the fear neural unit , wF , E , was critical for the PREE ( S1 Fig ) . In addition , we found that the PREE-like effect could be observed during successive full reinforcement conditioning and extinction , being equivalent to partial reinforcement conditioning overall ( S2 Fig ) . As conditioning and extinction repeat , the residual activity of the fear neural unit accumulates and becomes saturated . In fact , it has been seen in the literature that the re-conditioned fear memory exhibited substantial resistance to re-extinction [51–54] . Hence , the basic model based on crosstalk between the fear , persistent and extinction neurons processed the probabilistic nature of the pairing , i . e . , the uncertainty . What causes the difference in the extinction of fully and partially reinforced fear memories ? After fear conditioning with full reinforcement , CS-evoked activity of the persistent neural unit converged to the US intensity , i . e . , P = 1 , whereas the extinction neural unit showed no activity , i . e . , E = 0 ( Fig 2C ) . Thus , the no-US at the beginning of the extinction , i . e . , US = 0 , led to the maximum level of the learning signal to the extinction neural unit ( red line in Fig 2G ) , which was proportional to P-US-E , resulting in the extinction and fear neural units showing a rapid increase and decrease in their respective CS-evoked activities . After fear conditioning with partial reinforcement , P = 1 , the same as after fear conditioning with full reinforcement , whereas the extinction neural unit showed a certain level of activity , i . e . , E≧0 ( Fig 2D ) . Thus , at the beginning of the extinction , i . e . , US = 0 , the learning signal to the extinction neural unit , which was proportional to P-US-E , exhibited a lower level than that after fear conditioning with full reinforcement ( red line in Fig 2H ) . Therefore , the extinction neural unit could not produce enough of an increase in the CS-evoked activity to inhibit the fear neural unit . This is a scenario of the PREE . In addition , we found that the learning signal to the extinction neural unit was correlated with the degree of ‘surprise’ from a statistical standpoint; after fear conditioning with full reinforcement , the no-US input at the beginning of the extinction phase was unpredictable , leading to a relatively large degree of surprise ( red line in Fig 2G ) . In contrast , after fear conditioning with partial reinforcement , the no-US input at the beginning of the extinction phase was predictable , leading to a relatively small degree of surprise ( red line in Fig 2H ) . We also quantitatively evaluated the degree of surprise by developing a statistical inference model based on sequential updating of Bayesian logistic regression ( see S1 Text ) . Then , we found that the learning signal to the extinction neural unit was positively correlated with the degree of surprise ( see S3 and S4 Figs ) . We further investigated the effect of uncertainty during fear conditioning on the PREE . The uncertainty was evaluated in terms of the Shannon entropy of the probability distribution for the waiting time ( number of trials: n∈{1 , 2 , …} ) until the next US observation , P ( n ) = P ( US = 1|CS = 1 ) ( 1−P ( US = 1|CS = 1 ) ) n−1 [55] . Obviously , the uncertainty monotonically increased as P ( US = 1|CS = 1 ) decreased ( Fig 2I ) . With a high degree of uncertainty ( low P ( US = 1|CS = 1 ) ) , the fear neural unit was highly resistant to extinction with a longer time constant ( Fig 2J ) , and its residual activity at the end of the extinction phase remained high ( Fig 2K ) , indicating that uncertainty facilitated the PREE . This is because the activity of the extinction neural unit was almost saturated after fear conditioning and did not increase enough to inhibit the fear neural unit during the extinction phase ( Fig 2D ) due to the weakness of the learning signal to the extinction neural unit ( Fig 2L ) . In our basic model , the extinction neural unit , which presumably corresponds to the vmPFC , was the unique source of inhibition of the fear memory , indicating that the extinction neural unit was required for both the formation and retrieval of the extinction memory . Consistently , it has been shown that optogenetic activation of the vmPFC ( i . e . , IL in rodents ) reduces the fear response; in particular , vmPFC activation during the extinction phase facilitated the consolidation of the extinction memory [40] . However , vmPFC silencing experiments did not produce data consistent with the idea that the vmPFC is the unique source of inhibition of the fear memory; optogenetic silencing of the vmPFC during extinction impaired the retrieval of the extinction memory the next day , whereas silencing the vmPFC during retrieval had no effect , indicating that the vmPFC is necessary for the formation of the extinction memory but not for its retrieval [40] . This hypothesis has also been supported by a vmPFC lesion study [56] . Moreover , silencing the vmPFC during the extinction phase did not change the CS-evoked behavioral responses compared with those in the normal condition , although it impaired the retrieval of the extinction memory the next day , suggesting that formation and consolidation of the extinction memory are regulated by fast and slow timescales of synaptic plasticity . Here , we aimed to reproduce the new findings of this optogenetic study by extending the basic model . In the extended model , we considered a neural circuit consisting of nuclei in the amygdala and mPFC; in this model , the LA , CEA and vmPFC were simply represented by the persistent , fear , and extinction neural units , respectively . The extended model also included the ITC as another extinction neural unit ( Fig 1E ) ( see Model ) . The extended model could provide minimal understanding of the amygdala-mPFC neural circuit , although a simple correspondence between brain regions and functions , e . g . , the CEA as fear neurons , the LA as persistent neurons and both the ITC and vmPFC as extinction neurons , could be an oversimplification because fear , persistent and extinction neurons are distributed throughout the amygdala and mPFC . In this model , we also developed a synaptic plasticity mechanism for early-phase memory formation and late-phase memory consolidation ( see Model ) . In the simulation , after the schedule of CS-US pairings used in the basic model was repeated , no-CS and no-US resting trials as well as subsequent retrieval trials with only the CS were performed ( Fig 3A ) . This extended model showed essential behaviors of fear memory acquired through fear conditioning with full reinforcement and extinction ( Fig 3B ) . At the retrieval of the extinction memory after the resting phase , spontaneous recovery of the fear memory occurred to a small extent , as commonly observed after long intervals [57–59] . This is because during the resting phase , the synaptic weight to ITC , wE2 , settled down to the weight capacity , wE2∞ , due to the slow dynamics of the late-phase LTP ( S8B Fig ) . We also confirmed that the extended model generated the PREE in partial reinforcement conditioning ( S5 Fig ) and the PREE-like effect in successive full reinforcement conditioning and extinction ( S6 Fig ) , consistent with the basic model ( Fig 2D and S2 Fig ) . The extended model consistently reproduced the experimental results ( see Figs 2B , 2C , 3C , and 4B in Do-Monte et al . [40] ) observed for the activation and silencing of the vmPFC during extinction and retrieval ( Fig 3C–3F ) . Activation of the vmPFC reduced the expression of fear during both extinction ( Fig 3C ) and retrieval ( Fig 3D ) , and the vmPFC activation during extinction also facilitated the subsequent consolidation of the extinction memory during the resting phase , causing no recovery of the fear memory at the retrieval ( Fig 3C ) . When the vmPFC was suppressed during extinction , even though there was no effect on the extinction of the fear memory , the extinction memory was not consolidated during the resting phase , leading to significant spontaneous recovery of the fear memory ( Fig 3E ) . This finding is consistent with recent reports [40 , 56] and suggests that extinction learning is regulated by separate synaptic plasticity mechanisms consisting of early-phase memory formation that is independent of the vmPFC and late-phase memory consolidation that depends on the vmPFC , as assumed in our model . In addition , consistent with recent reports [40] , suppressing the vmPFC during retrieval did not affect the extinction memory ( Fig 3F ) . Taken together , the consistency between previous experimental reports and our simulation supports the validity of our extended model , which included another inhibitory source of fear memory in addition to the vmPFC as well as separate synaptic plasticity mechanisms underlying early-phase memory formation independent of the vmPFC and late-phase memory consolidation dependent on the vmPFC . The partially reinforced fear memory could not be fully inhibited by the extinction training ( Fig 2D and S5D Fig ) , which is reminiscent of exposure therapy-resistant anxiety disorder , panic disorder and post-traumatic stress disorder ( PTSD ) [60] . Here , we explored a new procedure to relieve the resistance to extinction , and we then tested it based on our model . Diminishing the extinction-resistant fear response would require a considerable increase in the activity of the extinction neural unit , which is driven by the learning signal , as described by eq ( 6 ) . According to this equation , increasing the learning signal requires an increase in the activity of the fear and persistent neural units . Thus , this observation suggested a ‘shock procedure’ in which the CS was paired with a stronger US; this pairing was applied before further extinction training . We then tested this shock procedure by using the extended model . The activity of the fear neural unit was first rapidly elevated due to high intensity of the US ( Fig 4A ) and then rapidly decreased to almost 0 during the subsequent extinction ( black line in Fig 4C ) , indicating that the extinction-resistant fear memory was completely inhibited . When the shock procedure employing an US of the same intensity was applied ( Fig 4B ) , the fear memory was conversely reinforced ( Fig 4D ) , suggesting that the intensity of the US is a critical determinant for the effectiveness of the shock procedure . The differences in the outcomes of these cases were due to different levels of learning signals to two extinction neural units ( E1 and E2 ) at the beginning of the second extinction ( green and red lines in Fig 4E and 4F ) . We evaluated the effectiveness of the shock procedure for various US intensities by final activity level of CEA ( the fear neural unit ) after re-extinction training ( Fig 4G ) . We confirmed that the shock procedure was also successful in the basic model ( S7 Fig ) .
It has been generally accepted that extinction is a form of inhibitory learning , which is opposite from an erasure or forgetting of fear memory [61] . In fact , extinguished fear responses recover under various circumstances . For example , an extinguished fear response spontaneously recovers after a long time , e . g . , several days [58] , and also reappears after exposure to the US without the CS , known as reinstatement [62] . It has been reported that extinction training inhibited the fear responses but could not erase the fear memory in adult rat , although erasure of fear memory may occur during early stages of postnatal development [63 , 64] . Consistently , conditioned fear memory in our model was not erased by a decrease in synaptic weights but was inhibited by extinction neurons . Recent physiological studies have identified neural populations with distinct firing characteristics , such as fear , persistent and extinction neurons , in the amygdala [27–32] . These findings suggest that information processing in the amygdala may take place through these neural populations . However , the computational role of each neural population in fear conditioning and extinction has not been well studied . In this study , we presented the neural implementation based on these neural populations and proposed the types of information that are encoded and processed through interactions between these neural populations: the fear , persistent , and extinction neurons encode the prediction of net severity , of US intensity and of safety ( no-US ) , respectively , and the weights of their synaptic inputs are modulated by the corresponding prediction errors . Consistent with the persistent neurons in our model , a previous report showed that CS-evoked activity of the persistent neurons in the LA does not further increase after reconditioning , suggesting that persistent neurons represent the memory of the US intensity [30] . Consistent with the extinction neuron in our model , a human fMRI study showed that the vmPFC uniquely encodes safety accompanied by the CS during extinction [24] , suggesting that the CS synaptic input to the vmPFC could be plastically regulated by ‘prediction error of safety’ . This proposed encoding mechanism could be further validated by experiments , such as electrophysiological recording of neural firing or the labeling of cFos immunoreactivity during partial reinforcement fear conditioning and extinction . It can be speculated that the learning signals in our model could be implemented through neuromodulators , e . g . , dopamine , serotonin , noradrenaline , acetylcholine , norepinephrine and oxytocin [42] . In general , the release of neuromodulators is associated with particular mental states , e . g . , reward , positive and negative emotions , happiness , motivation , attention and arousal [65] . Neuromodulators regulate neuronal firing and the efficacy of synaptic plasticity [66] . In particular , dopamine has been extensively investigated , and it is widely accepted that dopamine release from the ventral tegmental area ( VTA ) is a specific response to reward-related prediction error , i . e . , the acquisition of a greater reward than expected [67 , 68] and that the dopamine release facilitates the synaptic plasticity that underlies the association between sensory input ( the CS ) and reward ( the US ) [69–72] . Based on these facts , reinforcement learning theory has suggested that animals perform temporal difference ( TD ) learning [73–75] because the basal ganglia , which is involved in decision making , receives dense axonal projections from the VTA and exhibits dopamine-dependent plasticity of synaptic inputs from the cortex [71] . However , it has been known that dopaminergic neurons show firing responses not only to rewards but also to aversive stimuli [76] and , moreover , show diverse firing patterns that may encode prediction errors of other valences [77 , 78] . In addition , the VTA was recently suggested to be composed of anatomically and functionally heterogeneous dopaminergic neurons whose axons project to different regions , including the amygdala and mPFC [79] . Taken together , dopamine signals to different neural populations may represent different meanings , such as the prediction error of net severity , of US intensity and of safety , as assumed in our model . Fear conditioning has been used as a model system for anxiety disorders such as panic disorder , PTSD and obsessive-compulsive disorder ( OCD ) [80] . Traditional exposure therapy , which corresponds to extinction training , is an effective cure for anxiety disorders in some patients [81] , but some severe patients also show strong resistance to exposure therapy [82 , 83] . Moreover , anxiety disorders may be worsened by occasionally experienced negative social reactions [84 , 85] , which are akin to partial reinforcement experiences , and become strongly resistant to exposure therapy , similar to the PREE . Thus , we think that the widely used fear conditioning with full reinforcement , in which the acquired fear memory can be easily diminished by extinction training , is rare in real life and thus is not a good model for understanding anxiety disorders; instead , fear conditioning with partial reinforcement , which results in an extinction-resistant fear memory , is a more realistic model for neuroscience research on anxiety disorders and should improve the translatability of results [18 , 26] . To relieve extinction-resistant fear memory , we proposed a ‘shock procedure’ based on our model . The fear memory was diminished by extinction if a stronger US was paired with the CS before the extinction procedure ( Fig 4 and S7 Fig ) . In the shock procedure , although the fear memory is temporarily strengthened by the stronger US , the subsequent extinction training becomes effective , suggesting that an increase in activity in the amygdala ( persistent and fear neurons ) or in the learning signal to the vmPFC is key for effective extinction training . The shock procedure can be tested in animal experiments , but employing a stronger US as part of the shock procedure may raise ethical concerns for humans . It can be intuitively interpreted that the animal cannot comprehend the rule change to extinction after fear conditioning with partial reinforcement , whereas after the shock procedure with a strong US , in contrast , the animal can internalize the pronounced rule change to extinction , thus allowing the fear memory to be extinguished . The proposed ‘shock procedure’ may provide insight not only for the development of new therapies but also for understanding the neural mechanisms of fear memory extinction . Classical conditioning has been computationally modeled in a number of ways . In the field of behavioral psychology , ‘former learning theory’ , which defines mathematical embodiments to describe learning and behavioral phenomena , has been tested [86] . The Rescorla-Wagner model was a seminal former learning theory that described an association between CS and US controlled by prediction error as a learning signal [43] . Since then , many alternative models have been proposed to reproduce many observed phenomena in classical conditioning and extinction [87–91] . However , these models failed to explain the PREE . Moreover , these models did not fully describe their neural mechanisms , although several models can be implemented using neural networks [92 , 93] . ‘Reinforcement learning’ was proposed as an extension of the Rescorla-Wagner model; in this system , which animals explore optimal behavioral strategies by interacting with their environment to maximize the accumulated reward over time [94] . Sutton and Barto proposed temporal difference ( TD ) learning , in which the prediction of expected cumulative future reward was updated by its prediction error , called TD error [94] . The framework of TD learning reproduced classical conditioning and extinction [75] but not the PREE . To account for the PREE , TD learning was extended by two models [9 , 10] . Redish et al . [9] introduced a categorization process for inexperienced observations into new latent states , whereas Song et al . [10] introduced arousal signal-dependent learning to the existing TD learning model [95] . Although these TD learning-based models were successful in reproducing the PREE , how neural computation is performed by fear , persistent and extinction neurons has remained unclear . Another aspect of computational modeling is ‘statistical decision theory’ [96] . The PREE has been addressed by Bayesian estimation of the US probability per trial or unit time [12 , 97 , 98] . This framework was extended to introduce latent causes [8] . Related to latent causes , Gershman et al . [99] developed a Bayesian inference model based on a categorization process of contexts [9] . This model was further extended by introducing a hidden Markov model ( HMM ) , in which a particular context tended to persist over time , and this model successfully generated the PREE [11] . Although these models provided important concepts in light of statistical decision making , they did not describe the underlying neural mechanisms . There have also been two types of approaches to computational models of neural circuits consisting of the amygdala and other brain regions . One approach is the firing-rate model , in which neural units represent the average firing rate of neurons , neural populations or brain regions [100–102] . Balkenius et al . [102] first developed a mathematical model in which fear memory was extinguished by inhibition of amygdalar activity by the orbitofrontal cortex , which is subdivision of the vmPFC , but their model did not focus on and hence failed to reproduce the PREE . Moustafa et al . [101] developed a neural circuit model consisting of the BLA , CEA , ITC and vmPFC ( i . e . , IL ) , combined with TD learning , in which synaptic plasticity was regulated by the TD error as a learning signal . In their study , however , the PREE was not explicitly addressed , though it was mentioned that their model exhibited the PREE only when extensive training trials were performed in the acquisition phase , with no detailed results . The other approach is based on a spiking-neuron model , in which action potentials are simulated based on membrane voltage dynamics . Nair’s group has extensively developed biophysically realistic conductance-based models that express several types of firing patters that have been experimentally observed [103–106] . These studies addressed fear conditioning only with full reinforcement , not with partial reinforcement , while investigating the roles of synaptic input from the vmPFC to the ITC [106] , interaction between prelimbic cortex ( PC ) in the mPFC and the BA [104] , and synaptic inputs from thalamus and cortex to the LA [105] . On the other hand , Vlachos et al . [107] first proposed a large-scale neural network model of the BA by introducing populations of fear , persistent and extinction neurons , but that work did not address the PREE . Compared with these previous models , our model is the first neural network model of the amygdala/vmPFC circuit that could satisfactorily explain the PREE , and it is based on the three types of neurons ( fear , persistent and extinction neurons ) . Finally , limitations of our model are worth mentioning . Recently , it has been reported that gradually reducing the frequency of the US ( called gradual extinction ) prevents the spontaneous recovery of fear memory when compared with full reinforcement of the US [108] . This fact suggested that not only frequency but also the temporal pattern of the US affects consolidation of extinction memory . However , our model cannot demonstrate the effect of gradual extinction on the spontaneous recovery of fear memory . This effect would be addressed by modeling leaky integration of learning signals , which we leave to our future work . Although the mPFC is divided into several subregions including the dorsal mPFC ( dmPFC ) and vmPFC , our model only addressed the vmPFC . In contrast to the vmPFC , the dmPFC plays an important role in the acquisition of fear memory [16 , 109] . It has been reported that sustained activity of the dmPFC is correlated with extinction failure , which should be related to resistance to extinction [110] . Moreover , a recent electrophysiological study of monkeys investigated activity in the amygdala and dorsal anterior cingulate cortex ( dACC ) , namely , the dmPFC in primates , during partial reinforcement fear conditioning and showed that correlated amygdala-dACC activity during fear conditioning determines resistance to extinction [26] . Further extension of the current model , possibly introducing additional fear neural units corresponding to the dmPFC , will be required . This study addressed only cued conditioning , not contextual conditioning , in which context information is provided to the BA through the hippocampus [111 , 112] . In the BA , fear and persistent neurons are unidirectionally innervated from and reciprocally interact with the ventral hippocampus , respectively , whereas extinction neurons have no connectivity with the hippocampus [28] . In addition , BA fear neurons activate excitatory neurons in the dmPFC , whereas the ventral hippocampus inhibits the dmPFC via innervating inhibitory neurons in the dmPFC [113] . These facts suggest that three types of neurons in different nuclei play differential roles in integrating cue and context information . Modeling such differential roles is also left to our future work .
|
Animals live in environments that contain uncertainty . To adapt to uncertain situations , they flexibly learn to associate environmental cues with rewards and punishments . Understanding how the brain processes uncertainty has remained an important issue in neuroscience . To address this question , we focused on neural processing in the amygdala and mPFC during fear conditioning and extinction . We developed a neural circuit model that incorporates distinct neural populations in the amygdala and mPFC . Our model first successfully reproduced uncertainty-dependent resistance to the extinction of fear memory . An extension of the model provided a possible explanation for observations made during optogenetic manipulation of the ventral mPFC . Finally , we proposed a procedure to accelerate the efficacy of subsequent extinction based on our model .
|
[
"Abstract",
"Introduction",
"Models",
"Results",
"Discussion"
] |
[
"learning",
"medicine",
"and",
"health",
"sciences",
"brain",
"social",
"sciences",
"neuroscience",
"learning",
"and",
"memory",
"fear",
"synaptic",
"plasticity",
"cognitive",
"psychology",
"cognition",
"memory",
"developmental",
"neuroscience",
"emotions",
"animal",
"cells",
"behavior",
"memory",
"consolidation",
"amygdala",
"fear",
"conditioning",
"behavioral",
"conditioning",
"cellular",
"neuroscience",
"psychology",
"cell",
"biology",
"anatomy",
"neurons",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"cognitive",
"science"
] |
2016
|
Uncertainty-Dependent Extinction of Fear Memory in an Amygdala-mPFC Neural Circuit Model
|
We here compared pathogenic ( p ) and non-pathogenic ( np ) isolates of Entamoeba histolytica to identify molecules involved in the ability of this parasite to induce amoebic liver abscess ( ALA ) -like lesions in two rodent models for the disease . We performed a comprehensive analysis of 12 clones ( A1–A12 ) derived from a non-pathogenic isolate HM-1:IMSS-A and 12 clones ( B1–B12 ) derived from a pathogenic isolate HM-1:IMSS-B . “Non-pathogenicity” included the induction of small and quickly resolved lesions while “pathogenicity” comprised larger abscess development that overstayed day 7 post infection . All A-clones were designated as non-pathogenic , whereas 4 out of 12 B-clones lost their ability to induce ALAs in gerbils . No correlation between ALA formation and cysteine peptidase ( CP ) activity , haemolytic activity , erythrophagocytosis , motility or cytopathic activity was found . To identify the molecular framework underlying different pathogenic phenotypes , three clones were selected for in-depth transcriptome analyses . Comparison of a non-pathogenic clone A1np with pathogenic clone B2p revealed 76 differentially expressed genes , whereas comparison of a non-pathogenic clone B8np with B2p revealed only 19 differentially expressed genes . Only six genes were found to be similarly regulated in the two non-pathogenic clones A1np and B8np in comparison with the pathogenic clone B2p . Based on these analyses , we chose 20 candidate genes and evaluated their roles in ALA formation using the respective gene-overexpressing transfectants . We conclude that different mechanisms lead to loss of pathogenicity . In total , we identified eight proteins , comprising a metallopeptidase , C2 domain proteins , alcohol dehydrogenases and hypothetical proteins , that affect the pathogenicity of E . histolytica .
The protozoan parasite Entamoeba histolytica is responsible for approximately 50 million cases of invasive amoebiasis per year , resulting in an annual death toll of 40 , 000–100 , 000 [1] . The parasite life cycle is relatively simple , comprising infectious cysts that can survive outside the host and vegetative trophozoites that proliferate in the human gut . After infection , E . histolytica trophozoites can asymptomatically persist for months or years in its human host [2] . Under as yet unknown circumstances , E . histolytica escapes from the gut lumen , either by penetrating the intestinal mucosa and inducing colitis , or by disseminating to other organs , most commonly the liver , where it induces abscess formation . The factors that determine the clinical outcomes of E . histolytica infections are not well understood . Possible factors comprise genetic make-up of the parasite and/or host , the immune response mounted by the host , concomitant infections and host diet . Identification of E . histolytica pathogenicity factors is a major topic in the field . Recently , research dealing with E . histolytica pathogenicity factors has mainly focused on a triad of protein families , namely , galactose/N-acetyl d-galactosamine–inhibitable Gal/GalNAc-lectins , cysteine peptidases ( CPs ) and amoebapores . Results obtained using transgenic amoebae supported the hypothesis that these molecules are involved in amoebic liver abscess ( ALA ) formation [3–6] . Nevertheless , homologues of the majority of these potential pathogenicity factors are also present in the non-pathogenic sister species Entamoeba dispar , a commensal protozoan that is genetically closely related to E . histolytica . Therefore , it remains to be shown whether one of these factors or their combination is responsible for amoeba pathogenicity or whether additional factors are involved . Thus , the mechanisms and processes enabling E . histolytica to penetrate host tissues and induce colitis and/or liver abscesses are still not understood . One straight-forward approach of identifying pathogenicity factors is a direct comparison of pathogenic and non-pathogenic E . histolytica isolates that has been performed using comparative microarray and proteome approaches [7–10] . Unfortunately , these studies used two isolates with completely different genetic backgrounds ( pathogenic isolate HM-1:IMSS and non-pathogenic isolate Rahman ) . This rendered the straight-forward identification of pathogenicity factors almost impossible . In addition , an in-depth phenotypical characterisation of the Rahman isolate revealed a number of genomic defects that presumably interfere with its virulence capacity [10] . Recently , we identified two cell lines that were both derived from the clinical E . histolytica isolate HM-1:IMSS but which significantly differ in their pathogenicity . Whereas cell line HM-1:IMSS-A completely lost its ability to induce ALAs in gerbils ( Meriones unguiculatus ) and mice ( Mus musculus ) , cell line HM-1:IMSS-B is highly aggressive and induces large ALAs . Comparative transcriptomic and proteomic studies of these cell lines have already been performed [11] . The studies revealed 31 differentially ( ≥3-fold ) expressed genes [12] and 31 proteins with differential abundances in HM-1:IMSS-A and HM-1:IMSS-B [11] . However , an overlap of only two molecules was found between the proteomic and transcriptomic approaches . Until now , neither pathogenic nor non-pathogenic cell lines have been cloned , and therefore it is possible that they consist of a mixture of cells . Thus , in the present study , both cell lines were cloned to obtain homogenous cell populations , allowing analyses of the pathogenicity factors in a straight-forward fashion . Twelve clones derived from the non-pathogenic cell line HM-1:IMSS-A ( A1–A12 ) and 12 clones derived from the pathogenic cell line HM-1:IMSS-B ( B1–B12 ) were generated . All 24 clones were analysed for their ability to induce ALAs in gerbils and for their specific CP activity . One non-pathogenic A-clone ( A1np ) , one pathogenic B-clone ( B2p ) and one non-pathogenic B-clone ( B8np ) were analysed in more detail , including their time course of ALA formation , growth , size , motility , haemolytic activity , erythrophagocytosis , cytopathic activity and transcriptome profiles . Furthermore , transfectants overexpressing genes that were identified as differentially expressed between the pathogenic and non-pathogenic clones were tested for their involvement in ALA formation .
To analyse whether the E . histolytica cell lines consisted of a mixture of different cell types with different pathogenic phenotypes , the cell lines were cloned by limited dilution method . This resulted in 12 clones derived from cell line HM-1:IMSS-A and 12 clones derived from cell line HM-1:IMSS-B . The ALAs-generating ability of the different clones was analysed using the gerbil model . The animals were sacrificed 7 days post infection , and ALA sizes were determined . The results clearly indicated that the HM-1:IMSS-A cell line consists of a homogenous cell population . Except a few cases of small ALA formation , the majority of animals infected with different A-clones showed no ALA formation ( Fig 1 ) . The results were more divergent for clones derived from the HM-1:IMSS-B cell line . Eight clones , B2–B7 , B9 and B10 , showed a pathogenic phenotype comparable with the original cell line HM-1:IMSS-B . However , although clones B1 , B8 , B11 and B12 were derived from the pathogenic cell line , their ability to induce abscess formation was significantly reduced . This was especially evident for clone B8 that did not induce any abscess formation 7 days post infection ( Fig 1 ) . The non-pathogenic clones A1np and B8np and pathogenic clone B2p have been continuously cultivated for more than 5 years , without any change of the respective phenotypes . The pathogenic phenotype of clone B2p remained especially stable over the years without the need for animal passaging . To ensure that the observed phenotypes were indeed stable and uniform , the non-pathogenic clone B8np and the highly pathogenic clone B2p were sub-cloned . All sub-clones showed the same phenotype as the respective mother clone . All five sub-clones derived from the non-pathogenic clone B8np were unable to induce ALAs , whereas all five sub-clones of the pathogenic clone B2p produced large abscesses ( Fig 2 ) . Recently , it was shown that CP activity of the pathogenic cell line HM-1:IMSS-B is approximately ten times greater ( 110 ± 25 mU/mg ) than in the non-pathogenic cell line HM-1:IMSS-A ( 15 ± 5 mU/mg ) [11] . A similar difference has been measured for the non-pathogenic clone A1np ( 15 ± 10 mU/mg ) and the pathogenic clone B2p ( 123 ± 60 mU/mg ) [3] . To investigate if the observed correlation between CP activity and pathogenicity is generally valid , the activities of all A- and B-clones were determined and correlated with ALA formation ( Fig 3A ) . In general , the clones derived from HM-1:IMSS-B had a significantly higher CP activity ( 85 ± 50 mU/mg ) than clones derived from HM-1:IMSS-A ( 18 ± 11; p < 0 . 0001 ) . However , a direct correlation between the CP activity and ALA formation was not observed . This was obvious especially for clones B1 and B12 . Although these clones have a high CP activity ( 198 ± 52 mU/mg and 139 ± 89 mU/mg , respectively ) , they only induce small ALAs ( Figs 2 and 3A ) . CPs EhCP-A1 , EhCP-A2 , EhCP-A4 , EhCP-A5 and EhCP-A7 can be visualised by substrate gel electrophoresis [3] . Here , the results from substrate gel experiments clearly indicated that different CP activities of the various A- and B-clones are not linked to the expression of a single peptidase ( Fig 3B ) . To identify the underlying mechanisms of different virulence phenotypes , three clones were selected for in-depth analyses . These were as follows: a non-pathogenic clone derived from HM-1:IMSS-A ( clone A1np ) , pathogenic clone B2p that induced the largest ALAs among the B-clones , and clone B8np that completely lost its ability to produce ALAs . Both B-clones have been derived from HM-1:IMSS-B . Magnetic resonance imaging ( MRI ) was employed to follow post-infection abscess formation over time in more detail . Infection time course was analysed over 10 days in gerbil and mouse ALA models using the three clones , A1np , B2p and B8np ( Fig 4A and 4B ) . Our findings clearly confirmed the results of animal experiments described above ( Fig 1 ) . On day 7 post infection , no or very small ALAs were detected in both animal models with the two non-pathogenic clones A1np and B8np , whereas unequivocal abscess formation was observed with clone B2p ( Fig 4A and 4B ) . Nevertheless , it became apparent that clone A1np was also able to induce abscess formation initially , as lesions were detected on day 3 post infection . However , these ALAs were smaller compared with ALAs seen during clone B2p infection and were more rapidly resolved . In contrast to clone A1np , the pathogenicity of clone B8np was almost completely abolished . No ALAs were detected in gerbils infected with this clone , while , in mice , ALAs on day 3 post infection were significantly smaller compared with ALAs induced by clone B2p ( Fig 4A and 4B ) . Clones A1np , B2p and B8np were then phenotypically characterised . This included determination of size , growth rate , haemolytic activity , erythrophagocytosis and cytopathic activity . Microscopic analyses indicated that trophozoite sizes of clone B2p ( 818 ± 235 μm2 ) and clone B8np ( 782 ± 193 μm2 ) differed significantly and were larger in comparison with cells from clone A1np ( 591 ± 154 μM2 ) ( Table 1 ) . Clone A1np grew significantly slower in comparison with clone B2p and clone B8np . The doubling time of clone A1np was approximately 12 ± 2 . 7 h , whereas it was approximately 8 ± 1 . 6 h for clone B2p and 9 . 5 ± 2 . 2 h for clone B8np ( Table 1 ) . With an accumulated distance of 228 ( ±156 ) μm/10 min the amoebae of clone A1np move significantly slower in comparison to amoebae of clone B2p and clone B8np ( 376 ±172 μm/10 min ( p < 0 . 0001 ) and 453 ±184 μm/10 min ( p < 0 . 0001 ) , respectively ) . While clone B8np moved significantly faster than B2p ( p < 0 . 0013 ) ( Table 1 ) . Clone A1np and clone B2p were able to lyse erythrocytes , but no haemolytic activity was detected for clone B8np . In addition , no correlation of haemolytic activity with pathogenicity was observed , since the non-pathogenic clone A1np had the highest activity ( Table 1 ) . By contrast , clone B8np displayed the highest erythrophagocytosis rate , followed by clone A1np and clone B2p . Cytopathic activity ( percentage of monolayer disruption ) was highest for clone A1np followed by clone B2p . Clone B8np was unable to disrupt a cell monolayer ( Table 1 ) . RNAseq experiments were performed to identify differences in gene expression profiles of clones A1np , B2p and B8np . Comparison of the non-pathogenic clone A1np and pathogenic clone B2p revealed 76 differentially expressed genes ( threshold ≥ 3-fold , p-value adjusted ( padj ) < 0 . 05 ) . Some genes ( 46 ) were expressed more highly in clone A1np and some ( 30 ) in clone B2p ( Tables 2 and 3 , S2 Table ) . From the 46 genes with higher expression levels in clone A1np 10 code for surface proteins ( EHI_015290 , EHI_082070 , EHI_118130 , EHI_169280 , EHI_074080 , EHI_075660 EHI_164900 , EHI_039020 , EHI_006170 , EHI_086540 ) [13] . Amongst them are 2 members of the C2 domain protein family and 2 members of the Rab family . Since the analysis of the surface proteome referred to was performed with trophozoites of cell line A , it was not surprising that genes with higher expression levels in clone B2p could not be identified as surface associated [13] . The greatest differences , with expression fold changes ~20–200 , concerned genes encoding three C2 domain proteins , three Rab family GTPases , cell surface protease gp63 and one hypothetical protein ( EHI_074080 ) , whose expression was higher in clone A1np; and five genes encoding hypothetical proteins EHI_127670 , EHI_144490 , EHI_169670 , EHI_050490 and EHI_062080 , with higher expression in clone B2p . The majority of the identified genes showed 3-4-fold differential expression ( Tables 2 and 3 ) . The detected differential expression was verified by quantitative real-time PCR ( qPCR ) for 26 genes . Results of next generation sequencing were confirmed for all the analysed genes , with the exception of EHI_056490 ( threshold ≥ 2 fold , Table 4 ) . Most of the identified genes encode proteins with unknown function . Some of these hypothetical proteins contain functional domains , e . g . , C2 domains , phosphatase domains , tyrosine kinase domains , RecF/RecN/SMC domains and lecithin:cholesterol acyltransferase domains . Proteins with known functions up-regulated in clone A1np , when compared with clone B2p , mainly included Rab family proteins , peptidases and heat shock proteins . Putative function could be assigned to only 6/30 genes identified as more highly expressed in clone B2p in comparison with clone A1np ( e . g . , tyrosine kinase , methionine gamma-lyase , thioredoxin ) ( Tables 2 and 3 ) . Comparisons of the two B-clones , B2p and B8np , revealed only 19 differentially expressed genes . Twelve genes were expressed at higher levels in clone B8np in comparison with clone B2p , and seven genes were expressed more highly in B2p in comparison with B8np ( Tables 5 and 6 , S3 Table ) . The corresponding proteins assigned to EHI_039020 and EHI_088020 were found to be part of the surface proteome of E . histolytica [13] . Fold change ≥ 10 was detected for only three genes . These genes encoded two hypothetical proteins and a leucine-rich repeat-containing protein . All these genes were more highly expressed in B2p in comparison with B8np . As observed in the A1np/B2p comparison , the majority of the identified genes showed 3-4-fold differential expression ( Tables 5 and 6 ) . The majority ( 11/19 ) of the genes encoded hypothetical proteins . The remaining genes were annotated as galactose-inhibitable lectin 35 kDa subunit , phosphoserine aminotransferase , actobindin , alcohol dehydrogenase , AIG1 family protein and methionine gamma-lyase . However , only galactose-inhibitable lectin 35 kDa subunit and methionine gamma-lyase have been biochemically characterised in E . histolytica [14–17] . Interestingly , only six genes in the two non-pathogenic clones A1np and B8np were similarly regulated vs . pathogenic clone B2p . Genes EHI_026360 , EHI_039020 and EHI_056490 were up-regulated , and genes EHI_127670 , EHI_144490 and EHI_144610 were down-regulated , in clones A1np and B8np in relation to clone B2p . This suggests different mechanisms accounting for the inability of clones A1np and B8np to induce ALAs . In total 89 genes were found to be differentially expressed between clone A1np and clone B2p and/or between clone B8np and clone B2p . In a previous study , the transcriptomes of the non-clonal cell lines A and B were compared using a microarray approach [12] . Here , in total 31 genes were differentially expressed ( threshold ≥3-fold ) . Of the 12 genes with higher expression levels in cell line B in comparison to cell line A , 7 genes had also higher expression levels in clone B2p in comparison to clone A1np . Out of the 19 genes with higher expression levels in cell line A in comparison to cell line B , 11 genes showed also higher expression in clone A1np in comparison to clone B2p ( S4 Table ) . Only 8 of the identified 89 genes encoded for proteins containing a signal peptide and 15 genes encoded for proteins containing between 1–7 transmembrane domains ( S4 Table ) . In a previous study in which the surface proteome of cell line A was analysed , 693 putative surface-associated proteins were identified [13] . Out of them 11 showed differential expression between the different clones ( S4 Table ) . From the 89 identified genes , 32 encode hypothetical proteins , where no homology to other proteins or protein domains could be identified . Furthermore , 3 genes encode for proteins of the C2 superfamily , 4 genes encode for members of the small GTPase superfamily , 10 genes encode for heat shock proteins , 6 genes encode for AIG1 family proteins , 3 genes encode for kinases 2 genes encode for cysteine synthases and 2 genes encode for proteases . Additional 18 genes encode for proteins with other known functions ( S4 Table ) . To investigate whether the differentially expressed genes play a role in ALA formation , their respective overexpressing transfectants were generated . For genes that were more highly expressed in clone A1np in comparison with clone B2p , 13 overexpressing transfectants of clone B2p were generated . These included eight genes that displayed the highest differential expression ( >20-fold ) and three genes that were also up-regulated in clone B8np ( EHI_026360 , EHI_056490 , EHI_039020 ) ( Tables 2 and 5 ) . We were unable to generate transfectants overexpressing genes EHI_169280 ( ehrab7e ) , EHI_074080 , EHI_187090 ( ehrab7g ) and EHI_075660 ( ehcaax ) . For all other genes , a relative 2 . 4–235-fold overexpression was obtained ( S5 Table ) . The pathogenic phenotype of the majority of B2p transfectants overexpressing genes that were more highly expressed in the non-pathogenic clone A1np vs . pathogenic clone B2p was unaffected . These also included transfectants that overexpressed three genes regulated in the same manner in clones A1np and B8np . All these B2p transfectants induced ALA formation in mice . Four genes were identified whose overexpression had a dramatic impact on ALA formation . When EHI_015290 ( ehc2-3 ) , EHI_059860 ( ehc2-5 ) , EHI_042870 ( ehmp8-2 ) or EHI_075690 were overexpressed in clone B2p , these clones lost their pathogenic phenotype and produced significantly smaller ALAs than the respective controls ( Fig 5A ) . B2p transfectants were generated for 9/12 genes that were expressed at higher levels in clone B8np in comparison with clone B2p . They showed 4–300-fold increased expression in comparison with the control ( S5 Table ) . Overexpression of genes EHI_058920 , EHI_088020 and EHI_160670 significantly reduced pathogenicity of clone B2p ( Fig 5A ) . In silico analyses indicated that nucleotide sequences of EHI_088020 and EHI_160670 were identical and that the genes encode an alcohol dehydrogenase . However , the first 480 nucleotides of the EHI_088020 coding region were missing from the EHI_160670 sequence . This may be because the E . histolytica genome is not yet fully annotated ( AmoebaDB , http://amoebadb . org/amoeba/ ) . Regardless , increased expression of the full-length or truncated gene impacted abscess formation ( Fig 5A ) . For genes expressed more highly in clone B2p in comparison with clone A1np , five gene-overexpressing clone A1np transfectants were generated , including three genes with an initially detected >90-fold differential expression . Overexpression of EHI_169670 was unsuccessful; however , for all other genes a relative 4–490-fold expression was obtained ( S6 Table ) . Overexpression of these genes did not significantly affect ALA formation . However , strikingly , 4/9 mice infected with EHI_127670-overexpressing A1np transfectant produced large ALAs ( Fig 5B ) . Seven genes were more highly expressed in clone B8np in comparison with clone B2p . Two , EHI_144610 and EHI_057550 , were identical and encode methionine gamma-lyase . Four B8np transfectants were generated , and all of them showed 3–70-fold overexpression in comparison with the control ( S7 Table ) . Similarly to A1np transfectants , no significant influence on ALA formation was observed . Interestingly , large ALAs were detected 7 days post infection in 6/9 mice infected with B8np transfectant overexpressing EHI_127670 ( p = 0 . 0589 ) , as was observed for infections with A1np-EHI_127670 transfectants ( p = 0 . 292 ) ( Fig 5B and 5C ) .
In this study , no correlation was found between the ability of E . histolytica clones to produce amoebic liver abscesses and their cysteine protease , haemolytic , erythrophagocytosis , or cytopathic activities , or their sizes or growth characteristics . However , the clones showed different expression profiles . We conclude that different mechanisms result in the loss of E . histolytica pathogenicity , because only a few genes were found to be differentially regulated in the same way when either of the two non-pathogenic clones A1np and B8np were compared with the pathogenic clone B2p . However , overexpression of seven different genes , encoding a metallopeptidase , C2 domain proteins , alcohol dehydrogenases , and hypothetical proteins in the pathogenic clone B2p correlated with reduced ability of E . histolytica to produce amoebic liver abscesses . Only one gene was identified whose overexpression transformed a non-pathogenic phenotype into a pathogenic one .
Animal experiments were carried out in accordance with the guidelines from the German National Board for Laboratory Animals and ARRIVE guidelines ( https://www . nc3rs . org . uk/arrive-guidelines ) and approved by the review board of the State of Hamburg , Germany ( Ministry of Health and Consumer Protection/Behörde für Gesundheit und Verbraucherschutz—ethical permits 145/13 , 20 . 01 . 2014 ) E . histolytica trophozoites were cultured axenically in TYI-S-33 medium in plastic tissue culture flasks [65] . E . histolytica cell lines HM-1:IMSS-A and HM-1:IMSS-B were derived from the isolate HM-1:IMSS and both were originally obtained from the American Type Culture Collection ( ATCC ) under the catalogue number 30459[11] . HM-1:IMSS was originally isolated from a colonic biopsy of rectal ulcer from an adult male patient with amoebic dysentry in 1967 ( Mexico City , Mexico ) . The monoxenic cultured HM-1:IMSS isolate was passed from Margarita de la Torre to Louis S . Diamond who adapted it to axenic cultivation . Thereafter , this axenically cultivated HM-1:IMSS isolate was transferred to the ATCC library . Cell line A was sent to us in 2001 by Barbara Mann ( Charlottesville , University of Virginia ) , as a batch of cells from the same culture that was used for DNA preparation to sequence the E . histolytica genome [66] . The pathogenic cell line B was obtained directly from ATCC in 1991 . Since then , the ability of cell line B to induce liver pathology remained stable . Both cell lines were cloned by limited dilution . For this , a dilution of 120 amoebae/24 ml TYI-S-33 medium was prepared and 200 μl of this dilution was added to each well of a 96-well plate . The presence of only one amoebae/well was analysed microscopically and the trophozoites were cultivated under anaerobic conditions using Anaerocult ( Merck ) for one week . Afterwards the clones were transferred for further cultivation to tissue culture flasks . For individual experiments , 1 × 106 trophozoites were cultivated for 24 h in 75 mL culture flasks . Subsequently , after chilling on ice for 5 min , trophozoites were harvested by sedimentation at 430 × g at 4°C for 5 min . The resulting cell pellets were washed twice either in phosphate-buffered saline ( PBS; 6 . 7 mM NaHPO4 , 3 . 3 mM NaH2PO4 , 140 mM NaCl , pH 7 . 2 ) or in incomplete TYI-S-33 medium ( medium without serum ) . To prepare amoeba extracts , cells were lysed over four freeze-thaw cycles in CO2/ethanol and sedimented by centrifugation ( 9000 × g at 4°C for 15 min ) . Animal infections were performed with 10- to 12-week-old female gerbils obtained from JANVIER LABS ( Saint Berthevin Cedex 53941 France ) or with 10- to 12-week-old C57BL/6 male mice bred in the animal facility of the Bernhard Nocht Institute for Tropical Medicine , Hamburg , Germany . All animals were maintained in a specific pathogen-free environment . Animal experiments were approved by the review board of the State of Hamburg , Germany ( Ministry of Health and Consumer Protection/Behörde für Gesundheit und Verbraucherschutz , ( 145/13 , 20 . 01 . 2014 ) and conducted in accordance with institutional and ARRIVE guidelines ( https://www . nc3rs . org . uk/arrive-guidelines ) . For gerbil infections , 1 × 106 trophozoites in 50 μL of incomplete TYI-S-33 medium ( without serum ) were injected into the left liver lobe , as described previously [67] . For mice infection experiments , 1 . 25 × 105 trophozoites in 25 μL of incomplete TYI-S-33 medium were injected into the liver , as described by Lotter and colleagues [23] . To analyse ALA formation of the various amoeba clones , gerbils were sacrificed at 7 days post infection and the extent of the abscessed liver area was measured manually using a caliper and determined as size in mm2 . For each E . histolytica clone , ALA formation was analysed in at least four animals . Significance ( p-values ) was established using the Mann-Whitney U test . MRI was performed to analyse the time course of ALA formation using a small animal 7 Tesla MR scanner ( ClinScan , Bruker Biospin GmbH , Ettlingen , Germany ) . For these experiments , gerbil and mouse livers were imaged in vivo on days 3 , 5 , 7 and 10 after intrahepatic injection of E . histolytica . Anaesthesia was performed as described by Ernst and colleagues [68] . Images were acquired using T2-weighted fast spin echo ( T2w FSE ) sequences for high-resolution anatomical reference . Total abscess volume was calculated by measuring the region of interest ( ROI ) in each abscess-containing slice , using transversal sections of the abdomen and the OsiriX Imaging Software DICOM Viewer ( Open-source version 32-bit 4 . 1 . 1 ) . Significance ( p-values ) was established using an unpaired t test . E . histolytica trophozoites ( 1 × 106 ) were cultivated in 75 mL culture flasks for 24 h . The cells were harvested after chilling on ice for 5 min and sedimented at 200 × g for 5 min at 4°C . Cell pellets were washed twice with PBS . To isolate total RNA , trophozoites were treated with TRIzol reagent ( Thermo Fisher Scientific , Schwerte , Germany ) following the manufacturer’s instructions . Extracted RNA was further purified using the RNeasy mini kit ( Qiagen , Hilden , Germany ) but without β-mercaptoethanol in buffer RLT , and DNA was digested with DNase I ( Qiagen , Hilden , Germany ) . Total RNA for transcriptomics analyses was purified using the mirVana miRNA isolation kit ( Ambion-Thermo Fisher Scientific , Schwerte , Germany ) , according to the manufacturer’s instructions . cDNA synthesis was obtained using the SuperScript III Reverse Transcriptase system ( Thermo Fisher Scientific , Schwerte , Germany ) . Briefly , RNase-free and DNase-treated total RNA ( 1 μg in a 20 μL final volume ) was mixed with 5 × First-Strand buffer , 1 mM dNTPs , 500 nM OdT-T71 ( 5′-GAG AGA GGA TCC AAG TAC TAA TAC GAC TCA CTA TAG GGA GAT24 ) , 2 mM DTT , 0 . 5 mM MgCl2 , 40 U RNaseOut ( Thermo Fisher Scientific , Schwerte , Germany ) and SuperScript III ( 200 U/μL ) . cDNA was synthesised for 1 h at 42°C . For qRT-PCR experiments , sense and antisense primers were designed to amplify 100–120 bp fragments of the respective genes ( S8 and S9 Tables ) . Quantitative amplifications were performed in Rotor-Gene PCR apparatus ( Qiagen , Hilden , Germany ) using RealMasterMix SYBR ROX Kit ( 5PRIME , Hilden , Germany ) . cDNA ( 1 μL ) was mixed with 2 . 5 × RealMasterMix/20 × SYBR and 5 pmol/μL appropriate sense and antisense primers , to a final 20 μL volume . Amplification conditions were as follows: 40 cycles of 95°C for 15 s , 58°C for 20 s and 68°C for 20 s , and an adjacent melting step ( 67–95°C ) . Two biological replicates were analysed in duplicate . Relative differences in gene expression were calculated using the 2-ΔΔCT method with Rotor-Gene software [69] . Depending on the experiment , clone A1np , clone B2p , or clone B8np was used as the calibrator ( = 1 ) , and actin was used as the house-keeping gene for normalisation . RNA for RNA-Seq library preparation was purified as described above . RNA quantity and quality were evaluated spectrophotometrically ( NanoDrop 2000 , Thermo Fisher Scientific , Schwerte , Germany ) and with an Agilent 2100 Bioanalyzer with RNA 6000 Pico Assays kit ( Agilent Technologies , Waldbronn , Germany ) . Samples were Turbo DNase-treated using TURBO DNA-free Kit ( Ambion-Thermo Fisher Scientific , Schwerte , Germany ) . After quality control , rRNA was depleted using RiboZero Magnetic Gold kit ( Human/Mouse/Rat; Epicentre-Illumina , Munich , Germany ) and Agencourt RNAClean XP kit ( Beckmann Coulter , Krefeld , Germany ) , according to the manufacturers’ protocol . RNA-Seq libraries were then generated using ScriptSeq v2 kit ( Epicentre-Illumina , Munich , Germany ) according to the manufacturer’s instructions . Each library was indexed with Illumina-compatible barcodes to allow multiplexing . The individual libraries were assessed using Qubit dsDNA high sensitivity kit and Bioanalyzer DNA HS chips to ascertain the concentration ( 4 nM ) and fragment size distribution , respectively , prior to library multiplexing . Libraries were denatured and diluted to final concentration of 8 pM for sequencing on the MiSeq platform following the manufacturer’s instructions . Reads were aligned to E . histolytica transcriptome ( AmoebaDB 28 , released 30 March 2016 ) using Bowtie 2 version 2 . 2 . 3 [70] and differential expression was analysed using DESeq [71] version 1 . 18 . To determine amoeba size , the circumference of 80 trophozoites of each clone was measured using a BZ9000 Keyence microscope ( Keyence , Neu-Isenburg , Germany ) . To determine growth rate , 500 trophozoites of each clone were seeded into a 24-well plate and the cells were counted every 24 h over 72 h . The growth rate was determined three times in triplicate for each clone . The movement of the amoebae was directly filmed over a time period of 10 min using Evos FL Auto microscope from Life Technologies . A picture was taken every 5 sec . The movement of 80 amoebae/ clone was analysed manually using ImageJ version 2 . 0 . 0-rc-43/1 . 51d with plugins for manual tracking and chemotaxis . Significance ( p-values ) was established using the Mann-Whitney U test . CP activity was measured using the synthetic peptide Z-Arg-Arg-pNA ( Bachem , Bubendorf , Switzerland ) as substrate [72] . One unit of enzymatic activity is defined as the amount of protein that catalyses the generation of 1 μmoL p-nitroaniline per min . Substrate gel electrophoresis was performed as described previously [73] . Briefly , amoebic extract ( 2 μg ) was incubated in Laemmli buffer with 20 mM DTT , for 5 min at 37°C . For the substrate gel , 12% SDS-polyacrylamide gel was co-polymerised with 0 . 1% gelatine . After protein separation and incubation in buffer A ( 2 . 5% Triton X-100; 1 h ) and buffer B ( 100 mM sodium acetate , pH 4 . 5 , 1% Triton X-100 , 20 mM DTT; 3 h ) , at 37°C , the gel was stained with Coomassie blue . Erythrophagocytosis assay was performed as described by Biller and colleagues [11] . Human 0+ erythrocytes were provided by the blood bank of the University Medical Center Hamburg-Eppendorf ( UKE ) –Transfusion Medicine–Germany . Human erythrocytes and trophozoites were washed twice with serum-free TYI-S-33 medium . Erythrocytes and amoebae were mixed at a 1000:1 ratio ( 2 × 108 erythrocytes , 2 × 105 amoebae ) , to a final volume of 400 μL , in serum-free TYI-S-33 medium and incubated in parallel at 37°C for 30 min . To stop phagocytosis and lyse non-phagocytosed erythrocytes , 1 mL of distilled water was added , twice . Trophozoites were washed twice with PBS . Average numbers of ingested erythrocytes were quantified by measuring the absorbance at 397 nm after trophozoite lysis in 90% formic acid . The experiment was performed three times in triplicate . Significance ( p-values ) was established using the Mann-Whitney U test . Haemolytic activity assay was performed as described by Biller and colleagues [11] . Human erythrocytes and trophozoites were washed three times with PBS . The assay was performed by mixing trophozoites and erythrocytes in a 1:2000 ratio ( 2 × 105 amoebae with 4 × 108 erythrocytes per mL of PBS ) , followed by incubation for 1 h at 37°C . After incubation , the cells were sedimented for 1 min at 2000 × g . Haemoglobin released into the supernatant was measured at 570 nm in a spectrophotometer . Separately incubated erythrocytes and trophozoites were used as negative controls . To determine 100% haemoglobin release , 4 × 108 erythrocytes were lysed in 1 mL of water . The experiment was performed three times in triplicate . Significance ( p-values ) was established using the Mann-Whitney U test . Interaction of trophozoites and Chinese hamster ovarian ( CHO ) cells was determined by a modified method of Bracha and Mirelman [74] . CHO cells defective in glycosaminoglycan biosynthesis ( CHO-745; American Type Culture Collection No . CRL-2242 ) were used . CHO cells ( 1 × 105 per well ) were grown for 24 h in 24-well plates in Ham’s F12 ( with l-Glutamine ) medium supplemented with 10% fetal calf serum ( FCS ) and penicillin-streptomycin . After washing the CHO cells with preheated ( 37°C ) Ham’s medium , 500 μL of Ham’s medium was added . E . histolytica trophozoites ( 1 × 105 ) were washed twice with serum-free TYI-S-33 medium , resuspended in 500 μL of ABS-free TYI-S-33 and added to the CHO cells . The mixture was incubated for 20 min at 37°C under 5% CO2 . Cells were washed with 1 mL of ice-cold PBS and treated with 0 . 5 mL of 4% paraformaldehyde in PBS for 2 min . After another PBS wash , the cells were stained with 500 μL of 0 . 1% methylene blue for 2 min . Finally , the cells were washed with 0 . 01% methylene blue and PBS . Cells were lysed with 1 mL of 0 . 1 M HCl for 30 min at 37°C . Samples were photometrically analysed at 660 nm . As a control , methylene blue concentration was determined for CHO cells that had not been co-cultivated with trophozoites ( i . e . , no destruction of cell monolayer ) . Experiments were performed three times in sextuplicate . Significance ( p-values ) was established using the Mann-Whitney U test . All plasmids used for E . histolytica trophozoite transfections are derivatives of the expression vector pEhNEO/CAT ( pNC ) [75 , 76] . Genes of interest were amplified by PCR using genomic E . histolytica DNA as a template , cloned into TOPO TA vector , sequenced and cloned into pNC using KpnI and BamHI restriction sites ( S9 Table ) . For overexpression , coding sequences of the genes of interest were flanked by 485 bp 5′-untranslated sequence of the E . histolytica lectin gene and 600 bp 3′-untranslated region of the actin gene . Neomycin phosphotransferase was used as a selectable marker . Transfections were performed by electroporation as described previously [76] . Two days post transfection , cells were transferred to a selection medium containing 10 μg/mL G-418 sulphate , for approximately 2 weeks . Subsequently , the cells were cloned by a limited dilution method and cultivated in the presence of 20 μg/mL G418 . Successful overexpression of at least four clones was checked by qRT-PCR . For infection experiments , trophozoites were cultivated for 24 h in the absence of G418 .
|
The pathogen Entamoeba histolytica can live asymptomatically in the human gut , or it can disrupt the intestinal barrier and induce life-threatening abscesses in different organs , most often in the liver . The molecular framework that enables this invasive , highly pathogenic phenotype is still not well understood . In order to identify factors that are positively or negatively correlated for invasion and destruction of the liver , we used a unique tool , E . histolytica clones that differ dramatically in their pathogenicity , while sharing almost identical genetic background . Based on comprehensive transcriptome studies of these clones , we identified a set of candidate genes that are potentially involved in pathogenicity . Using ectopic overexpression of the most promising candidates , either in pathogenic or in non-pathogenic Entamoeba clones , we identified genes where high expression reduced pathogenicity and only one gene that increased pathogenicity to a certain extend . Taken together , the current study identifies novel pathogenicity factors of E . histolytica and highlights the observation that various different genes contribute to pathogenicity .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"trophozoites",
"medicine",
"and",
"health",
"sciences",
"parasite",
"groups",
"pathology",
"and",
"laboratory",
"medicine",
"enzymes",
"enzymology",
"cloning",
"parasitic",
"protozoans",
"parasitology",
"apicomplexa",
"protozoans",
"signs",
"and",
"symptoms",
"molecular",
"biology",
"techniques",
"research",
"and",
"analysis",
"methods",
"abscesses",
"entamoeba",
"histolytica",
"proteins",
"gene",
"expression",
"pathogenesis",
"molecular",
"biology",
"biochemistry",
"diagnostic",
"medicine",
"genetics",
"protein",
"domains",
"biology",
"and",
"life",
"sciences",
"proteases",
"organisms"
] |
2016
|
Overexpression of Differentially Expressed Genes Identified in Non-pathogenic and Pathogenic Entamoeba histolytica Clones Allow Identification of New Pathogenicity Factors Involved in Amoebic Liver Abscess Formation
|
While most processes in biology are highly deterministic , stochastic mechanisms are sometimes used to increase cellular diversity . In human and Drosophila eyes , photoreceptors sensitive to different wavelengths of light are distributed in stochastic patterns , and one such patterning system has been analyzed in detail in the Drosophila retina . Interestingly , some species in the dipteran family Dolichopodidae ( the “long legged” flies , or “Doli” ) instead exhibit highly orderly deterministic eye patterns . In these species , alternating columns of ommatidia ( unit eyes ) produce corneal lenses of different colors . Occasional perturbations in some individuals disrupt the regular columns in a way that suggests that patterning occurs via a posterior-to-anterior signaling relay during development , and that specification follows a local , cellular-automaton-like rule . We hypothesize that the regulatory mechanisms that pattern the eye are largely conserved among flies and that the difference between unordered Drosophila and ordered dolichopodid eyes can be explained in terms of relative strengths of signaling interactions rather than a rewiring of the regulatory network itself . We present a simple stochastic model that is capable of explaining both the stochastic Drosophila eye and the striped pattern of Dolichopodidae eyes and thereby characterize the least number of underlying developmental rules necessary to produce both stochastic and deterministic patterns . We show that only small changes to model parameters are needed to also reproduce intermediate , semi-random patterns observed in another Doli species , and quantification of ommatidial distributions in these eyes suggests that their patterning follows similar rules .
The development of multicellular animals is highly reproducible , with deterministic and orderly processes generating reliable outcomes . Segment boundaries form in the proper place and cell types are set aside in specific proportions in differentiating tissues . Underlying these seemingly precise developmental outcomes , though , are inherently stochastic transcriptional events , e . g . decisions to express or not express key regulators of cell fate [1 , 2] . Varying amounts of activating or repressive input can bias these decisions strongly one way or the other , producing seemingly deterministic on or off outcomes , resulting in distinct boundaries and specific spatial patterns [3 , 4] . The distribution of these inputs depends largely on lineage and positional information within an embryo or tissue . In another class of cell fate decisions , stochastic cell-intrinsic mechanisms instead produce a particular probability of taking one fate or another in otherwise equivalent cells [5] . In their own way , these stochastic decisions are highly regulated to take place in specific tissue types and to produce reliable proportions of one cell fate vs . another . How such probabilistic patterning mechanisms might be switched between stochastic and deterministic is a question to which the tools of statistical physics can meaningfully be applied . An example of stochastic patterning occurs in the Drosophila eye [5 , 6] , a complex organ whose development has been the subject of great scrutiny [7 , 8] . Our interest focuses on the patterning of photoreceptors ( PRs ) that are involved in color vision: two types of ommatidia express different combinations of color-sensitive photopigments Rhodopsins in their “inner” R7 and R8 PRs , and these two ommatidial types are randomly distributed across the retina [9] . This can be visualized via staining with antibodies against the green-sensitive ( Rh6 ) or blue-sensitive ( Rh5 ) Rhodopsins expressed in R8 PR cells ( Fig 1a ) . A similar stochastic pattern exists in R7 PR cells for two UV Rhodopsins , Rh3 and Rh4 [10] . Stochastic on or off expression of the transcription factor Spineless in the R7 PRs controls ommatidial type , and therefore the overall random pattern [5] . In contrast , another group of flies in the family Dolichopodidae ( referred to here as “Doli” ) have ordered retinal patterning with alternating columns of ommatidia ( the individual units of the adult compound eye ) that produce two distinct corneal lens colors ( Fig 1b ) . The patterning mechanisms that underlie both differentiation of PR types ( e . g . R7 vs . R8 ) and stochastic patterning across ommatidia have been shown to be largely conserved between Drosophila and butterflies [11 , 12] . Considering the apparent similarities between the Drosophila and Doli eye , it is tempting to suggest that the cell fate decisions involved in stochastic vs . non-stochastic patterning may share the same underlying regulatory mechanisms with similar downstream effectors , but which differ in how the expression of few critical upstream regulators is controlled . Thus , it might be possible to predict the rules that underlie the control of stochastic vs . deterministic patterning , and might underlie the evolutionary conversion from one mode of patterning to the other . In this work , we present a simple mathematical model for such a regulatory mechanism , and compare our results with experimental data from three fly species . Our model is also predictive and applicable to patterns observed in the eyes of other flies; as an example , we present predictions for the eyes of another species of Dolichopodidae that displays intermediate patterns . The adult eye is a geometrically regular structure composed of hexagonal unit eyes packed into a grid . Patterning begins with the progression of the morphogenetic furrow , a posterior-to-anterior wave of differentiation . Sequential rounds of signaling produce ∼25 highly ordered rows of ∼30 ommatidia each to make up a total of 800 ommatidia per eye [13] . Each ommatidium contains eight PRs and accessory cells: the six “outer PRs” ( R1- R6 ) express a broad-spectrum Rhodopsin , Rh1 , and are required for motion and dim light vision . The two “inner PRs” ( R7 and R8 ) each express different Rhodopsins and are used for color discrimination and polarized light vision [8 , 14–16] . A detailed mathematical model for much of the process of eye formation has recently been formulated [17] . However , that model does not address the stochastic distribution of color photoreceptors , which is the subject of this paper . There are three main ommatidial subtypes in Drosophila , which are defined by the combination of Rhodopsin photopigments expressed in their R7 and R8 PRs [8] . Two types of so-called pale and yellow ommatidia , are randomly distributed across the retina in a 35:65 ratio [5 , 9] . The pale ommatidia express UV-sensitive Rh3 in R7 and blue-sensitive Rh5 in R8 and are used for the discrimination of short-wavelength light [5 , 8] . The yellow ommatidia express longer UV-sensitive Rh4 in R7 and green-sensitive Rh6 in R8 and are used for the discrimination of longer wavelengths [18] ( Fig 1a ) . A third subtype found in the dorsal rim area ( DRA ) is used for the detection of the vector of light polarization [19] . The stochastic distribution of yellow and pale ommatidia in Drosophila is controlled at a single upstream node in the retinal regulatory network by the stochastic expression of the transcription factor Spineless ( Ss ) in a subset of R7 cells [5] . In Doli , where the patterning is highly ordered , Ss might also be responsible for Rhodopsin expression as the eyes appear to develop in highly similar ways; it also seems likely that many of the interactions in the eye regulatory network are conserved between Doli and Drosophila [11 , 12] . The generation of the very different patterns observed might thus be due to changes in the initial expression of Ss . In this paper , we present a simple mathematical model that captures the essence of these ideas , by attributing the diverse patterning in the three fly species ( stochastic , ordered and semi-ordered ) to a single switching mechanism .
We assume that the choice of a color in each column is essentially complete by the time the morphogenetic furrow moves on to the next column . Thus , we regard fate establishment as instantaneous on the timescale at which the full ordering process occurs in the next column . Consistent with this assumption , Ss expression in Drosophila starts almost immediately after the morphogenetic furrow [21] . In Doli , the two alternate colors , green and red , are denoted by 1 and 0 respectively . In the hexagonal lattice formed by the ommatidia , the color choice of the ommatidium in the ith row and jth column is defined by the element aij ( which is either 1 or 0 ) in an n × m matrix . Two competing effects determine the value of a given element: The first is a default probability of being in the state 1 ( green ) for every element in a vertical column . The second is that of the subtype correlation between the ommatidium and its nearest neighbors in the preceding column . The default probability pj ( S|X ) of all sites in a column j to be green , can be expressed as: p j ( S | X ) = { P 0 X j ≤ X 0 1 - P 0 X j > X 0 ( 1 ) with P0 being a constant . This expresses the fact that , for all values smaller than a threshold X0 , the probability for the column to be green is given by P0: this changes to 1 − P0 when the threshold is exceeded . High probabilities of being green , as mentioned above , are related to the expression of Spineless . The expression of S leads in its turn to a rapid change in the value of X over column j + 1 , which can be simply modelled by the following linear equation: X j + 1 = γ - β p j ( S | X ) ( 2 ) where γ and β are constants . If β is negative , successive columns are anticorrelated , in order to model the case of Doli eyes ( Fig 1a ) , while the positive sign is appropriate for a hypothetical fly that would have a fully homogeneous retina . On the other hand , the constant γ , which parametrizes the sensitivity of the switch , is strictly positive . In order to build an extremely ordered alternating pattern such as that of Doli shown in Fig 1b , we can choose the constant P0 to be very small which guarantees almost uniform color in each column . We illustrate the case for P0 being very small in the following: In Fig 2a , P0 = 0 . 0001 , so that if , at the jth column , Xj is greater than the threshold , then the default probability pj of the column to be green is very close to 1 ( 1 − P0 = 0 . 9999 , from Eq 1 ) . With suitably chosen constants ( e . g . β = 8 , γ = 10 , X0 = 5 as chosen here ) , Eq 2 implies that the value of X at column j + 1 , Xj+1 ∼ 2 , which is less than the threshold X0 . In turn , the first line of Eq 1 yields pj+1 = 0 . 0001 , i . e . column j + 1 is red with very high probability . The ordered stripes of red and green are thus built across the eye as depicted in Fig 2a . For the ordered retina , the negative sign of β means that an ordered initial column with an above-threshold value of X , will , from Eq 2 , give rise to a constant value of X in all successive columns ( γ − β ( 1 − P0 ) ) , so that a column that is initially green will stay green forever , that is , all subsequent columns become the same color as the very first one . Some Doli species , such as this example from the Chrysosoma genus ( Fig 1d ) , have partially ordered eyes that are intermediate between Condylostylus and Drosophila retinas . These patterns can be modelled as perturbations of the uniform fly , as seen in Fig 2d . Fig 3 is an illustration of the quantitative behavior of Eqs 1 and 2 as a function of all the parameters β , γ , P0 and X0 , with γ and X0 scaled by β for convenience . Fixing the scaled value X0/β at a sample value of 0 . 2 , we obtain the regions of phase space dominated by Doli-type or uniform retinal pattern . The blue and purple regions correspond to uniform retinas , the green to the Doli retina , while the white region corresponds to physically inadmissible values where X0 ( as a biological factor ) takes negative values . The overlap of the blue , green and purple region for P0 > 0 . 5 indicates that the green region or the oscillating Xj will not occur anymore . The value of Xj will be fixed at either γ − β ( 1 − P0 ) or γ − βP0 in the overlapped region depending on its initial value . At the other end of the spectrum from Doli , the Drosophila retina contains ommatidia where R8 cells express either green- or blue-sensitive Rhodopsins that are randomly distributed , with a bias for green-sensitive ( yellow , 65% ) vs . blue-sensitive ( pale , 35% ) . This pattern is fully random [5 , 22–24] . This implies that the choice in every ommatidium is fully independent of all its neighbors , indicating that there is a total absence of correlations and full stochasticity: no auxiliary equation like Eq 2 is therefore needed . On the other hand , the bias can be accommodated by choosing the default probabilities of the two subtypes to be unequal . The analog of Eq 3 in this case becomes P i j D r o s o ( S , l ) = p j D r o s o ( S | X ) ( 4 ) where Pij is the probability that the ommatidium ij is green . Fig 2b shows a sample configuration simulated with P0 = 0 . 35 ( see Eq 1 ) , to be compared with an image of the Drosophila eye ( Fig 1a ) . Retinal patterns in other fly species may be conceived as a “mixture” of characteristics of the disorder in Drosophila and Doli or a fully ordered retina , respectively . The simplest way to achieve this from a mathematical perspective is to consider linear combinations of the transition probabilities of the form P j α ( S | X ) = α p j O r d e r e d F l y ( S | X ) + ( 1 - α ) p j D r o s o ( S | X ) , ( 5 ) and P j α ( S | X ) = α p j D o l i ( S | X ) + ( 1 - α ) p j D r o s o ( S | X ) , ( 6 ) respectively . The parameter α ∈ [0 , 1] quantifies the relative influence of the stochastic , Drosophila-like mechanism . Eq 5 applies for β < 0 , and Eq 6 for β > 0 . We use this formalism to analyze the eye of a partially ordered fly from the Chrysosoma genus ( Fig 1d ) where , as for Doli eyes , the pattern is that of corneal colors rather than Rhodopsin expressed in photoreceptors , which are not known in this species . Mathematically , this is a finite α case with β negative , corresponding to the first of Eq 5 . Our simulations ( Fig 2d ) agree qualitatively with the image of a real fly shown in Fig 1d; however , a detailed experimental study of its geometrical correlations needs to be performed to get more quantitative agreement . This formalism allows us to present a generic phase diagram ( Fig 4 ) , in which any given fly eye can be defined; the specific location , along with the theory above , enables us to identify the specific mechanisms associated with the morphogenesis , e . g . whether the fly is derived from a perturbation of the Uniform Fly , the hypothetical fly with completely ordered eye colors , or from Doli , and so on . This phase diagram is calculated based on the horizontal and vertical autocorrelations detailed below ( see Materials and Methods ) . Fig 4G corresponds to the case when one of two colors ( in this case green ) is predominant . The horizontal axis shows the value of α that varies from 0 to 1 symmetrically on each side of the figure; the right side corresponds to positive β , and the left side to negative β . The vertical axis gives the value of P 0 D r o s o . Panels A-I show sample configurations . Configuration A is pure Doli-like , where β is positive and α = 1 ( the two arrows indicate that the value of P 0 D r o s o is irrelevant , see Eq 6 ) . Configuration B represents a slight deviation from Doli , with β > 0 , α = 0 . 9 and P0 = 0 . 4 . In configuration C , the reduced value of α = 0 . 65 leads to more and more deviations from Doli; here , P0 = 0 . 9 . Configuration D corresponds to parameters α = 0 . 65 and P0 = 0 . 1; this is a near mirror image of C , because of the flipped value of P0 with respect to C . The β < 0 region in the negative x-axis is largely disordered . The probability of the so-called Uniform Fly is considered as p j O r d e r e d F l y ( S | X ) = 0 . 99 so that , e . g . the extreme configuration which occurs at α = 1 is 99 percent green . Configurations E and F are drawn based on identical values of α close to zero but for two opposite points on the vertical axis corresponding to P 0 D r o s o = 0 . 1 and P 0 D r o s o = 0 . 9 . Therefore , a Drosophila-dominant behavior with different percentages of greens and reds is observed in these panels . As α increases to the left , the effect of the Drosophila term in the first of Eq 6 decreases , so that configurations tend to become more and more a homogeneous green as will be seen in configurations G , I and H which correspond to α = 0 . 9 and P0 = 0 . 9 , 0 . 55 and 0 . 1 respectively . We next discuss the significance of our results and seek to relate the patterns we predict theoretically to simulated as well as real data . Since the stochasticity inherent in the model suggests that almost all parameter combinations generate almost all patterns with non-zero probability , the pertinent question to ask is: which parameter combination most likely generates a given pattern ? The aim of this exercise is twofold . First , we show using data simulated by our model , to what extent it is possible to infer the model parameters consistently from the patterns generated by the same model . This provides a baseline for the analysis of real-life images where we strive to infer the most likely parameters for a given image of the fly eye . These parameters then provide the mostly likely explanation for the mechanism that has generated the observed pattern via the phase diagram ( Fig 4 ) given above . A few technical considerations are in order: The parameters β and γ , cf . Eq ( 2 ) , themselves are not well-suited for inference because large fractions of the ( β , γ ) parameter space generate the same behaviour ( see Fig 4 ) . We replace β and γ by a discrete variable m with only 4 distinct values m ∈ {fr , fg , ar , ag} . These represent the combination of two binary variables ( i ) dynamic mode: fixed point / alternation ( ii ) initial condition: green probability high / low . Furthermore , we discretize the values of the parameter α to be multiples of 0 . 01 , i . e . α ∈ V ( α ) ≔ {0 . 0 , 0 . 0 . 1 , 0 . 02 , … , 1 . 00} . We restrict ϵ ∈ V ( ϵ ) ≔ {0 . 0 , 0 . 1 , 0 . 2 , 0 . 3 , 0 . 4 , 0 . 5} and P0 ∈ V ( P0 ) ≔ {0 . 0 , 0 . 1} . This discretization of the inference problem makes it computationally easier . It is justified by the fact that small variations in P0 affect probabilities in a way similar to small variations in α . In particular the parameters α and P0 are fully interdependent for a fixed point behavior ( m ∈ {fr , fg} ) . In this case , we set P0 = 0 . We keep the spatial range parameter at k = 1 . 0 throughout , thus not subjecting it to inference . We first apply parameter inference to simulated patterns generated by the model itself in order verify that parameters can indeed be inferred consistently . For each realization , we ( i ) draw parameter values uniformly at random from the set of eligible combinations defined in the preceding paragraph; ( ii ) generate a pattern by the model with these parameter values , which we call the true values; ( iii ) find an eligible combination of parameter values that maximizes the probability of the model generating the pattern from step ( ii ) ( maximum likelihood ) . Fig 5 shows the results for 10000 independent realizations of the inference using patterns of 50 columns and 30 rows . The mixing parameter α is inferred with good precision over the entire range . Values of the other parameters are also inferred with large precision for sufficiently large α . For small α , the color assignment is dominated by the fixed probabilities of Drosophila , so that the parameters m , ϵ and P0 have less influence on the pattern . Inference is expected to be more difficult for smaller values of α . As an application of the model to real data , we perform similar parameter inference on photoreceptor distribution data collected from the eyes of individuals in the “intermediate” Chrysosoma species . See Fig . S1 in S1 Text for details on images and processing of data . Parameters are discretized for the inference; α is in the set {0 . 01 , 0 . 02 , … , 0 . 99}; P0 ∈ {0 . 01 , 0 . 02 , 0 . 04 , 0 . 08 , 0 . 16}; ϵ ∈ {0 , 0 . 001 , 0 . 002 , … , 0 . 010}; and k ∈ {1 . 0 , 4 . 0 , 9 . 0 , 16 . 0 , 32 . 0 , … , 625 . 0} . The results of maximum likelihood parameter inference are as follows: All in all , the basic behaviour ( red fixed point ) is found consistently in all eyes . In the other parameters , we have strong fluctuations between samples; in some cases ( large ϵ and large α ) , the stochasticity is described by the large probability of error propagation; in the case of small α , the patterning resembles Drosophila . In future studies , stronger support for the model could be obtained by analyzing data from additional species . Ideally , the species could be distinguished by how their parameters fall into disjoint regions .
Statistical physics can be usefully applied to biology when searching for organizational principles [25–27] . The present study is such an example: we present a minimal model that describes patterns observed in the retinas of Drosophila and Dolichopodidae flies . Via a generic phase diagram , we are able to predict parameters that might underlie a range of patterns found on the eyes of various fly species . Conversely , model parameters can be estimated from images of retinal patterns . The essential ingredients of the model are stochasticity and correlations: an ommatidium chooses its color state depending on the competition between its gene-expression-induced probability to be a specific color ( e . g . the expression of Spineless ) and the influence of the ommatidium in the preceding column along the progression path of the morphogenetic furrow . Intercolumnar correlations greatly outweigh the effect of stochasticity in Doli , while the reverse is the case for Drosophila . This formalism also allows us to probe how mistakes propagate in the retina , for example in Doli . For instance , supposing an ommatidium in a column is a “mistake” , i . e . is of the wrong color , our formalism suggests that two scenarios are possible . If the mistake is isolated , it will soon be resolved as the morphogenetic furrow progresses . However , a cluster of mistakes can lead to interesting domain formation , with interfacial roughness as in , say , crystalline systems where dislocations define the area of interfacial disorder [28] . This is in accord with experimental observations [5 , 20] . The richness of our formalism , however , goes well beyond the description of known species to the prediction of those that are as yet undiscovered . In terms of simple model parameters , we are able to generate a global phase diagram for flies with binary color choices and allow for important clues to the morphogenetic mechanisms at play . The suggested parameter values for the Chrysosoma species are an example of this power , although we emphasize that in the absence of quantitative experimental data about the molecules involved and variables such as their diffusion constants , we have here aimed at only qualitative agreement . We hope this work will motivate detailed quantitative analysis of experimentally observed patterns , as well as genetic analysis of factors involved in the expression of color in fly eyes , so that our predictions can be put to the test . Once our model is adequately tested for two-color retinae , we will be able to extend our analysis to flies and other insects with more complex color patterning .
|
A simple model is able to account for a diversity of photoreceptor patterns in different fly species , ranging from highly deterministic to fully random .
|
[
"Abstract",
"Introduction",
"Model",
"Discussion"
] |
[
"invertebrates",
"medicine",
"and",
"health",
"sciences",
"ocular",
"anatomy",
"social",
"sciences",
"light",
"neuroscience",
"animals",
"electromagnetic",
"radiation",
"animal",
"models",
"drosophila",
"melanogaster",
"model",
"organisms",
"phase",
"diagrams",
"network",
"analysis",
"experimental",
"organism",
"systems",
"eyes",
"drosophila",
"research",
"and",
"analysis",
"methods",
"computer",
"and",
"information",
"sciences",
"animal",
"studies",
"animal",
"cells",
"regulatory",
"networks",
"sensory",
"receptors",
"head",
"insects",
"arthropoda",
"physics",
"signal",
"transduction",
"cellular",
"neuroscience",
"psychology",
"eukaryota",
"retina",
"anatomy",
"cell",
"biology",
"data",
"visualization",
"cornea",
"neurons",
"photoreceptors",
"biology",
"and",
"life",
"sciences",
"ocular",
"system",
"cellular",
"types",
"afferent",
"neurons",
"sensory",
"perception",
"physical",
"sciences",
"organisms"
] |
2018
|
Patterning the insect eye: From stochastic to deterministic mechanisms
|
In a forward genetic screen for regulators of pancreas development in zebrafish , we identified donuts908 , a mutant which exhibits failed outgrowth of the exocrine pancreas . The s908 mutation leads to a leucine to arginine substitution in the ectodomain of the hepatocyte growth factor ( HGF ) tyrosine kinase receptor , Met . This missense mutation impedes the proteolytic maturation of the receptor , its trafficking to the plasma membrane , and diminishes the phospho-activation of its kinase domain . Interestingly , during pancreatogenesis , met and its hgf ligands are expressed in pancreatic epithelia and mesenchyme , respectively . Although Met signaling elicits mitogenic and migratory responses in varied contexts , normal proliferation rates in donut mutant pancreata together with dysmorphic , mislocalized ductal cells suggest that met primarily functions motogenically in pancreatic tail formation . Treatment with PI3K and STAT3 inhibitors , but not with MAPK inhibitors , phenocopies the donut pancreatic defect , further indicating that Met signals through migratory pathways during pancreas development . Chimera analyses showed that Met-deficient cells were excluded from the duct , but not acinar , compartment in the pancreatic tail . Conversely , wild-type intrapancreatic duct and “tip cells” at the leading edge of the growing pancreas rescued the donut phenotype . Altogether , these results reveal a novel and essential role for HGF signaling in the intrapancreatic ducts during exocrine morphogenesis .
The vertebrate pancreas is an endodermal organ that is part endocrine , releasing hormones that regulate glucose metabolism , and part exocrine , releasing pancreatic juices that aid in digestion . Pancreatic endocrine and exocrine developmental dysmorphogenesis and dysregulation , including diabetes mellitus and pancreatic adenocarcinoma , can result in human diseases with high morbidity and mortality . Thus , a more sophisticated understanding of molecular mechanisms mediating pancreatic development and homeostasis will certainly refine the treatment of these diseases . In zebrafish as in mammals , all pancreatic endocrine and exocrine tissues derive from the fusion of a dorsal and ventral bud arising at the level of somites 2–9 [1] , [2] , [3] . In zebrafish , the dorsal bud generates the principal islet by 24 hours post fertilization ( hpf ) , and fuses with the emerging ventral bud between 40–44 hpf [4] , [5] . Around 52 hpf , acinar and ductal cells start to expand caudally to form the tail of the pancreas [5] , [6] , [7] . The pancreatic mesenchyme is essential for the induction , growth , branching , and cytodifferentiation of the pancreatic epithelium [8] . While several mesenchymal signals mediating pancreatic induction have been identified ( reviewed in [9] ) , our knowledge of how the mesenchymal/epithelial signaling pathways regulate pancreatic growth and branching is more limited [8] . Hepatocyte Growth Factor ( HGF ) is a stromally-produced ligand which binds Met , a receptor tyrosine kinase that is predominantly expressed in epithelia . Upon receptor dimerization and autophosphorylation , Met activates a bevy of cellular processes including motogenesis , tubulogenesis , mitosis , chemotaxis , and cell survival [10] . During organogenesis , HGF/Met signaling has been shown to be involved in liver and placenta formation , as well as in the migration of hypaxial muscle precursors into limbs [11] , [12] , [13] , [14] . However , the role of HGF/Met signaling in vertebrate pancreas development remains unclear . Both HGF and Met are expressed in the developing rodent pancreas [15] , [16] , but pancreatic phenotypes have not been characterized in global knockout mice . Studies have been mostly focused on the role of HGF/Met signaling in pancreatic tumorigenesis and beta-cell survival . Indeed , pancreas-specific Met knockout mice are euglycemic and morphologically unaffected at maturity , but show impaired beta-cell homeostasis during pregnancy [17] and following STZ-induced islet inflammation [18] . Even though HGF/Met signaling has been shown to activate the PI3K/Akt and ERK pathways in acinar cells [19] , its biological role during exocrine pancreas development remains undetermined .
To find novel regulators of endodermal organ morphogenesis and differentiation , we conducted a forward genetic screen utilizing doubly transgenic ( 2CLIP: 2-color liver , insulin , exocrine pancreas ) zebrafish with EGFP-expressing pancreatic acinar cells and DsRed-expressing pancreatic beta-cells and hepatocytes [20] . Using this approach , we identified donuts908 mutants , which show shortened ( type 1 ) or absent ( type 2 ) exocrine pancreatic ( xp ) tails . Tg ( ptf1a:GFP ) jh1 [21] was crossed into the donuts908 background to reveal the short exocrine pancreas phenotype at 3 days post fertilization ( dpf; Fig . 1A–E ) , prior to the terminal differentiation of acinar cells . Differentiation of the principal islet ( beta-cells , β ) , acinar and ductal cells , and liver ( li ) appeared unaffected in donut mutants ( Fig . 1B , C; and see below ) . Nearly 25% of progeny from heterozygous intercrosses showed the donut phenotype , evenly split between types 1 and 2 ( Fig . 1F ) . The variable expressivity suggested either a hypomorphic mutation in donut , and/or a genetic interaction with other loci . To isolate the molecular lesion responsible for the donut phenotype , we used bulk segregant analysis and localized donut on linkage group 25; we then used z-markers and customized CA-repeat markers to define a critical interval containing 14 candidate genes ( Fig . 1G ) . In sequencing these candidates , we identified a T2324G variant in the coding sequence of the met gene , resulting in an L775R missense substitution ( Fig . 1H ) . The MetL775R substitution in donuts908 mutants represents a significant shift in amino acid charge and polarity at a residue that is conserved in vertebrates as a leucine or an isoleucine ( Fig . 1I ) . Importantly , L775 is localized to the IPT3 ( immunoglobulin-like regions in plexins and transcription factors 3 ) domain in the Met ectodomain , which together with IPT4 comprises a high affinity HGF binding site ( Fig . 1L ) [22] , [23] . Moreover , protein domain modeling based on human MET IPT3 indicates that the analogous residue , I777 , resides in a solvent inaccessible region of IPT3 , and that its substitution with arginine may be sterically unfavorable ( Fig . 1J , K ) , suggesting that Met folding may be compromised in zebrafish donut mutants . We hypothesized that the donuts908 mutation would result in a Met loss of function . To test this hypothesis , we injected zygotes with translation-blocking met morpholinos ( metMO ) . We found that even low doses of metMO ( 2 ng ) could reproduce the shortened pancreatic tail phenotype observed in donut mutants with high penetrance ( Fig . 1M , N ) . Importantly , when donuts908+/− embryos were injected with a near-threshold ( 1 ng ) dose of MO , 53 . 4% displayed a short pancreas phenotype , as compared to 29 . 1% of injected wild-type embryos ( Fig . S1 ) . Since HGF is the only demonstrated ligand for Met , we reasoned that knockdown of both zebrafish hgf paralogs , hgfa and hgfb , should also mimic donut mutant phenotypes . Similar to donut mutants , we found that embryos injected with a mixture of MOs directed against hgfa and hgfb ( hgfMO ) also developed shortened pancreata ( Fig . 1O , P ) . We observed a positive correlation between the injected quantity of each morpholino and the penetrance and expressivity of the pancreatic tail phenotypes ( Fig . 1Q ) , with expressivity variation resolving towards the more severe type 2 phenotype at higher doses . Finally , donut mutants frequently lack muscle tone in the pectoral fins , suggesting a failed morphogenesis of hypaxial muscle ( Fig . 1D , E , arrowheads; Fig . S2 ) , which corroborates previous studies of Met function in myogenic precursor cells in mouse and zebrafish [11] , [12] . In sum , these data are consistent with the interpretation that the phenotypic spectrum observed in donuts908 mutants is due to a hypomorphic effect of the L775R substitution . To clarify how HGF signaling is implemented in pancreatogenesis , we examined the expression of met , hgfa and hgfb during several stages of pancreas formation . We also analyzed the expression of both furin genes , furina and furinb , which encode the proteases that cleave Met into a mature form . At 34 hpf , which is prior to pancreatic bud fusion , met is expressed throughout the endodermal organ forming region , including pancreas , liver , and intestine anlagen ( Fig . 2A ) . Additionally , met expression is detected in pancreatic endoderm before ( 44 hpf ) , during ( 52 hpf ) , and after its caudal extension ( 80 hpf; Fig . 2A ) . We found that hgfa and hgfb were expressed adjacent to the dorsal pancreatic bud at 34 hpf , and that their expression remains in register with the leading edge of the growing pancreatic tail ( red arrowhead ) between 52 and 80 hpf ( Fig . 2B , C ) . Importantly , both furina and furinb are expressed in endodermal tissue throughout pancreatogenesis , with furina present in liver and pancreas at 75 hpf ( Fig . S3 ) , indicating that the process of Met maturation is active in these organs . To determine which exocrine tissues express met during pancreatic tail morphogenesis , we performed fluorescent in situ hybridization in the Tg ( nkx2 . 2a ( -3 . 5 kb ) :GFP ) line ( hereafter duct:GFP ) which highlights the intrapancreatic ducts ( IPDs ) [6] . We found that within the pancreatic tail , met was expressed in both IPD and acinar cells at 60 hpf , a time of active pancreatic tail extension ( Fig . 2D–F ) . A role for Met signaling in pancreatic exocrine morphogenesis has not been reported in mammalian model systems . Hgf and Met knockout mice die during organogenesis precluding analysis [11] , [13] , [14] , and a pancreas specific knockout ( PancMet KO ) showed no clear morphological defect of the adult pancreas [18] . However , as pancreas development was not described in PancMet KO , it is possible that a transient phenotype or developmental delay could have been overlooked . Furthermore , it is likely that early mosaic recombinase activity inherent to the Pdx1:Cre line that was utilized in these experiments leaves a population of wild-type cells in these mice . These cells may be competent to effect normal morphogenesis , and would mirror what we observed in chimera studies ( see below ) . Alternately , residual Met protein , which is expressed in endoderm prior to the onset of Pdx1 expression , or a parallel/redundant mechanism could drive exocrine outgrowth in mammals . Additional targeted studies are needed in mice to determine whether the role of Met in exocrine pancreas growth is conserved . To model how the donut mutation affects Met function , we generated an I776R point mutation in murine Met , an analogous lesion to the donuts908 mutation , and then transfected either MetWT or MetI776R into TOV112D cells . This cell line lacks endogenous Met activity , but expresses the intracellular components of the Met signaling pathway [22] , [24] . Met is translated as a single polypeptide chain which is cleaved by furin into alpha and beta chains [25] . In TOV112D cells transfected with MetWT , most of the nascent polypeptide was cleaved to the mature form ( 150 kDa+30 kDa ) , while a small proportion of the receptor remained unprocessed ( 190 kDa ) . Since the I776R lesion lies far from the furin cleavage site ( residues 302–307 ) [26] , we were surprised to find that the MetI776R precursor polypeptide failed to be cleaved ( Fig . 3A , bottom panels ) . Since cleavage-incompetent Met mutants ( i . e . MetR306A ) can signal normally [27] , we next assessed the signaling efficacy of MetWT and MetI776R using specific anti-phosphorylated Met antibodies directed against several key tyrosine residues . Upon HGF binding , Met dimerizes and its cytoplasmic catalytic region is activated by trans-phosphorylation of tyrosines Y1234 and Y1235 . Subsequently , residues Y1349 and Y1356 are phosphorylated in the Met multifunctional docking site , which connects Met to multiple downstream branches , including Ras/MAPK and PI3K , through several adapter proteins . In addition , phosphorylation of tyrosine Y1003 negatively regulates Met signaling by promoting receptor turnover . Compared to MetWT , phosphorylation of Y1234-5 , Y1349 , and Y1003 were significantly diminished in MetI776R ( Fig . 3A , B ) . Finally , since cleavage by Furin occurs in the Golgi apparatus [28] , we hypothesized that intracellular trafficking of MetI776R from the endoplasmic reticulum to the plasma membrane was impaired . To investigate this hypothesis , we next examined the presentation of Met at the plasma membrane using an antibody that binds specifically to the Met ectodomain ( α-Met ab1 ) . Antibody staining was evident at the membrane of unfixed , unpermeabilized HEK293T cells expressing MetWT or cleavage incompetent MetR306A , but not MetI776R ( Fig . 3C–E ) . However , all of these Met variants were detected in fixed and detergent-permeabilized cells using an antibody that recognizes an intracellular epitope ( α-Met ab2 ) , thereby indicating similar transfection efficiencies of the three constructs ( Fig . 3F–H ) . Thus , these data suggest that MetI776R is retained in an intracellular compartment , and that the lack of MetI776R cleavage per se is not causing this retention . To further test this hypothesis , RNA encoding zebrafish MetWT or MetL775R fused to mCherry were injected in zebrafish embryos and the localization of the proteins analyzed at blastula stages ( Fig . 3I–J ) . As observed in cell culture , MetL775R was not detected at the plasma membrane in vivo . This defect is not simply a delay of membrane targeting , as similar results were observed at later time points ( data not shown ) . Together , these data support the hypothesis that deficient signaling through MetL775R causes the hypomorphic effect observed in donuts908 mutants . Likely , reduced Met signaling is due to ( 1 ) localization of MetL775R away from the plasma membrane , ( 2 ) impaired binding of HGF to the high-affinity binding site in IPT3-4 , and/or ( 3 ) impaired assembly of Met co-receptor moieties , that would then lead to defective phospho-activation of Met . Growth and elongation of the pancreatic tail involves both proliferation and directed migration of exocrine cells [6] , [7] . To analyze which mechanism is impaired in donut mutants , we first assayed cell proliferation using 30 min incorporation of the thymidine analog EdU at two time points during pancreatic tail growth: 56 and 75 hpf ( Fig . 4A–D ) . Quantification of labeled cells showed no significant difference in exocrine cell proliferation between WT and donut mutant animals ( Fig . 4E ) . To determine whether acinar or ductal cell migration was affected in donut mutants , we next examined the distribution of acinar and ductal cells using ptf1a:GFP and duct:GFP , respectively , in 84 hpf wild-type ( WT ) and metMO-injected ( Fig . 4F–I ) larvae . Although ptf1a:GFP+ acinar cells were always confined to the pancreas , duct:GFP+ cells were clearly localized in the hepatic region in both type 1 and type 2 mutants . Furthermore , while IPD cells were elongated and spindle-shaped in WT , they showed a more rounded morphology in metMO-injected larvae ( Fig . 4H , I , insets ) . We marked hepatocytes and biliary ducts with Prox1 and Alcam antibodies , respectively , and found that duct:GFP+ pancreatic cells had extensively invaded the liver , tracking along the biliary duct network ( Fig . 4J , K ) . Taken together , our data show that HGF/Met signaling promotes invasive cell behavior during pancreatic tail morphogenesis , rather than the proliferation of exocrine cells . The invasive cell behaviors activated by HGF/Met signaling , such as motility and proliferation , work through multiple parallel downstream branches of Met signaling . Three key branches involve the PI3K , ERK , and STAT3 pathways , which may be activated by direct interaction with Met , or through adapter proteins: the PI3K/Akt pathway elicits cell migration via Rac1 , as well as cell survival; the ERK branch promotes proliferation and transformation; and STAT3 signaling mediates branching morphogenesis and proliferation ( reviewed in [10] ) . To dissect which specific branches might be critical for Met-mediated exocrine morphogenesis , we treated larvae during the formation of the pancreatic tail with specific inhibitors ( Fig . 4L ) . Inhibition of ERK signaling with the MEK inhibitor UO126 or the p38 MAPK inhibitor SB203580 resulted in a mild phenocopy of donut type 1 and type 2 mutants , ranging from 25–40% . However , treatment with the potent PI3K inhibitor LY294002 or the STAT3/Src family kinase inhibitor SU6656 generated a nearly perfect phenocopy of type 1 and 2 donut mutants . These data confirm our previous observations that cell proliferation via MAPK/MEK is not the primary mechanism driving pancreatic tail outgrowth . The furin protease inhibitor CMK caused only a minor effect on pancreatic tail formation confirming that failure of proteolytic cleavage of Met may not cause the reduced signal transduction capacity of MetL775R . The differentiating exocrine pancreas is highly organized at the onset of pancreatic tail elongation [6] , [7] . Acinar and ductal cells are the only two cell types found in the growing pancreatic tail , and they are always closely associated . Even though met is expressed in both cell types , we hypothesized that it was required primarily in ductal cells , since these cells were mislocalized and dysmorphic in metMO-injected larvae . To test this hypothesis , we performed cell transplantation experiments to generate chimeras and determine the autonomy of Met function in the pancreas ( Fig . 5A ) . First , blastomeres isolated from control or metMO-injected Tg ( hs:mCherry ) ( see Methods ) embryos were directed toward an endodermal fate by cas mRNA coinjection [29] , [30] , and were transplanted into WT hosts . Incorporated donor cells ubiquitously express mCherry RFP upon heat shock induction , and were quantified based on their contribution to ptf1a:GFP+ acinar or ptf1a:GFP− ductal cell compartments in the distal or proximal portion of the exocrine pancreas ( Fig . 5B ) . In contrast to WT cells that contributed evenly to both acinar and ductal compartments , metMO-injected cells were strongly biased toward the acinar cell and proximal ductal compartments , and were largely excluded from the distal ductal compartment ( Fig . 5C–E; Tables 1 , 2 ) . We observed similar results when we transplanted endodermal cells from donut mutant embryos into Tg ( hs:mCherry ) hosts ( data not shown ) . Next , we assessed the rescue of exocrine pancreatic tail outgrowth by transplanting varying numbers of WT Tg ( hs:mCherry ) cells into the endoderm of metMO-injected hosts ( Tables 3 , 4 ) . metMO-injected larvae lacking WT contribution showed a type 1 donut phenotype ( Fig . 5F ) , whereas a high contribution of WT endoderm cells led to a complete rescue of the pancreatic tail outgrowth ( Fig . 5G ) . Notably , the tip of the rescued pancreatic tail appeared to be composed exclusively of WT donor cells ( Fig . 5G , inset ) . In larvae with fewer integrated WT cells , the pancreatic tail was also rescued , with the WT cells being preferentially found in the IPD ( Fig . 5H , inset ) . Finally , we confirmed that WT endoderm could not similarly rescue the short exocrine pancreas phenotype observed in hgfMO-injected embryos , consistent with a mesenchyme-specific role for hgfa and hgfb ( Fig . S4 ) . These data indicate that endodermal Met signaling , probably preferentially in distal IPD cells , directs the caudal extension of the pancreatic tail . We thus propose a model for pancreatic tail outgrowth in which ductal cells would respond to HGF secreted from the mesenchyme adjacent to the pancreatic primordium and initiate migration caudally . The HGF signal may constitute a chemotactic signal for the ductal cells , or simply a motogenic signal , with directionality imparted by distinct signaling pathways . Even though Met signaling appears to be required in IPD cells for pancreatic outgrowth , Met signaling may also be active in a small population of “tip” cells that are mature or differentiating acinar cells . From the analysis of a specific subset of chimeric embryos , ductal cells may thus be directing the migration of the pancreatic tail from a trailing position , possibly exhibiting cellular extensions , such as cytonemes or filipodia , that were not resolved in our studies . Hence , our data expand the understanding of the role of HGF/Met signaling in mitogenic , motogenic and morphogenetic events required for the development , homeostasis and regeneration of different tissues . In addition , they provide a developmental framework to dissect the role of Met in pancreatic cancer stem cells [31] , [32] .
Fish were raised and maintained under standard conditions [33] . Pigmentation was inhibited with 0 . 02 mM phenlythiourea ( Sigma ) . We used the following published strains: Tg ( fabp10:RFP , ela3l:EGFP ) gz12 [34] , Tg ( ins:CFP-NTR ) s892 [35] , Tg ( ins:dsRed ) m1018 [36] , Tg ( nkx2 . 2a ( -3 . 5 kb ) :GFP ) ia3 ( a . k . a . “duct:GFP” ) [6] , and Tg ( ptf1a:eGFP ) jh1 [21] . To generate the Tg ( hsp70l:loxP-mCherry-stop-loxP-NICD-P2A-Emerald ) line ( a . k . a . Tg ( hs:mCherry ) ) , H2B-GFP was replaced in the plasmid hsp70l:loxP-mCherry-loxP-H2B-GFP-cryaa:Cerulean [37] by an in-frame PCR fusion product of NICD [38] , P2A [39] , and Emerald GFP ( Invitrogen , Carlsbad , CA ) . Transgenesis was achieved as described [40] . An ENU mutagenesis screen was executed as described [20] . Bulk segregant analysis was performed using pooled DNA extracted from 20 WT or donut mutant embryos . donut mutants were subsequently genotyped using a dCAPS strategy: a PCR product amplified by the primer set: 5′-TCCAG CCCAA ACATT CTTTC and 5′- CGTTT GTGTG GGTTG TATAG ACTCA CCACT TGGAA GAGTT TGCCC TCAGT GGCGG CAGcG was digested with HhaI . Pymol software ( Schrödinger ) was used to manipulate the PDB model of human IPT1-4 [41] . Whole mount antibody staining , in situ hybridization , and proliferation analysis with EdU were performed as described [20] . Proliferation index was calculated using the number of EdU+ cells divided by the total number of exocrine pancreas nuclei , per section; >5 slices were counted per animal . Significance was assessed using student's t test . Fluorescent in situ hybridizations were performed as described [42] . Templates for antisense RNA probes were amplified from embryonic cDNA with the following primers: met: CGGAG AGAGA GGGAG GAAG and TAATA CGACT CACTA TAGGG AGACA TTGAT GTCCG TGATG GAG; hgfa: TGTGT GCTTG AGAAA GAGAG AGA and TAATA CGACT CACTA TAGGG AGATC GACAA ATTGC CACGA TAA; hgfb: AGCCA CTGCA GGGAG ACTAC and TAATA CGACT CACTA TAGGG AGAGG GGTAC CTTTT AGGGT GGA; furina: GTGTC GGAGT GGCCT ACAAT and TAATA CGACT CACTA TAGGG AGGGT CTTCA TCCCA GGAGT; furinb: TGACC TGGAG AGACA TGCAG and TAATA CGACT CACTA TAGGG AATGC TGGGG GATTT TCTCT . For cytoimmunofluorescence , human embryonic kidney cells ( HEK293T ) were maintained in DMEM supplemented with 10% FBS . HEK293T cells were plated on polylysine-coated coverslips and transfected with lipofectamine with either murine pBabe-puro-MetWT ( Addgene ) or site directed mutagenesis-generated pBabe-puro-MetI776R . Cells were transfected 48 h before immunostaining: unfixed cells , as well as 4% paraformaldehyde-fixed and 0 . 1% Triton-permeabilized cells were incubated with indicated antibodies for 2 hours prior to incubation with appropriate fluorescent secondary antibodies . TOV112D human ovarian carcinoma cells were obtained from ATCC ( Manassas , VA ) and cultured using a 1∶1 mixture of MCDB 105 medium and medium 199 supplemented with 15% FBS . MetWT and MetI776R were transfected into TOV112D cells with lipofectamine . Coimmunoprecipitations and protein blot analyses were performed as previously described [43] . The antibodies used for immunoprecipitation , protein blot and immunofluorescence are: anti-Met ( 25H2 ) and anti-phosphoMet ( Cell Signaling ) , goat anti-Met ( extracellular epitope , R&D Systems ) , rabbit anti-Met ( SP260 – intracellular epitope , Santa Cruz ) . Tg ( nkx2 . 2a ( -3 . 5 kb ) :GFP ) ia3; Tg ( ptf1a:dsRed ) jh1 embryos , distributed in 12-well plates ( 20 embryos per well ) , were incubated with different chemicals from 60 to 76 hpf , during the growth of the pancreatic tail . Chemicals were specific inhibitors of MEK ( UO126 , at 100 µM ) , p38MAPK ( SB203580 , at 100 µM ) , PI3K ( LY294002 , at 30 µM ) , Src kinase ( SU6656 , at 50 µM ) and furin ( CMK , at 65 µM ) . The morpholinos ( Gene Tools ) used for gene knockdown were described in [12] , [44] . For experiments at the blastula stage , mRNAs encoding zebrafish wild-type or L775R Met fused to mCherry were synthesized in vitro , by kit ( Ambion ) , and 150 pg of mRNA was injected . Live imaging of embryos was performed 5 hours post-injection . For transplantation , capped cas/sox32 [29] mRNA was synthesized in vitro , and donor Tg ( hsp70l:loxP-mCherry-stop-loxP-NICD-P2A-Emerald ) embryos , which exhibit strong , ubiquitous RFP expression in the absence of Cre recombinase , were injected with 200 pg of cas mRNA alone or with 4 ng of met morpholino . Cell transplantations were performed as described [42] . When cas-injected putative met mutant cells were transplanted , donors were genotyped after transplantation .
|
The pancreas functions as an endocrine and exocrine gland that secretes hormones regulating blood glucose homeostasis , and pancreatic juice that aids the digestion and absorption of nutrients , respectively . Contrary to endocrine tissue development , that of the exocrine pancreas has received less attention . We conducted a forward genetic screen in zebrafish and identified HGF/Met signaling as a key regulator of exocrine development . We called the mutant donut because the body of the pancreas fails to elongate and thus remains rounded . The mutation leading to this phenotype affects the extracellular domain of Met , the hepatocyte growth factor ( HGF ) receptor , impairing its maturation , plasma membrane localization and phospho-activation . Although HGF/Met signaling may elicit many context-dependant cellular responses , our data indicate that HGF/Met signaling triggers the migration , but not the proliferation , of the pancreatic ductal cells to drive the extension of the pancreatic tail .
|
[
"Abstract",
"Introduction",
"Results/Discussion",
"Materials",
"and",
"Methods"
] |
[
"developmental",
"biology",
"cell",
"migration",
"genetics",
"biology",
"morphogenesis",
"gene",
"function"
] |
2013
|
Hepatocyte Growth Factor Signaling in Intrapancreatic Ductal Cells Drives Pancreatic Morphogenesis
|
The four dengue virus serotypes ( DENV1–4 ) cause the most prevalent mosquito-borne viral disease affecting humans worldwide . In 2009 , Nicaragua experienced the largest dengue epidemic in over a decade , marked by unusual clinical presentation , as observed in two prospective studies of pediatric dengue in Managua . From August 2009–January 2010 , 212 dengue cases were confirmed among 396 study participants at the National Pediatric Reference Hospital . In our parallel community-based cohort study , 170 dengue cases were recorded in 2009–10 , compared to 13–65 cases in 2004–9 . In both studies , significantly more patients experienced “compensated shock” ( poor capillary refill plus cold extremities , tachycardia , tachypnea , and/or weak pulse ) in 2009–10 than in previous years ( 42 . 5% [90/212] vs . 24 . 7% [82/332] in the hospital study ( p<0 . 001 ) and 17% [29/170] vs . 2 . 2% [4/181] in the cohort study ( p<0 . 001 ) . Signs of poor peripheral perfusion presented significantly earlier ( 1–2 days ) in 2009–10 than in previous years according to Kaplan-Meier survival analysis . In the hospital study , 19 . 8% of subjects were transferred to intensive care , compared to 7 . 1% in previous years – similar to the cohort study . DENV-3 predominated in 2008–9 , 2009–10 , and 2010–11 , and full-length sequencing revealed no major genetic changes from 2008–9 to 2010–11 . In 2008–9 and 2010–11 , typical dengue was observed; only in 2009–10 was unusual presentation noted . Multivariate analysis revealed only “2009–10” as a significant risk factor for Dengue Fever with Compensated Shock . Interestingly , circulation of pandemic influenza A-H1N1 2009 in Managua was shifted such that it overlapped with the dengue epidemic . We hypothesize that prior influenza A H1N1 2009 infection may have modulated subsequent DENV infection , and initial results of an ongoing study suggest increased risk of shock among children with anti-H1N1-2009 antibodies . This study demonstrates that parameters other than serotype , viral genomic sequence , immune status , and sequence of serotypes can play a role in modulating dengue disease outcome .
Dengue is an increasing public health problem in tropical and sub-tropical regions , with tens of millions of cases estimated to occur annually [1] . The four serotypes of dengue virus ( DENV-1–4 ) , a mosquito-borne Flavivirus , cause a range of clinical manifestations , from undifferentiated illness and classic Dengue Fever , to more severe syndromes characterized by plasma leakage , shock , and death , referred to as Dengue Hemorrhagic Fever and Dengue Shock Syndrome ( DHF/DSS ) [2] . While serotype and dengue immune status have been shown to affect severity of disease [3] , [4] , [5] , [6] , many of the epidemiologic and clinical variations in the presentation of dengue remain poorly understood . Dengue transmission has increased dramatically in Nicaragua and much of the rest of the Americas in the past three decades . All four serotypes now circulate in Nicaragua , though unlike Asia , where dengue is hyperendemic [3] , [7] , one serotype usually predominates each year [8] , [9] , [10] , [11] , [12] . The dengue season starts some months after the first rains , and typically lasts from August to January . In 2004–2008 , several thousand cases of laboratory-confirmed dengue were reported annually by the National Epidemiologic Surveillance program , though actual numbers of cases are suspected to be much higher [13] . Approximately 2–10% of laboratory-confirmed dengue cases were the more severe dengue hemorrhagic fever ( DHF ) or dengue shock syndrome ( DSS ) , and less than 1% resulted in death ( A . Nuñez and A . Balmaseda , unpublished data ) . In 2009–10 , Nicaragua experienced one of the largest dengue epidemics since the virus was reintroduced into the country in the 1980s , with approximately 3 times as many laboratory-confirmed cases reported by national authorities and documented in our studies than in the previous five years . The same serotype , DENV-3 , was responsible for the majority of cases in the 2008–9 , 2009–10 , and 2010–11 seasons , but caused an atypical clinical presentation only in 2009–10 , characterized primarily by early signs of poor peripheral perfusion and what is designated “compensated shock” . Here we present epidemiologic and clinical data from two on-going prospective studies of pediatric dengue in Managua , Nicaragua , that define the characteristics of the epidemic and begin to investigate possible explanations .
The Pediatric Dengue Cohort Study is a community-based prospective study that was initiated in 2004 in the low- to middle-income District II of Managua , Nicaragua , close to Lake Nicaragua , in which most residents attend the local municipal health center , the Health Center Socrates Flores Vivas ( HCSFV ) . Children aged two to nine years old living in the catchment area of the HCSFV were initially enrolled in August–September 2004 , and new 2-year olds have been enrolled each year since then . Children are withdrawn from the study when they reach 15 years of age . Study design and methods have been previously described [14] . Participants are encouraged to present at the first sign of illness to the HCSFV , where study physicians screen them for signs and symptoms of dengue and presence of warning signs for severity using a standardized data collection form . Subjects are followed daily during the acute phase of illness by physicians at the HCSFV or via home visits by study nurses . Acute and convalescent ( 14 days after onset of fever ) blood samples are drawn for virological , serological , and molecular biological testing for dengue , and additional blood samples are drawn every other day during the acute phase of illness for Complete Blood Count ( CBC ) and blood chemistry tests as warranted . Participants are transferred to the National Pediatric Reference Hospital in Managua ( Hospital Infantil Manuel de Jesús Rivera , HIMJR ) if warning signs or risk factors are present ( see below ) . Trained study physicians and nurses also collect data using standardized forms at the HIMJR . Additionally , each year in July–August , a healthy blood sample is drawn from all subjects . Sera from consecutive annual samples are tested for the presence of anti-DENV antibodies to identify silent transmission during the year and to determine dengue immune status . Beginning in 1998 , a prospective study of pediatric dengue has been carried out in the Infectious Disease Ward of the HIMJR to study clinical , immunological and viral risk factors for severe dengue [9] , [10] , [15] , [16] , [17] , [18] . This report focuses on data from August 2005 through January 2010 . Children between six months and 14 years of age with suspected dengue ( <7 days of illness ) and who are not actively enrolled in the concurrent cohort study are eligible to participate in the hospital study . Both in-patient and out-patient subjects are enrolled each year during the dengue season ( August–January ) and followed clinically through the acute phase of illness . Upon enrollment , a medical history is taken and a complete physical exam is performed . Clinical data , including vital signs , symptoms , fluid balance and treatment , are recorded daily on standardized data collection forms during hospitalization or through daily ambulatory follow-up visits by the same team of study physicians and nurses responsible for care of hospitalized cohort study participants . Acute blood samples are taken daily for CBC and serological , virological , and molecular biological testing for DENV infection , and ultrasound and/or X-rays are performed daily . Participants requiring more intensive therapies are transferred to the intensive care unit ( ICU ) . A convalescent-phase blood sample ( two weeks after presentation to hospital ) is also collected . The protocols for both studies were reviewed and approved by the Institutional Review Boards ( IRB ) of the University of California , Berkeley , and of the Nicaraguan Ministry of Health; additionally , the cohort study protocol was reviewed and approved by the IRB of the International Vaccine Institute in Seoul , South Korea . Parents or legal guardians of all subjects in both studies provided written informed consent , and subjects 6 years of age and older provided assent . Acute-phase serum samples are tested for DENV RNA using a nested reverse transcriptase–polymerase chain reaction ( RT-PCR ) directed to the capsid gene , which also permits identification of serotype [19] . Samples positive by RT-PCR are processed for viral isolation by inoculation onto Ae . albopictus C6/36 cells [20] . Paired acute- and convalescent-phase samples are tested for anti-DENV IgM antibodies using an in-house IgM capture ELISA [21] and for total anti-DENV antibodies by Inhibition ELISA [11] , [22] . The presence of anti-influenza A H1N1 2009 antibodies in paired acute- and convalescent-phase samples from 2009–10 DENV-positive hospital subjects was determined using the hemagglutination inhibition assay [23] . The antigen used was prepared using Nicaraguan strain A/Managua/2339 . 03 09-H1N1-SW and is specific for the H1N9 2009 strain of Influenza A ( S . Saborio and A . Balmaseda , unpublished data ) . DENV isolates were sequenced at the Broad institute using a combination of Sanger sequencing and high-throughput ( 454/Roche ) pyrosequencing . Sequences were aligned using Muscle [24] , with parameters optimized for maximum accuracy ( i . e . , no limit was specified for runtime or iteration count ) . Aligned sequences were clustered using phyML [25] with the following parameters: ( a ) substitution model = HKY85 , ( b ) number of substitution categories = four , ( c ) proportion of invariant sites = zero , and ( d ) estimated values for equilibrium frequency , transition/transversion ratio and gamma shape parameter . In these studies , a suspected dengue case was considered positive if it met one of the following criteria: DENV was isolated; DENV RNA was detected through RT-PCR; seroconversion of DENV-specific IgM was detected in paired acute- and convalescent-phase samples; or antibody titer by Inhibition ELISA demonstrated a 4-fold or greater increase in titer between acute and convalescent sera [9] . Primary DENV infections were those in which acute antibody titer was <10 or convalescent antibody titer was <2 , 560 and secondary infections were those in which antibody titer was ≥10 ( acute ) or ≥2 , 560 ( convalescent ) as determined by Inhibition ELISA [12] . Hospital cases from 2009–10 were considered positive for anti-influenza A H1N1 2009 antibodies if the HI titer was ≥20 in acute samples or if seroconversion or a ≥4-fold increase in HI titer was observed in paired acute and convalescent samples . In 2004–9 , cohort subjects presenting to the HCSFV were referred to the HIMJR , and hospital study participants were hospitalized , if they presented any of the following warning signs: persistent vomiting; moderate-to-severe dehydration; signs or symptoms of shock; abdominal pain; breathing difficulties; moderate-to-severe hemorrhagic manifestations; neurological manifestations; thrombocytopenia ( platelet count ≤100 , 000 platelets/µL ) ; or hematocrit ≥20% of normal value for age and sex . Children with known risk factors were also referred to the hospital , including those who were overweight and/or under one year of age . During the 2009–10 dengue epidemic , national health authorities mandated primary care physicians to refer all suspected dengue cases with the above signs of alarm to the hospital , regardless of clinical laboratory results , and hospitals to admit all referrals from health centers for observation . All laboratory-confirmed cases without signs of severity were classified as dengue fever ( DF ) , including cases presenting as undifferentiated fever . Severity was categorized according to the 1997 WHO guidelines [2]: Dengue Hemorrhagic Fever ( DHF; hemorrhagic manifestations , platelet count ≤100 , 000 platelets/µL , and evidence of plasma leakage ) and Dengue Shock Syndrome ( DSS; DHF with circulatory disturbance evidenced by hypotension for age or narrow pulse pressure accompanied by clinical signs of shock ) . In addition , included here among categories of severe dengue are “Dengue with Signs Associated with Shock” ( DSAS; DF with hypotension or narrow pulse pressure plus one or more of the following: capillary refill >2 seconds , cold extremities , or weak pulse ) [9] and “Dengue Fever with Compensated Shock ( DFCS; DF with capillary refill >2 sec plus cold extremities , poor or impalpable pulse , tachycardia , and/or tachypnea ( increased breathing rate for age ) on the same day ) . While clinical classifications by study physicians were assigned at discharge , here we present classifications determined by applying algorithms for DHF , DSS , DSAS and DFCS to clinical data ex post facto . Separate from DFCS as a disease classification , patients were considered to have experienced “compensated shock” if they presented with capillary refill >2 sec plus cold extremities , poor or impalpable pulse , tachycardia , and/or tachypnea on the same day; this state could or could not evolve into hypotensive shock depending on treatment and patient response . Thus , cases classified as DHF , DSS or DSAS may have experienced compensated shock during the course of illness . Delayed capillary refill , a sign of vasoconstriction and low blood volume , was the key clinical criteria for compensated shock , along with signs of circulatory distress , including cold extremities . In this study , study physicians and nurses used standard techniques of observation . Capillary refill time was determined by the clinician by briefly pressing on the pad of the patient's index finger and counting the number of seconds before normal color returned . Cold extremities is defined as lower than normal skin temperature and was observed by the clinician by touching with the dorsal side of the hand the patient's extremities , most commonly the plantar side of the foot . Cold and clammy skin , which is included in the variable cold extremities here , was also observed by palpating the plantar region of the inferior extremities with the dorsal side of the hand to sense coldness , clamminess , and sweat . Demographic and clinical characteristics of dengue cases in both studies are presented by dengue season , defined in the year-round cohort study as July 1–June 30 of each year in years 2005–2010 and August 1 , 2004–June 30 , 2005 in the first year of the study . In the hospital-based study , the dengue season is defined as August 1–January 31 , months corresponding to the highest incidence of symptomatic DENV infection , when study enrollment occurs ( clinical follow-up continues through the month of February each year ) . Total and age-stratified incidence in the cohort study were calculated using laboratory-confirmed dengue cases and the number of participants active at the time of the annual healthy blood sample collected in July of each year . Univariate and bivariate analyses were performed comparing cases during the 2009–10 dengue season to cases in previous dengue seasons . For bivariate comparison of cases across seasons , chi squared and Fisher's exact tests were applied to categorical variables , while the t-test and non-parametric Mann-Whitney test were applied to continuous variables . Kaplan-Meyer survival analysis of symptom presentation by day was performed controlling for early presentation , defined as presenting to hospital or health center in the first 3 days after onset of symptoms . Log-rank tests were performed to compare survival curves , and Cox regression models were created to determine risk factors associated with “compensated shock” , poor capillary refill , and cold extremities , controlling for dengue season , serotype , immune response , age , sex , and early presentation . Dummy variables were created for categorical variables with more than 2 values , utilizing as the reference the variable that presented least risk . Risk factors associated with DFCS as a disease classification were estimated using generalized linear model ( GLM ) multivariate analysis , controlling for the same variables as above . Confidence levels of 95% were used . In both studies , data were stored in Microsoft Access 2003 , and statistical analyses were performed using Intercooled Stata , Version 9 ( StataCorp ) .
Between August 2009 and June 2010 , 170 laboratory-confirmed , symptomatic cases of dengue were identified among the 3 , 711 active cohort participants , with 97 . 7% occurring between August and February ( Fig . 1A ) . Overall , a dengue case incidence of 4 . 6% was observed , compared to 0 . 4–1 . 8% in 2004–9 ( between 13 and 65 cases each dengue season , Table 1 ) . In the concurrent hospital-based study , 212 ( 54% of enrolled participants ) dengue cases were confirmed among 396 study participants during the enrollment period spanning August 2009–January 2010 ( Fig . 1B ) . While the design of the hospital study does not allow calculation of incidence , surveillance records indicate that 5–10 times more suspected dengue cases presented to the HIMJR in 2009–10 than in previous study years ( A . Balmaseda and M . A . Perez , unpublished data ) . The greatest numbers of cases were observed in September and October in both studies ( Fig . 1A and B ) . As in previous years , dengue was distributed evenly by gender ( 50 . 0% were female in both studies ) , and cases were encountered in subjects of all ages . The mean age of dengue cases in the cohort was greater in 2009–10 due to aging of the cohort , which by then extended to 13–14 years of age; no such age variation is seen in the hospital study , where subject ages were constant through all 5 years of the study . DENV-3 caused 88 . 8% ( hospital study ) and 83 . 9% ( cohort ) of cases in 2009–2010 . DENV-1 and -2 were also responsible for cases in both studies in 2009–10 , and one subject in the hospital study was positive for both DENV-3 and DENV-4 ( Table 1 ) . Roughly half of dengue cases in both studies experienced a primary immune response ( 47 . 6% in the cohort study and 52 . 2% in the hospital study ) . More secondary cases were observed in both studies prior to 2008 due to the dominant circulation of DENV-2 in 2005–8 , although only in the hospital study was this difference significant ( p<0 . 001 ) . A larger proportion of DENV infections in 2009–10 presented symptomatically ( 1 symptomatic case for every 1 . 2 inapparent DENV infections , as measured via the annual healthy blood sample ) , as compared to 2008–9 , when DENV-3 also predominated ( 1∶9 . 5 symptomatic to inapparent DENV infection ratio; p<0 . 001 ) . In both studies , the clinical presentation of dengue was markedly different in 2009–10 than in previous years ( Table 2 ) . Unique to the 2009–10 season was the significantly more frequent presentation of poor peripheral perfusion , specifically poor capillary refill ( >2 seconds ) and cold extremities . Significantly more patients experienced “compensated shock” ( poor capillary refill plus cold extremities , tachycardia , tachypnea , and/or weak pulse ) in 2009–10 than in 2005–8 ( 42 . 5% [90/212] vs . 24 . 7% [82/332] in the hospital study ( p<0 . 001 ) and 17 . 0% [29/170] vs . 2 . 2% [4/181] in the cohort study ( p<0 . 001 ) ) ( Table 2 ) . In Kaplan-Meier survival analysis , poor capillary refill and cold extremities presented significantly earlier ( 1–2 days ) in both studies when controlling for day of presentation ( log rank p<0 . 001 in both studies , Figs . 2A and B , 3A and B ) . Kaplan-Meier survival analysis of “compensated shock” as an entity comprising these signs naturally yielded similar results ( Fig . 2C ) . When controlling for serotype , immune response and early presentation in a Cox regression model , hospital study dengue cases in 2009–10 had 2 . 80 times more risk of presenting poor capillary refill ( 95% Confidence Intervals ( CI ) 1 . 85–4 . 23; p<0 . 001 ) , 2 . 13 times more risk of presenting cold extremities ( 95%CI 1 . 52–3 . 00; p<0 . 001 ) , and 2 . 81 times more risk of presenting “compensated shock” ( 95%CI 1 . 85–4 . 27; p<0 . 001 ) ; performing the same analysis among cohort cases produced similar results ( Table 3 ) . In addition , thrombocytopenia was seen less frequently in 2009–10 than in previous years in the hospital study ( 26 . 5% vs . 52 . 4% in 2005–9 , p<0 . 001 , Table 2 ) , although there was a trend toward more thrombocytopenia in the cohort study in 2009–10 . Clinical manifestations of hemorrhage were more frequently observed in both studies in 2009–10 , though this difference was only significant in the cohort study ( 70 . 0% vs . 53 . 0% , p<0 . 01 , and 84 . 1% vs . 78 . 3% , p>0 . 05 in the cohort and hospital studies , respectively ) . In both studies , more dengue-positive children required intensive care in 2009–10 than in previous years . In the cohort study , 11 . 2% ( 19/170 ) of all cases were admitted to the ICU , while only 1 . 1% ( 2/181 ) cases were transferred to intensive care in 2004–9 ( p<0 . 001 ) . In the hospital study , 19 . 8% ( 42/212 ) of patients with confirmed dengue were transferred to intensive care in 2009–10 compared to 7 . 1% ( 16/225 ) in 2007–9 ( p<0 . 001; ICU data is available in the hospital study beginning in 2007 ) . This increase in ICU transfer was due to markedly greater presentation of “compensated shock” in 2009–10 than in previous years . Children with signs of poor peripheral perfusion ( “compensated shock” ) were administered crystalloid fluid IV and if unresponsive , then colloids were given . When signs of poor peripheral perfusion persisted , children were transferred to the ICU . In the hospital study , of subjects presenting “compensated shock” , 96% ( 86/90 ) in 2009–10 received IV fluid therapy , and 40 . 7% ( 35/86 ) of those given liquids were transferred to the ICU . In 2005–9 , 83% ( 68/82 ) of patients with “compensated shock” received IV fluids , of which 23% ( 15/68 ) were transferred to the ICU ( during 2007–9 ) . The early administration of IV crystalloid and colloid fluids may well have contributed to preventing progression to hypotensive shock . For instance , in the hospital study , only 30% ( 27/90 ) of children experiencing “compensated shock” progressed to hypotensive shock; this could explain the reduced numbers of DHF/DSS noted in 2009–10 . In 2005–9 , 51% ( 42/82 ) of children with “compensated shock” progressed to hypotensive shock . When children who experienced “compensated shock” did not progress to hypotensive shock , they were classified as “Dengue Fever with Compensated Shock” ( DFCS ) . Significantly more cases of DFCS were observed in 2009–10 as compared to 2004–9 in the cohort study ( 15/170 [8 . 8%] vs . 4/181 [2 . 2%] ) and in the hospital study ( 50/212 [23 . 6%] vs . 23/332 [6 . 9%] ) ( Table 1 and Fig . 4 ) . Conversely , in the hospital study , significantly less DHF and DSS were seen in 2009–10 ( 11 . 3% and 2 . 8% , respectively ) compared to 2005–9 ( 24 . 7% and 14 . 2% , respectively , p<0 . 001 , Table 1 and Fig . 4B ) . However , more DSAS cases were seen in 2009–10 ( 28/212; 13 . 2% ) than in 2005–9 ( 16/332; 4 . 8% ) . Interestingly , in 2009–10 , DFCS , DSAS , and DSS cases were comprised of equivalent numbers of primary and secondary DENV infections . When multivariate analysis was performed to examine risk factors associated with classification as DFCS as compared to all other dengue cases , controlling for immune status , year 2009–10 versus 2005–9 , serotype , sex , and age ( <5 years old ) , only year of study 2009–10 emerged as significant ( RR 3 . 42 , 95%CI 1 . 67–7 . 01; p<0 . 001; Table 4 ) . Likewise , in multivariate analysis controlling for the same variables as above , only study year 2009–10 was a significant risk factor for DFCS as compared to uncomplicated dengue fever ( RR 3 . 02 , 95%CI 1 . 64–5 . 59; p<0 . 001 ) or for shock ( both compensated and hypotensive ) as compared to non-shock dengue cases ( RR 2 . 42 , 95%CI 1 . 62–3 . 11; p<0 . 001 ) . Beginning in 2008–9 and continuing through 2010–11 , DENV-3 caused the majority of dengue cases in both the cohort and hospital studies , differentiating these years from the previous study years in which DENV-2 predominated . In 2008–9 , 2009–10 and 2010–11 , DENV-3 was responsible for between 74 and 97% of cases in the hospital study , and between 80 and 90% of cases in the cohort study . We sequenced the full-length genome of 17 DENV-3 strains from the 2008–9 dengue season , 82 strains from the 2009–10 dengue season , and 28 samples from the 2010–11 dengue season; phylogenetic analysis revealed no changes in genotype or clade ( Fig . 5 ) . We then performed a sub-analysis among DENV-3 cases in the hospital study to control for the effect of serotype on clinical presentation . Compared to cases in 2005–9 and 2010–11 , more cases in 2009–10 were referred to the ICU ( 3 . 6% and 4 . 8% vs . 18 . 8% , p<0 . 001 ) and were classified as DFCS ( 7 . 1% and 10 . 6% vs . 21 . 2% , p<0 . 01 ) or DSAS ( 1 . 8% and 2 . 9% vs . 13 . 7% , p<0 . 001; Table 5 ) . Likewise , more DENV-3 cases in 2009–10 presented compensated shock and its associated signs , delayed capillary refill and cold extremities , than in 2005–9 or 2010–11 ( p<0 . 001 ) . Kaplan-Meier survivor curves indicate that these same signs presented significantly earlier in 2009–10 than 2005–9 and 2010–11 ( Fig . 6A–C ) . In Cox regression models controlling for early presentation and immune responses , cases in 2009–10 had increased risk of compensated shock , delayed capillary refill and cold extremities ( Table 6 ) . Finally , in multivariate analysis adjusting for age , sex and immune status , year 2009–10 cases had a 2 . 62 greater risk of developing DFCS compared with cases in 2005–9 or 2010–11 ( 95% CI 1 . 25–5 . 49 , p = 0 . 011 ) ( Table 7 ) . In the cohort , similar trends were seen among cases of DENV-3 in 2009–10 vs . 2010–11 ( data not shown ) , though the small number of cases of DENV-3 in the cohort in 2004–9 precluded a complete analysis as in the hospital study . The atypical , early onset of compensated shock and poor peripheral perfusion we observed is unique to the 2009–10 dengue season , when neither DENV serotype nor clade changed . Our leading hypothesis for what may have caused this unusual clinical presentation is the concurrent circulation of pandemic Influenza A during the 2009 dengue season , the only notable epidemiological distinction that occurred in 2009 . Usually , influenza peaks in June–July [26] , 2–3 months before the beginning of the annual dengue season . However , in 2009 , although Influenza A H1N1 virus began circulating in Managua in June , the influenza pandemic was more prolonged that year [26]; incidence reached a maximum in August and overlapped for 8–10 weeks with the 2009–10 dengue epidemic ( Fig . 7A and B ) . One hypothesis is that recent infection with Influenza A H1N1 2009 might have modulated the immune response to a subsequent DENV infection . Therefore we tested for the presence of anti-H1N1 2009 antibodies in 2009–10 hospital dengue cases . In serum samples from 187 DENV-positive subjects in the hospital study in 2009–10 , 65 . 7% ( 123 ) contained antibodies specific to the Influenza A H1N1 2009 strain , indicating recent infection with the virus . Subjects with H1N1 antibodies had significantly greater odds of developing compensated or hypotensive shock ( DFCS , DSAS or DSS ) compared with subjects without H1N1 antibodies ( 43% vs . 28% , respectively; OR 1 . 93 , 95% CI 1 . 01–3 . 31; p = 0 . 045 ) .
In this paper , we have described an unusually large dengue epidemic with a unique clinical presentation characterized by early presentation of poor peripheral perfusion . In our pediatric cohort study , we observed incidence rates more than twice as high as in previous study years ( 4 . 6% vs . <2% ) , a finding that reflected observations in our parallel hospital-based study . Both the epidemiologic and clinical trends reported in this study were also observed by national authorities and reflected in the national surveillance statistics for the 2009–10 dengue season in Managua . This epidemic was associated with atypical presentation of dengue , with signs of poor peripheral perfusion presenting 1–2 days earlier than usual in the course of illness and less progression to DHF/DSS in both studies . The usual evolution of dengue toward severe disease involves rising hematocrit and falling platelet count followed by onset of compensated shock leading to hypotensive shock ( hypotension for age or narrow pulse pressure ) accompanied by clinical signs of shock ( e . g . , poor capillary refill , cold extremities ) . This latter is termed the critical phase and generally occurs on days 4–5 ( range 3–7 ) after onset of symptoms [2] , [27] . Compensated shock had not previously been seen on a large scale in Nicaragua , nor has it been reported elsewhere as a defining clinical feature of dengue without progression to hypotensive shock . In 2009–10 , study physicians treated 96% of subjects presenting “compensated shock” with IV therapy compared to 83% in previous study years . Those cases that did not respond to crystalloid fluid therapy were administered colloids , and if still no improvement was seen , patients were transferred to the ICU , and if necessary , administered amine vasopressor agents . Only 31% of those administered crystalloids were not responsive , suggesting that such therapy limited progression to more severe disease . The recently revised WHO guidelines for dengue management suggest treatment of compensated shock with crystalloids [27] . The early presentation of compensated shock in the 2009–10 epidemic coupled with early IV fluid administration may well have enhanced the effect of IV fluid treatment on the course of disease . DENV serotype and immune response have been linked to variations in clinical presentation of dengue . However , in this study , in multivariate analysis controlling for serotype , immune response , sex , and age less than 5 years , presenting in 2009–10 was the most significant risk factor for DFCS compared with uncomplicated dengue or all other dengue outcomes , and similar results were obtained with DFCS , DSAS , and DSS compared to dengue without shock . Presenting in 2009–10 was also the most significant risk factor for early presentation of signs of poor peripheral perfusion and compensated shock in Cox regression models . DENV-3 was responsible for the majority of cases in the 2008–9 and 2009–10 seasons , with no change in frequency or viral sequence . Another large dengue epidemic during the 2010–11 season was dominated by the same clade of DENV-3 , yet clinical presentation was once again typical and involved neither high frequencies nor early presentation of “compensated shock” as in 2009–10 . In addition , the sequence of DENV infections and immune status appear not to be factors in the unique clinical presentation of DENV-3 in 2009–10 . In the hospital , where more severe secondary cases are typically seen , 2009–10 had significantly more primary infections than previous study years ( p<0 . 001 ) and similar frequencies of primary cases as in 2008–9 – consistent with the higher proportion of primary cases often observed with DENV-3 [3] , [4] , [9] . This suggests that DENV immune response did not impact severity , as the clinically severe cases of 2009–10 were just as likely to be primary as secondary DENV infections . Lastly , multivariate analysis ruled out immune status ( i . e . , secondary infection and thus sequence of DENV serotypes ) as responsible for the unusual clinical characteristics of dengue in 2009–10 . Thus , our data indicate that the distinct disease phenotype observed in 2009–10 was not due to DENV serotype , viral sequence , immune status , or serotype sequence of sequential DENV infections and lends support to the hypothesis that something about the year 2009–10 was different . In addition , the ratio of symptomatic to inapparent DENV infections increased in 2009–10 ( 1∶1 . 2 ) relative to 2008–9 ( 1∶9 . 5 ) , implying that the overall incidence of symptomatic DENV infection was substantially higher in 2009–10 . While early IV fluid intervention and increased hospitalization may have affected disease progression , decreasing plasma leakage and other typical symptoms of severe disease in 2009–10 , case management could not have affected the early presentation of the symptoms of poor peripheral perfusion in the first place , since the children arrived at the hospital with these signs and symptoms prior to any interventions . Currently , our favored hypothesis is that there may be an interaction between a recent Influenza A H1N1 2009 infection and the subsequent DENV infection , as high rates of Influenza A H1N1 2009 in the cohort and in Managua as well as a number of identified DENV/H1N1 2009 co-infections [28] suggest that concurrent or previous H1N1-2009 infection may have influenced the atypical clinical presentation of dengue cases in 2009–10 . The influenza pandemic occurred later than usual ( August–September rather than June–July ) in 2009–10 and thus preceded/overlapped the annual dengue epidemic in an unusual way that is not normally seen in Nicaragua [26] , evidenced in both the cohort and the national surveillance system . We posit that a prior influenza infection may be modulating the subsequent immune response to DENV infection , perhaps by reducing interferon levels [29] , which could lead to greater DENV infection , as it is known that type I interferons are powerful antagonists of DENV infection [30] , [31] . Influenza infections are known to predispose towards and enhance the severity of secondary bacterial infections [32] , [33] , [34] , [35] , [36]; perhaps in close temporal proximity , a similar effect could occur on a subsequent DENV infection . To investigate this hypothesis , a case control study is underway in which acute samples from 2009 dengue cases in the cohort and hospital studies with or without compensated/hyptotensive shock are being tested for the presence of antibodies specific to the Influenza A H1N1 2009 strain . Initial analyses presented here support this hypothesis . Cytokine profiles and viremia in these samples are also being investigated . In addition , we are establishing a mouse model of influenza virus followed by DENV infection to explore the existence and then mechanistic underpinnings of this proposed immunomodulation . In summary , the 2009–10 dengue epidemic in Managua was large and involved an atypical presentation of the disease , with early onset of signs of poor peripheral perfusion ( “compensated shock” ) , as observed in both long-term community-based and hospital studies . Multivariate analysis controlling for usual risk factors associated with severe dengue revealed only study year 2009–10 as a significant risk factor for DFCS . Another unusual feature of 2009 in Managua was the circulation of pandemic influenza A H1N1 , which for the first time overlapped with the dengue season . We postulate that the unusual presentation of dengue in 2009 may have been due in part to immunomodulation by a prior influenza H1N1-2009 infection and are currently testing this hypothesis . Overall , this study demonstrates that parameters other than DENV serotype , viral genomic sequence , immune status , and sequence of serotypes can play a role in modulating dengue disease outcome .
|
Dengue is the most prevalent mosquito-borne viral disease affecting humans worldwide . The four dengue virus serotypes ( DENV1–4 ) cause Dengue Fever and more severe life-threatening syndromes . In 2009 , Nicaragua experienced the largest dengue epidemic in over a decade . In a hospital-based study and community-based prospective cohort study of pediatric dengue in the capital , Managua , we observed unusual clinical presentation of dengue . Significantly more patients experienced “compensated shock” ( poor capillary refill plus cold extremities , rapid heartbeat , elevated respiratory rate , and/or weak pulse ) in 2009–10 than in previous years . These signs of poor peripheral perfusion presented significantly earlier and more children were transferred to intensive care in 2009–10 than in previous years . DENV-3 was the predominant serotype in Managua in 2008–9 , 2009–10 and 2010–11 , but full-length sequencing revealed no major genetic changes . In 2008–9 and 2010–11 , typical dengue was observed; only in 2009–10 was unusual presentation noted . Since pandemic influenza A H1N1-2009 overlapped with the dengue epidemic in Nicaragua , we hypothesize that prior influenza A H1N1-2009 infection may have modulated subsequent DENV infection , and preliminary results appear to support this hypothesis . This study demonstrates that parameters other than DENV serotype , viral genomic sequence , immune status , and sequence of serotypes can play a role in modulating dengue disease .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"Methods",
"Results",
"Discussion"
] |
[
"medicine",
"infectious",
"diseases",
"clinical",
"research",
"design",
"virology",
"epidemiology",
"global",
"health",
"neglected",
"tropical",
"diseases",
"biology",
"microbiology",
"viral",
"diseases"
] |
2011
|
Unusual Dengue Virus 3 Epidemic in Nicaragua, 2009
|
The heavy consumption of ethanol can lead to alcohol use disorders ( AUDs ) which impact patients , their families , and societies . Yet the genetic and physiological factors that predispose humans to AUDs remain unclear . One hypothesis is that alterations in mitochondrial function modulate neuronal sensitivity to ethanol exposure . Using Drosophila genetics we report that inactivation of the mitochondrial outer membrane translocator protein 18kDa ( TSPO ) , also known as the peripheral benzodiazepine receptor , affects ethanol sedation and tolerance in male flies . Knockdown of dTSPO in adult male neurons results in increased sensitivity to ethanol sedation , and this effect requires the dTSPO depletion-mediated increase in reactive oxygen species ( ROS ) production and inhibition of caspase activity in fly heads . Systemic loss of dTSPO in male flies blocks the development of tolerance to repeated ethanol exposures , an effect that is not seen when dTSPO is only inactivated in neurons . Female flies are naturally more sensitive to ethanol than males , and female fly heads have strikingly lower levels of dTSPO mRNA than males . Hence , mitochondrial TSPO function plays an important role in ethanol sensitivity and tolerance . Since a large array of benzodiazepine analogues have been developed that interact with the peripheral benzodiazepine receptor , the mitochondrial TSPO might provide an important new target for treating AUDs .
Alcohol is one of the most widely used drugs worldwide , but long term consumption leads to its abuse and dependence . An estimated 17 . 6 million adults in the United States have AUDs with associated health concerns of alcohol dependence , liver cirrhosis , cancer , and injuries . From 2006 through 2010 , this generated an annual average of about 88 , 000 alcohol-related deaths and 2 . 5 million years of potential life lost [1 , 2] . To develop therapeutic strategies for alcoholism it will be necessary to determine the molecular and cellular mechanisms underlying AUDs . Considerable effort has been invested in determining the role of the central nervous system in the etiology of AUD [3–5] but many features of the AUDs remain unexplained . Neuronal function is highly dependent on mitochondrial bioenergetics [6 , 7] . In addition to the direct metabolizing of ethanol , the mitochondria are central to a wide range of essential neuronal cell functions including ATP synthesis , ROS production and REDOX homeostasis , Ca2+ buffering , and the metabolic regulation of apoptosis [8–10] . In humans mitochondrial DNA ( mtDNA ) alterations have been correlated with alcoholism , involving both acute ethanol responses and chronic damage [11–16] . In rodents , hepatic mtDNA depletion is seen in alcohol exposed mice [17] and mtDNA complex I gene variants have been correlated with “non-drinker” versus “drinker” rat lines derived from the same founder strain [18] . Variation in the mtDNA genes have also been shown to have profound effects of nuclear gene expression [19] . In previous studies we showed that the nuclear DNA coded Drosophila translocator protein 18kDa ( dTSPO , CG2789 ) is localized in outer mitochondrial membrane ( OMM ) and important for regulating mitochondria bioenergetics , ROS production , caspase activity , and apoptotic function [20] . In humans , TSPO ligands are widely used in neuroimaging for neurodegenerative diseases and neuronal injuries , both of which are associated with increased brain TSPO levels and distribution [21] . As the previous nomenclature ( peripheral benzodiazepine receptor , PBR ) implies , TSPO binds the benzodiazepines and other psychotrophic drugs associated with tolerance and addiction [22] . Thus we hypothesized the TSPO may be an important factor in addiction to ethanol . Drosophila’s sensitivity and tolerance to ethanol are similar to humans and rodents . Ethanol results in biphasic locomotor alterations . At lower doses ethanol acts as a stimulant , but at higher doses it acts as a depressant [23] . After repeated alcohol stimulation , tolerance is developed , defined as acquired resistance . Tolerance is thought to be an intermediate step to alcohol dependence and addiction [24] . Here we report that in male Drosophila , neuronal inactivation of dTSPO sensitizes flies to ethanol sedation , mediated by increased ROS production and decreased caspase activation . Furthermore , systemic but not neuronal loss of dTSPO inhibits the development of tolerance . By contrast , females are constitutively more sensitive to ethanol sedation than males and they have much lower dTSPO mRNA in their brains . Therefore , the mitochondrial TSPO is an important mediator of ethanol sensitivity and tolerance and contributes to gender-specific differences in alcohol sensitivity .
Acute ethanol sensitivity was analyzed by placing flies in vials closed by cotton clogs soaked with varying concentrations of ethanol thus exposing the flies to ethanol vapor . During initial exposure the flies flew to the top of the vial , and exhibited hyperactivity for a few minutes . With longer exposure , the flies became sedated and remained at bottom of the vial without locomotion . Wild type flies became comparably sedated whether the ethanol-soaked clogs were at the top of the vials or the vials were inverted with clogs at the bottom ( S1 Fig ) . Moreover , the Drosophila showed a dose-dependent response to ethanol using this protocol ( Fig 1 ) . Therefore , the ethanol effects observed in the following experiments were due to the ethanol concentration rather than an environmental factor such as hypoxia due to ethanol vapor exclusion of air . The tspo[EY00814] mutant Drosophila has a P-element inserted into the tspo gene leading to loss of dTSPO expression [20] . Male tspo-/- flies exhibited higher sedation sensitivity than tspo +/+ flies when exposed to ethanol vapor from 34% ethanol solution ( Fig 1A ) while at 44% or 54% ethanol vapor both the tspo-/- and tspo +/+ flies exhibited the same sensitivity ( Fig 1B and 1C ) . Post sedation , we tested for the recovery from ethanol sedation by replacing the ethanol-soaked clogs with normal clogs . This revealed that at 54% ethanol exposure tspo-/- males were slower to recover than the tspo +/+ male flies ( Fig 1D ) . Thus , tspo-/- male flies are more sensitive to ethanol sedation than their tspo +/+ counterparts . While the rempA gene overlaps with the tspo gene , rempA-/- deficiency is not the cause of the ethanol phenotypes since rempA-/- flies exhibit comparable sensitivity to 34% ethanol vapor as wild type flies ( S2 Fig ) . Since the tspo mutation is present in all developmental stages of the fly , it could act through creating a developmental abnormality . However , Hematoxylin-Eosin histological comparison of the brains of male tspo-/- and +/+ flies did not reveal any gross anatomical differences ( S3 Fig ) . To determine whether the increased ethanol sensitivity was attributable to dTSPO function in neurons , we depleted dTSPO in neurons by inducing dsRNA ( RNAi ) to knockdown dTSPO mRNA in adult flies following eclosion ( days after eclosion , dae ) . We used the Gal4-GeneSwitch/UAS system [25] in which Gal4 is activated within the flies when fed with mifepristone ( RU486 ) . The activated Gal4 binds to the UAS of the UAS-dTSPO-RNAi which induces the dsRNA expression and inhibition of the dTSPO mRNA . Since the Gal4 element is expressed under the neuronal specific elav promoter ( elav-GeneSwitch ) , this switch was restricted to neurons . In this way , the flies were permitted to progress through larval and pupal development with normal TSPO activity , and following eclosion , the dTSPO RNAi was induced in neurons by exposure to RU486 . Male flies harboring both elav-GeneSwitch and UAS-dTSPO-RNAi ( elav-GS/+; TSPO-IR/+ ) ( GS means ‘Gene Switch’ and IR means ‘Inverted Repeats’ ) cassettes that were exposed to RU486 post eclosion had reduced head dTSPO mRNA as quantified by RT-PCR ( Fig 2A ) . Therefore , activation of the elav-GS/+; TSPO-IR/+ system with RU486 specifically depletes dTSPO mRNAs in the neurons . In parallel with the whole body knockouts , the elav-GS/+; TSPO-IR/+ RU486 knockdown male flies exhibited faster ethanol sedation in the presence of 44% ethanol vapor than did flies who were not exposed to RU486 ( Fig 2B ) . RU486 exposure of flies harboring only the neuronal elav-GeneSwitch ( elav-GS/+ ) or the UAS-dTSPO-RNAi ( TSPO-IR/+ ) cassette had no effect on the ethanol sensitivity ( Fig 2C and 2D ) . Similarly , elav-GS/+; TSPO-IR/+ male flies exposure to 34% ethanol also showed increase sedation after RU486 induction relative to uninduced flies ( S4A Fig ) . After 55% ethanol sedation ( S4B Fig ) , the RU486-induced flies were slower to recover ( S4C Fig ) . The difference between the RU486 induced and uninduced flies was not due to differential alcohol absorption or metabolism since after a brief exposure to 44% ethanol vapor both groups of fly heads ( with and without RU486 ) had the same ethanol concentration ( S4D Fig ) . Hence , dTSPO inactivation in adult neurons is sufficient to sensitize male flies to ethanol exposure . Male and female flies exhibit sexual dimorphic response to ethanol exposure [4] and this sexual dimorphism was also observed in the brains of the TSPO knockout and knockdown flies . Male tspo-/- flies showed an increased sensitivity to ethanol sedation relative to tspo +/+ flies with 34% ethanol exposure and delayed recovery from 54% sedation ( Fig 1A and 1D ) while female tspo-/- and tspo +/+ flies showed no difference in their response to 34% ethanol exposure ( Fig 3A–3C ) . In elav-GS/+; TSPO-IR/+ female flies , after neuronal inactivation of dTSPO by dsRNA expression , there was no effect on the sedation rate with exposure to 34% or 44% ethanol solution ( S5A and S5B Fig ) . Furthermore , only slightly delayed recovery was seen for female flies after 44% vapor sedation ( S5C Fig ) . The marked difference between male and female elav-GS/+; TSPO-IR/+ flies’ sensitivity to ethanol following dTSPO inactivation by RU486 induction correlated with male fly heads having about four times the level of dTSPO mRNA as female heads . Moreover , neuronal knockdown of dTSPO reduced male head TSPO mRNA level but had no effect on female head TSPO mRNA level ( Fig 2A ) . Therefore , the lack of sensitivity of female flies to neuronal inactivation of dTSPO is likely do to a gender-specific lack of TSPO in female fly brains . To determine what might be the physiological basis of the neuron-specific effects of dTSPO deficiency on male ethanol sedation , we examined the effects on ROS production , which we previously found was increased in dTSPO deficient mitochondria [20] . ROS has been identified as modulator of neuronal activity [26] . Using Amplex Ultrared to determine the amount of H2O2 in fly heads , we found that H2O2 levels were higher in male elav-GS/+; TSPO-IR/+ flies treated with RU486 than untreated flies ( Fig 4A ) . Hence , neuronal dTSPO knockdown increased fly head H2O2 production . When these flies were fed with N-Acetyl-L-Cysteine ( NAC ) , an efficient antioxidant , the enhanced sedation effect of the dTSPO knockdown flies to 44% ethanol vapor was negated ( Fig 4B ) . In tspo +/+ male flies , exposure of 44% ethanol vapor for 20 minutes resulted in sedation of most of the flies but did not significantly alter H2O2 content in fly heads ( Fig 4C ) . Hence , the increase in ROS is not caused by ethanol exposure . Rather , dTSPO inactivation in neurons up-regulates ROS and the increased ROS is responsible for the enhanced ethanol sensitivity of the dTSPO-depleted flies . Since ethanol sedation sensitivity was controlled by TSPO and TSPO expression declines in tspo +/+ flies to a minimum at 30 dae ( S5B Fig of [20] ) , we determined whether ethanol sensitivity changes during aging . Male tspo +/+ and tspo-/- flies were tested with 44% and 54% ethanol sedation at different ages i . e . young ( about 5 dae ) , mid-age ( about 20 dae ) , and old ( about 35 dae ) . Wild type ( tspo+/+ ) flies exhibited increased sedation as they aged , with the effect already evident by 20 dae . tspo-/- flies also displayed and increased predilection to sedation with age ( S6C Fig ) , but they were initially significantly more sensitive to ethanol . This is consistent with their higher level of oxidative stress as demonstrated by the marked reduction in their ROS-sensitive mitochondrial aconitase activity ( Fig 6 of [20] ) . Depletion of dTSPO in flies suppresses caspase activation and impedes apoptosis [20] . However , caspase also has cell death-independent functions which might be involved in neuronal control [27] . The activity of caspase 3/7 , the most downstream caspase in the intrinsic apoptosis pathway , was decreased in heads of flies with dTSPO-depleted neurons ( Fig 5A ) . Neuronal expression of the caspase inhibitor protein , p35 , also reduced caspase 3/7 activity to a similar degree as induction of the TSPO dsRNA ( Fig 5A ) . The level of caspase reduction in TSPO knockdown and p35 induced neurons is likely to be much greater than shown in Fig 5A where whole brain homogenates were assayed . Whole brain homogenates mix the enzymes of all cell types most of which are not neurons and thus not subject to dTSPO knockdown . Supporting this speculation , caspase 3/7 activity of whole body homogenate tspo-/- flies was tenfold lower than that of tspo +/+ flies ( Fig 5 , legend ) . Neuronal expression of the caspase inhibitor protein , p35 , also increased sensitivity of male flies to ethanol sedation when exposed to 34% ethanol vapor ( Fig 5B and 5C ) . This phenocopyied the dTSPO knockdown flies and confirmed the importance of reduced neuronal caspase 3/7 in ethanol sensitivity . Hence , both increased ROS production and decreased caspase activity in neurons are important in enhanced ethanol sensitivity . To investigate the development of ethanol tolerance ( reduced ethanol sensitivity following repeated ethanol exposures ) , we exposed flies to ethanol , allowed them to recover for 6 hours , and then exposed the flies to the same ethanol concentration again and monitored their sedation . In tspo +/+ male flies , the sedation for second exposure to 54% ethanol solution vapor was significantly delayed compared with first exposure ( Fig 6A ) , indicating tolerance formation . However , tspo-/- male flies exhibited no diminished sedation sensitivity between the first and second ethanol exposure ( Fig 6A ) . Hence , the systemic inactivation of dTSPO prevented male flies from developing tolerance . In contrast to male flies , female tspo-/- flies developed ethanol tolerance similar to tspo +/+ flies ( S7 Fig ) . Hence , loss of ethanol tolerance in tspo-/- flies is also gender-specific . To determine if the effect of dTSPO on tolerance is attributable to neurons , we compared elav-GS/+; TSPO-IR/+ flies with or without RU486 to induce TSPO dsRNA . Knockdown of dTSPO in adult male neurons had no effect on the development of tolerance following a second ethanol exposure to 54% ethanol vapor ( Fig 6B ) . Hence the suppression of tolerance in dTSPO-depleting flies was not driven by neuronal dTSPO levels . Since there might be other cell types in which dTSPO functions in tolerance formation , we isolated the heads and bodies of male tspo +/+ flies to examine the expression of dTSPO during tolerance . Within 4 hours after first exposure to 54% ethanol vapor , the amount of dTSPO mRNA in heads was decreased while the dTSPO mRNA in bodies was markedly increased . Both head and body dTSPO mRNA levels began to normalize at 6 hours post exposure ( Fig 6C and 6D ) . Therefore , tolerance is associated with the induction of dTSPO in fly bodies , which is consistent with the loss of the capacity to develop tolerance in tspo-/- flies but not in elav-GS/+; TSPO-IR/+ induced flies .
The involvement of TSPO-mediated increased neuronal ROS production and decreased caspase activity in the sensitivity to ethanol sedation is consistent with reports that oxidative stress and caspase-mediated apoptosis contribute to brain pathology [28] . Since TSPO controls mitochondrial ROS production and caspase activation [20] , it follows that modulation of ROS levels and caspase activity could mediate ethanol sensitivity . Inactivation of tspo increased ROS production and NAC negated the enhanced sensitivity to ethanol demonstrating that increased neuronal ROS is related to increased ethanol sedation sensitivity . Given the short exposure period of the flies to ethanol , the ROS effect is most likely due to its second messenger action [26] rather than due to a cell death mechanism . This is consistent with the recent report showing that expression of oxidative stress genes can be altered by ethanol exposure and their functions are essential for ethanol sensitivity [29 , 30] . It is possible that TSPO-deficiency induced ROS production could also participate in development of tolerance , but this effect must be mediated by cells other than neurons . Inactivation of dTSPO also inhibits caspase activity[20] and inhibition of neuronal caspase activity also sensitized flies to ethanol sedation . This was confirmed by expression of the caspase inhibitor p35 resulting in increased ethanol sensitivity . Since caspase has been shown to function in neuronal apoptosis-independent pathways to control neuronal activity in both developmental and adult stages[27] , it is reasonable to conclude that dTSPO depletion in fly neurons activates such pathways thus altering neuronal activity and ethanol response . The male-specific effects of TSPO inactivation were particularly striking . Previous studies have shown that male flies are more resistant to ethanol-induced sedation than females [31] , which we also observed . Inactivation of dTSPO in males increased their sensitivity to ethanol , bringing their sensitivity close to that of females . Furthermore , female flies were found to have much lower dTSPO mRNA in their heads than males and knockdown of neuronal dTSPO in male heads reduced dTSPO mRNA about 20% while having no effect on the dTSPO levels of female fly heads . Thus , female flies have inherently low expression of dTSPO in their neurons and this may account for to their increased sensitivity to ethanol sedation . In humans , men and women also exhibit different responses to acute and long-term ethanol exposure [32 , 33] . Men are at higher risk of AUD than women , but once AUD develops , women are more susceptible to ethanol-induced damage in multiple organs . Perhaps differences in TSPO expression contribute to human gender differences as well . The molecular basis for the differences in dTSPO expression in flies is unknown . Male flies express a male specific splicing isoform of neuronal sex determination gene fruitless ( fru ) , FruM . This may control the gender-specific production of neurotransmitters and neuropeptides [31] . Such a system might also regulate dTSPO expression . Additional environmental factors to which male and female animals are differentially exposed may also affect dTSPO expression . A variety of genes have been reported to control fly brain development and impact ethanol responses [5] . Since the tspo mutation affects all developmental stages in fly , it’s deficiency could create a developmental abnormality that alter ethanol sensitivity . However , Hematoxylin-Eosin histological staining of tspo-/- brains did not reveal any gross anatomical defect compared with tspo+/+ brains . Furthermore , by using the RU486-inducible gene switch system to knockdown dTSPO only after eclosion , we avoided any alterations in fly anatomy demonstrating that only physiological changes were important in ethanol sensitivity of adults . Hence , the ethanol sensitivity induced in male flies by the knockdown of dTSPO cannot be due to developmental alterations , but must be the product of the physiology of the adult neurons . This means that physiological modulation should be able to treat alcoholism . The knockdown of dTSPO in neurons demonstrates that neuronal expression of TSPO is important in determining ethanol sensitivity . This neuronal action of TSPO is at variance with reports in mammals that TSPO probes co-localize primarily with glial [34 , 35] . That dTSPO must be expressed in neurons is not only confirmed by the current ethanol studies but also by our previous observations that systemic depletion of dTSPO protects flies from toxicity of neuronally-expressed Aβ42 [20] . Unfortunately , our current data do not indicate if the ethanol sensitivity effects of dTSPO knockdown are related to a specific group of neurons . Mammalian TSPO has been reported to function in hippocampal neurons to affect long-term potentiation and learning[36] . Also , ethanol effects have been reported for the KCNQ channel expressed in dopaminergic neurons[37] and PKA expressed in insulin-producing neurons [38] . Benzodiazepines are widely used for treatment of anxiety , insomnia , seizures and other neural disorders , and are known to enhance the effects of GABA at the GABAA receptor . However , long-term use of these drugs is controversial due to decreasing effectiveness , physical dependence , and withdrawal [39 , 40] . TSPO is also a target of benzodiazepines and our results suggest that benzodiazepine-derived antagonists might increase sensitivity to ethanol and decrease neurological damage [20] while benzodiazepine-derived agonists could have the opposite effects . Consequently , the TSPO may provide an important drug target for treatment of drug abuse and alcoholism [36] which could be conveniently investigated with the current system . Our data demonstrate that the mitochondrial TSPO protein , also known as the peripheral benzodiazepine receptor , is important in determining both ethanol sensitivity and the development of ethanol tolerance . Given the existing of a broad range of benzodiazepine analogues , these compounds may provide a novel approach for treating AUDs .
Flies were raised on standard cornmeal medium in narrow ( 25x95mm ) vials at 25°C , with 12 hours/12 hours light/dark cycles . The tspo[EY00814] strain , obtained from the Bloomington Drosophila Stock Center ( Bloomington , IN , USA ) , has a P-element insertion in the 3' regulatory region of tspo gene . The UAS-dTSPO-RNAi stock was obtained from Vienna Drosophila RNAi Center ( VDRC , Vienna , Austria ) and contained a transgene which can be transcribed into a dsRNA that targets the dTSPO mRNA . Pan-neuronal gene switch Gal4 driver , elav-GeneSwitch , was also obtained from Bloomington Drosophila Stock Center . These strains were all backcrossed to w1118 ( isoCJ1 ) background . UAS-p35 stock was kindly provided by Dr . Nancy Bonini in University of Pennsylvania . The rempA[e02928] strain was also obtained from Bloomington Drosophila Stock Center . To induce gene switch , flies combining elav-GeneSwitch with UAS-dTSPO-RNAi ( elav-GS/+;TSPO-IR/+ ) or UAS-p35 ( elav-GS/+;UAS-p35 ) were raised in regular food with 50 μl 4 mg/ml ethanol solution of mifepristone ( RU486 , Sigma-Aldrich , St Louis , MO , USA ) added on the surface of food in vials for 3 days . As control , the flies were raised in regular food with 50 μl ethanol . The food vials were changed every 24 hours . In N-Acetyl-L-Cystein ( NAC ) experiments , 20 μl 500 mM NAC ( Sigma-Aldrich ) water solution or pure water was pre-mixed with 50 μl RU486 solution or ethanol solvent and then added on the surface of food in vials . In NAC experiments , the drug feeding was extended to 5 days . Flies at 4–7 dae age were used for all sedation , recovery and tolerance assays , except for the NAC experiment where 6–9 dae flies were used . In the aging experiments , 19–22 or 34–37 dae flies were studied . Flies were sorted under CO2 and loaded into empty narrow ( 25×95mm ) vials . Ten flies were loaded into a vial as a single trial and allowed to recover for at least 2 hours before use . Ethanol solutions of 34% , 44% , 54% ( weight/vol ) were made by mixing absolute ethanol ( Sigma-Aldrich , catalog number E7023 , for molecular biology ) and ultrapure distilled water ( Gibco , Grand Island , NY , USA ) at the ratio ( vol/vol ) of 4:6 , 5:5 , and 6:4 , respectively . For each vial , regular cotton clog was replaced with clog added with 1 ml ethanol solution at the vial-side surface . Recording to number of sedated flies started immediately . The interval for recording was 2 or 5 minutes . To monitor the recovery , the ethanol-containing clog was replaced with regular clog immediately after all flies were sedated . The number of flies remaining sedated was counted every 2 or 5 minutes . For tolerance assays , the flies were transferred into regular food vial with regular clog after all flies were sedated . Four hours later , the flies were transferred back into empty vial and the recording for sedation was performed as same as in naïve flies . Internal ethanol content was measured with Abcam Ethanol Assay Kit ( Ab65343 , Abcam , Cambridge , MA , USA ) . In brief , twenty flies at 4–7 dae age were CO2 anesthetized and loaded into empty narrow vial . After 2 hours recovery , flies were exposed to ethanol vapor from cotton clog soaked with 44% ethanol solution for 6 minutes when >90% flies were inactive . Then flies were quickly frozen in liquid nitrogen and homogenized in lysis buffer provided by kit and then centrifuged for 14000 g for 10 min in 4°C . The diluted sample together with standard ethanol samples were incubated in 96-well plate wells with ethanol oxidation reaction mix to produce H2O2 which further reacts with the probe in the mix to generate color . The absorbance at 570 nm was measured with a plate reader ( SpectraMax Paradigm , Molecular Devices , Sunnyvale , CA , USA ) . The content of ethanol was calculated based on the standard curve , and finally normalized by the total protein concentration measured by the Bradford method . Total RNA was extracted from bodies of 20–40 flies or 100 fly heads using RNeasy Mini Kit ( Qiagen , Valencia , CA , USA ) . The RNA was converted to cDNA using oligo ( dT ) 15 ( Invitrogen , Grand Island , NY , USA ) and SuperScript II reverse transcriptase ( Invitrogen ) . After reverse transcription , PCR reactions were performed using a ViiA7 Real-Time PCR System ( Applied Biosystems , Grand Island , NY , USA ) with SYBR Green Master Mix ( Applied Biosystems ) and primers for rp49 ( forward , 5- gctaagctgtcgcacaaatg -3 , and reverse , 5- ccaggaacttcttgaatccg -3 ) or dTSPO ( forward , 5- ctcttcgtaccctacgtcgc -3 , and reverse , 5- ctggttcgataggtcggaaa -3 ) . The PCR protocol involved denaturation at 95°C for 15 seconds and combined annealing and extension at 60°C for 1 min over 40 cycles . The melting curve was generated after these cycles to ensure that the amplification in each reaction was specific . Isolated fly heads or whole bodies were homogenized in Homogenization Buffer ( 225 mM mannitol , 75 mM sucrose , 10 mM MOPS , 1 mM EGTA , pH 7 . 2 ) on ice , then centrifuged at 300 g for 5 min . The supernatant was collected and added in 96-well plate wells together with an equal volume of reaction buffer ( ApoONE kit , Promega , Fitchburg , WI , USA ) . The plate was shaken gently for 5 min , and then incubated in dark for 15 hours in room temperature . Fluorescence was measured with a plate reader ( SpectraMax Paradigm , Molecular Devices ) with the excitation at 499 nm and emission at 521 nm . The fluorescent values were normalized by total protein concentration measured by the Bradford method , and the relative activity was calculated based on the ratio of normalized fluorescent signals between samples . Isolated fly heads were homogenized in Homogenization Buffer on ice . The samples were then centrifuged at 14000 g for 10 min in 4°C to collect the supernatant . The standard reaction solution containing 0 . 1 mM Amplex UltraRed , Invitrogen and 0 . 2 U/L horseradish peroxidase ( Thermo Scientific , Pittsburgh , PA , USA ) diluted in Homogenization Buffer was placed in 96-well plate wells . Then the fly extract samples or standard H2O2 samples were added to the plates and incubated for 15 min in the dark at room temperature . The fluorescence was measured with a plate reader ( SpectraMax Paradigm , Molecular Devices ) with the excitation at 530 nm and emission at 590 nm . The H2O2 content was calculated based on standard and normalized to the total protein concentration measured by the Bradford method . Fly heads were fixed in standard Bouin's Fixative , embed in paraffin blocks , and sectioned at a thickness of 6 μm . Sections were placed on slides , stained with haematoxylin and eosin ( Vector ) , and examined by bright-field microscopy .
|
Alcohol use disorders ( AUDs ) affect millions of patients worldwide and result in high social and economic burdens . Although environmental factors are involved , there are clear genetic components to AUDs . Both the acute sedating effect of alcohol exposure and alcohol tolerance contribute to long term risk for alcohol dependence and addiction . Yet the genetic etiology of AUDs remains to be determined . The mitochondria play a central role in ethanol metabolism and are important in many aspects of cellular physiology such as REDOX and ROS regulation , and apoptosis . The mitochondrial outer membrane translocator protein 18 kDa ( TSPO ) binds the benzodiazepines and perhaps other addictive drugs , and thus may play a role in AUDs . Since Drosophila is a well-established model for ethanol-related behaviors , we have developed systems for manipulating the Drosophila tspo gene and protein . With these systems , we have discovered that neuronal TSPO controls sensitivity to ethanol sedation via ROS and caspase-mediated signaling and that systemic TSPO levels are important in the development of tolerance to repeated ethanol exposure . Given the variety of known TSPO ligands , and the common mechanisms of various abusive substances , our studies suggest that TSPO might be a promising target to combat alcoholism as well as addiction to other drugs .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2015
|
TSPO, a Mitochondrial Outer Membrane Protein, Controls Ethanol-Related Behaviors in Drosophila
|
The intestinal microbiota is a microbial ecosystem of crucial importance to human health . Understanding how the microbiota confers resistance against enteric pathogens and how antibiotics disrupt that resistance is key to the prevention and cure of intestinal infections . We present a novel method to infer microbial community ecology directly from time-resolved metagenomics . This method extends generalized Lotka–Volterra dynamics to account for external perturbations . Data from recent experiments on antibiotic-mediated Clostridium difficile infection is analyzed to quantify microbial interactions , commensal-pathogen interactions , and the effect of the antibiotic on the community . Stability analysis reveals that the microbiota is intrinsically stable , explaining how antibiotic perturbations and C . difficile inoculation can produce catastrophic shifts that persist even after removal of the perturbations . Importantly , the analysis suggests a subnetwork of bacterial groups implicated in protection against C . difficile . Due to its generality , our method can be applied to any high-resolution ecological time-series data to infer community structure and response to external stimuli .
The intestinal microbiota has been receiving much attention lately . Recent studies , propelled by metagenomics and next-generation DNA sequencing technologies , establish novel connections between the intestinal microbial species composition and diseases [1]–[3] . An imbalance in bacterial composition has been linked to chronic diseases such as obesity [4] , Crohn's disease [5] and type 2 diabetes [6] . Even drug-induced transient changes in the microbial community can increase the risk of developing diseases such as acute intestinal infections [7] , or pulmonary viral infections [8] in mammalian hosts . Although its importance has long been acknowledged [9]–[12] studies of the microbiota had been limited by the fact that most microbes are uncultivable in the lab . The recent developments in metagenomic high-throughput sequencing allow this by enabling the investigation of species composition directly without the need for culturing [13] . This has opened a new window into the microbial ecosystem residing in the intestinal tract . Our present view is that the intestinal microbiota is a relatively resilient ecosystem [14] , with a composition that is quite stable over time [15] , [16] . External perturbations , such as dramatic changes in diet [17] or antibiotic administration [18] , can shift the composition . For example , broad-spectrum antibiotics can remove highly abundant species and allow less abundant , antibiotic-tolerant bacteria to dominate [7] . Antibiotic-induced losses of biodiversity increase the risk of bacterial infections [19] , [20] and the effects can persist for several days after antibiotic treatment [18] , [19] , [21] . Perturbation-induced composition shifts are often observed in multispecies microbial ecosystems and may affect macroscopic overall functionality [22] . The loss of protective species can be resolved by reintroducing normally resident ( or commensal ) microbes . Faecal transplantation , i . e . the reestablishment of a patient's intestinal microbiota by introducing the microbiota of a healthy donor , is highly effective against Clostridium difficile induced colitis [23] , [24] . Similarly , the reintroduction of anaerobic flora with high levels of Barnesiella sp . can clear intestines from highly abundant vancomycin-resistant Enteroccocus in mice [25] . In order to understand how commensal consortia confer resistance against pathogens it is crucial to identify the network of interactions between the species [26] . Interactions can be mediated by a direct secretion of substances such as bacteriocins [27] , or ecological competition between the microbes [28] , or even indirect interactions through immune system modulation [29] . Most quantitative studies of the intestinal microbiota so far focused on comparing the composition of different samples using quantitative indices and correspondence analyses [14] and cross-sectional statistical tests [1] , [30] . Likewise , associations between microbial species are often obtained using correlation-based algorithms [26] , [31]–[36] , which results in undirected interaction networks . Singular value decomposition [28] or mixture model engines [37] allow for individuating stereotypical modes of response to external perturbations ( i . e . grouping species positively or negatively affected by the stimulus ) but they provide no information on the interactions themselves ( Figure 1A ) . We recently introduced an ecological model of microbiota dynamics that considers both species interaction networks and extrinsic perturbations such as antibiotics [28] . The model can explain how relatively simple ecological interactions such as competition for nutrients can lead to complex phenomena as , for example , multi-stability or antibiotic-mediated catastrophic shifts . Importantly , we concluded that quantitative knowledge of the microbial interactions could enable the prediction of microbiota dynamics . Predictive models can be of great therapeutic value by guiding antibiotic selection to reduce the risk of antibiotic-induced enteric disease [20] . However , no study to date has generated predictive models of ecological interactions and antibiotic perturbations . Inspired by work on interaction inference in cheese-associated microbial communities [38] we extend the generalized Lotka–Volterra equations [39] , [40] to infer microbiota ecology and predict its temporal dynamics under time-dependent external perturbations . A related approach based on linear ordinary differential equations has already been applied to gene-interaction networks [41]–[44] . Specifically , our method enables the quantification of ( 1 ) growth rates of microbial species , ( 2 ) species–species interactions , and ( 3 ) susceptibilities of microbial groups to time-variable external perturbations such as antibiotics . Moreover , we can use these parameters to numerically predict dynamics of the microbiota and to characterize its stability ( Figure 1B ) . Using this method , we analyze data from a recent mouse study [19] , which shows that the antibiotic clindamycin increases susceptibility to Clostridium difficile colonization . Our results suggest the existence of resilience and multistability in the intestinal microbiota and lead to a hypothesis on a subnetwork of microbial groups involved in the native resistance against pathogen colonization . This study demonstrates that data-derived models of microbiota dynamics can have significant analytic and predictive power . As such , inference and prediction algorithms could be used in combination with metagenomics to assist in the rational design of therapies such as antibiotic or probiotic therapies [12] .
Extracting model parameters using a time-discrete Lotka–Volterra system has already been presented in the context microbial communities [38] , [45] , [46] . We extend this approach by introducing time-variable perturbations and applying Tikhonov regularization to solve the discretized Lotka–Volterra equations . Furthermore , we use the obtained parameters to predict dynamics and assess the system's stability . In this spirit , we adopt the general deterministic approach of modeling time-dependent ecological dynamics using generalized Lotka–Volterra equations [39] with the addition of external perturbations . Formally , this model consists of autonomous , non-linear , coupled first-order ordinary differential equations , ( 1 ) Here is the concentration of a focal species , , at time , is its specific growth rate , is the effect of the interaction of species on species and is the susceptibility to the time-dependent perturbation ( for instance , an antibiotic or diet ) . Ecological time-series data , such as longitudinal metagenomic sequencing data [15] , [47] , provide the composition of a community at discrete time points . Temporally resolved metadata , such as the timing of antibiotic administration [20] or of changes in diet regimes [17] , may also be available and provide information about processes that perturb the microbiota . In order to translate the time-discrete data to a time-continuous dynamical system we divide ( 1 ) by and discretize ( see Materials and Methods ) , ( 2 ) The model parameters are determined by a linear system of equations , which is then solved using Tikhonov regularization [48] in order to ensure uniqueness and stability of the solution , ( 3 ) The values for the regularization parameters , , can for example be found in -fold cross-validation ( we use ) as the minimizer of the mean-squared stepwise prediction error to set the optimal trade-off between data fit and robustness towards the introduction of unseen or missing data [49] . The inference method was first tested on in silico data by generating trajectories for a Lotka–Volterra model as defined in ( 1 ) . We created multiple trajectories of ecological systems characterized by different population sizes , random growth rates , interaction values and susceptibility parameters while ensuring system stability [50] , 51 . The simulations were also subjected to random perturbations of variable duration and white noise was added to simulate measurement uncertainty ( Figure S1 ) . The test confirms that the minimum of the stepwise prediction error can be used as a suitable proxy for the minimization of the parameter inference errors ( Figure S2 ) . Given the inferred parameters we can now predict the temporal dynamics by solving ( 1 ) . We applied this approach to in silico data . The results are presented in Figure S3 . In a recent study , Buffie et al . described experiments on the effect of the antibiotic clindamycin on the intestinal colonization with the spore-forming pathogen C . difficile [19] . The experiments were performed in a mouse model and high-throughput DNA sequencing was used to measure the relative abundance of bacterial species in cecum and ileum . The experiment consisted of three distinct populations of mice . The first population received spores of C . difficile , and was used to determine the susceptibility of the native microbiota to invasion by the pathogen . The second population received a single dose of clindamycin to assess the effect of the antibiotic alone . Finally , the third population received a single dose of clindamycin and , on the following day , was inoculated with C . difficile spores . The untreated mice challenged with C . difficile ( population #1 ) did not develop infection and maintained a stable microbiota throughout the entire experiment . The single dose of antibiotic ( population #2 ) resulted in a dramatic reduction in the microbiota biodiversity , with more than 90% of the initial species dropping below detection . The effects of this perturbation were long lasting , and biodiversity did not return to pre-treatment levels even 28 days after the clindamycin dose . Finally , mice that received C . difficile following the treatment with clindamycin ( population #3 ) were colonized by the pathogen , with 40% of those mice dying due to C . difficile induced colitis . The experiment was performed in three replicates: for each population three mouse colonies were uniformly treated , but separately housed . Each time point represents a mouse from each colony which was sacrificed to determine the intestinal microbiota composition . Mice from the same colony are biological replicates which justifies the interpretation of these compositions as one time line representing one co-housed mouse population [19] . We used the cecal content data to infer microbial interactions , growth rates and susceptibilities to clindamycin ( see Materials and Methods ) . Our mechanistically-based model presupposes absolute abundances . Therefore , we converted the normalized DNA sequence abundances obtained by metagenomics by multiplying with the number of universal 16S rRNA per gram of cecal content ( measured using qPCR ) multiplied by the sample density , [52] ( the actual density value has little importance for the inference of the interactions given the model scaling invariance , see Materials and Methods ) . For consistency with the previous study [19] we integrated only the ten most abundant genera including the pathogen C . difficile , together accounting for the vast majority ( approx . 90% ) of the total sequences obtained from 16S rRNA high-throughput DNA sequencing ( Figure S4 ) . The remaining lower abundance microbes were grouped into a category called “Other” ( see Materials and Methods ) . This choice resulted in less than 30% of undetected entries in the data matrix . The choice of a higher number of independently treated genera , e . g . 15 , could result in more than 50% of missing values in the data matrix ( Figure S5 ) . Consistent with the underlying biological assumptions , the specific growth rates obtained from our inference method ( Figure 2A ) are all positive , and concordant with values measured in vitro using representative species of human colonic microbiota ( 0 . 55–1 . 78 per day [53] compared to 0 . 2–0 . 9 from Figure 2A ) . The diagonal elements of the obtained interaction matrix ( Figure 2B ) are negative . This is again consistent with the underlying biology , since it means that each of these species would eventually reach carrying capacity even in the absence of other species . Coprobacillus is found to be the genus with strongest interactions by value in the ecological network . Specifically , it appears to primarily inhibit all the other microbes , including C . difficile , with the exception of Akkermansia and Blautia which also show inhibitory effect on C . difficile . The strongest inhibitory effect is on Enterococcus which together with the group of unclassified Mollicutes is inferred to positively interact with the pathogen C . difficile . This positive association is consistent with previous reports [54] , [55] . Intriguingly , our method also suggests Barnesiella to mildly inhibit Enterococcus , which agrees with previous findings in mice and humans [25] . Susceptibilities to clindamycin ( Figure 2C ) propose that the antibiotic inhibits all of the genera , except for Enterococcus and the group of undefined Enterobacteriaceae . C . difficile itself is mildly repressed by the antibiotic . Next , we investigated the implications of the inferred model parameters for microbiota dynamics . First , we tested the model's performance in predicting microbiota trajectories . To do so , we inferred the growth , interaction and susceptibility parameters on of the available data , leaving of the trajectories to test the model prediction . Subsequently , we solved eq . ( 1 ) numerically using the inferred parameters , initial compositions and the metadata of antibiotic and/or C . difficile inoculation ( see Materials and Methods for further details ) . In Figure 3 , we compare the observed dynamics of the second replicates with the dynamics inferred from the first and third replicate . Figure S6 shows the full comparison for all the three replicates . The simulated trajectories show a good agreement with the experimental data for all the three populations with respect to order of magnitude and qualitative behavior . There are , however , discrepancies especially in Figure 3B . Here , the experimental data shows a community take-over of Akkermansia and Blautia three days after clindamycin treatment . Our method predicts the same behavior but with several days delay ( see Discussion for possible explications and model limitations ) . The rank correlation between data and prediction is of 62% along time ( Figure 3D ) . We then investigated the long-term stability of the system . We calculated the steady-state composition of the microbiota , , as a solution of eq . ( 1 ) for vanishing the time-derivatives in the absence of any perturbation . Consequently , there are steady states where is the number of microbial groups in the system . Of these , one state corresponds to the trivial case of total extinction ( ) , one state corresponds to the case of total coexistence ( , for invertible ) , and states correspond to the permutations of existence or extinction for every other species [56] . A priori , we have no knowledge about which one of these states the system will attain . This depends on the initial composition , presence and duration of the external perturbations . Therefore , we determine the steady state by simulating long-term dynamics to obtain information on species extinction and coexistence . Once this information is obtained , we can analytically evaluate the steady state of the system and its qualitative behavior by determining the spectrum of the corresponding Jacobian matrix evaluated in that state ( see Materials and Methods ) . The principle of linearized stability states that if the real part of the largest eigenvalue of the Jacobian is negative then the composition represents a stable microbiota ( an asymptotically stable state ) . Otherwise , it is unstable [57] . For instance , the total extinction state , , is unstable if any of the growth rates is positive , which is true for our data ( Figure 2A ) . However , the dynamics of high-dimensional Lotka–Volterra systems allow for a large variety of different qualitative behaviors such as limit cycles , chaos or attractors [39] . We applied this analysis to our system and identified one unique steady state for each independent replicate ( Figure 4A ) . The replicate corresponding to untreated mice challenged with C . difficile ( population #1 ) is characterized by high abundance of clindamycin-sensitive bacteria such as Barnesiella , undefined Lachnospiraceae and unclassified Lachnospiraceae . The steady state corresponding to clindamycin application ( population #2 ) is characterized by a take-over by Blautia , unclassified Enterobacteriaceae and unclassified Mollicutes . Finally , for the case corresponding to C . difficile after clindamycin ( population #3 ) , the steady state predicts severe C . difficile colonization in addition to the genera emerging in population #2 . Interestingly , these steady states agree in order of magnitude , community profiles and composition with the last experimentally measured data point of Figure 3A–C . However , in the observed trajectories the composition still changes between the last two observed data points . This could be due to the fact that the microbiota is not yet stabilized ( i . e . still in transient dynamics ) or due to the effect of fluctuations [15] . Although this cannot be discerned from a simple observation of the data , assuming that our model captures the actual microbiota ecology our analysis suggests that the microbiota of the perturbed microbial communities did not recover their original composition within 28 days from treatment cessation . Rather , the microbiota stays in distinct , perturbation-history dependent equilibria . The intact microbiota is , by antibiotic administration , driven towards a composition which is more susceptible to C . difficile colonization . By subsequent introduction of the pathogen , the community is dragged into an alternative stable composition including the otherwise repelled C . difficile; this may be an example of “niche opportunity” [58] , [59] . Interestingly , when considering the landscape of all possible steady states of the inferred Lotka–Volterra model , unstable steady states , i . e . those referring to critical compositions which drive communities with similar compositions to a collapse or catastrophic shift [60] , are significantly more often observed than stable ones . Given the inferred parameters , we find that of the steady states which the system is able to attain from a composition of L initially present genera , about 98% are found to be unstable ( Figure 4B ) . Nonetheless , our model predicts the existence of multiple stable compositions in each of the three experimental arms . Our results , therefore , may indicate the existence of alternative stable compositions of the intestinal microbiota; switches between these states are induced by perturbation with clindamycin or C . difficile inoculation . This concept is reminiscent of ecological stability and resilience discussed by Connell and Sousa [61] . The inspection of the model inferred from mouse experiments [19] could suggest a possible ecological mechanism for C . difficile colonization ( Figure 5A ) . In the intact microbiota , our method proposes that Coprobacillus interacts positively with the genera of Akkermansia and Blautia . Additionally , Coprobacillus inhibits Enterococcus , which , when increasing in abundance , enhances C . difficile establishment . Without clindamycin , the three genera Coprobacillus , Akkermansia and Blautia , maintain intestinal stability and confer resistance against C . difficile colonization ( Figure 5B ) . However , when clindamycin is administered , Coprobacillus , Akkermansia and Blautia , are inhibited while Enterococcus is promoted . As the three protective groups decrease in abundance , our results suggest that Enterococcus increases in abundance and may facilitate colonization by C . difficile . We discuss the validity of this mechanism in the Discussion section .
We presented a general method for the inference and prediction of multispecies ecological community dynamics under perturbations . Although this method was primarily developed having in mind the intestinal microbiota , the same method may be potentially applied to time-resolved data from any ecological systems , such as bioreactors [62] , marine [63] or soil ecosystems [64] . Our method quantifies growth rates , community interactions and susceptibilities to external perturbations in a single inference . The modeling approach is based on the generalized Lotka–Volterra model ( eq . ( 1 ) ) , a system of non-linear ordinary differential equations , whose governing parameters can be stably determined by a regularized regression on the discretized version of the model ( eq . ( 2 ) ) . Microbiota metagenomics data often have a high number of microbial species which is much larger than the number of available time points . This presents a challenge to inference . We solved this problem in two steps . The first step was to group the bacterial sequences at the genus level of phylogenic classification and consider only the ten most abundant microbial genera including the pathogen C . difficile and merge all remainders to “Others” . The second step was to apply Tikhonov regularization , a procedure that provides a unique and stable solution and , in combination with cross-validation , reduces the risk of overfitting noisy data . Our inference method was tested using in silico data ( Figure S1 ) and evaluated by its ability to recover left-out data using a cross-validation approach ( Figures S2 , S3 ) . The application of inference methods to temporal metagenomic data shows great promise . Still , the development of accurate , predictive models , for example for clinical application , will require further developments and the next few years are sure to see major improvements in this area . For example , the method used here to group microbial sequences may be expanded by adding functional information in addition to taxonomic information . Future methods will benefit from deeper sequencing of the metagenome [65] to inform new ways to define functional microbial groups . Such analyses can shed new light , for example , on the mechanisms by which the abundance of certain species seem to correlate with susceptibility to colonization by closely related pathogenic bacteria [66] . Regarding antibiotic effects , even though we are not yet able to measure the effective concentrations of the antibiotic in the intestine in a high-throughput manner , more accurate information on the pharmacokinetics in vivo will greatly enhance the applicability of this method to clinical settings . Likewise , experimental advancements with animal models will also be crucial . The experiments analyzed here consisted of a single dose of clindamycin of by intraperitoneal injection [19] . Comparing antibiotic perturbed mice with intact mice in this case is similar to comparing a thriving forest with one that has burnt to the ground . The same antibiotic administered in gradual dosages , or the use of other antibiotics , will surely produce distinct effects and would allow for analyzing the communities with distinct compositions . Also , engineered artificial microbiota with defined numbers of bacteria in germ-free mice could be a valuable tool to test the resilience of communities with increasing complexity . Longitudinal data collected from such experimental models can give valuable new insight into the mechanisms of protection against C . difficile . Other differences between data and simulation results may stem from approximating the infinitesimal by time-discrete dynamics and the fact that the Lotka–Volterra model incorporates only pairwise , second-order interactions ( eq . ( 1 ) ) . This could be relaxed in the future by extending the model to third or higher-order interactions once more data becomes available . Furthermore , due to the requirements of the Lotka–Volterra framework our method cannot be applied directly to read count data without additional information on the total number of bacteria per volume unit . If this information is not available it needs to be estimated which can be a source of deviations between measured and predicted results . Nevertheless and even though we cannot claim that the inferred interactions are revealing real causative relationships among microbes , we believe that our results go beyond the explanatory power of widely-used correlations and other methods used . A major advantage of this method is its foundation on a mechanistic framework . This allows for the determination of directional interactions as well as the simulation of microbial dynamics with considerable agreement with the actual data . Based on our inference results , we also hypothesized on a mechanism of C . difficile colonization . However , making a substantiated statement on this mechanism would require further analysis across different host systems and under various antibiotic perturbations . Moreover , due to the limited phylogenetic resolution of the 16S rRNA sequencing , our approach would assign the effects of possibly few , interaction-mediating strains to the whole genus . Nevertheless , the analysis presented here suggests possible experiments focusing on the role of Enterococcus , Coprobacillus , Blautia and Akkermansia in mediating C . difficile colonization . This could be investigated , for example , in mice with engineered microbial consortia . Specifically , the microbiota of these mice could be manipulated to lack the genus of Enterococcus or to contain after antibiotic treatment representative strains of genera such as Coprobacillus , Blautia and Akkermansia which are predicted to have protective effect . Non-colonization and clearance of C . difficile in this system after clindamycin application would then support our hypothesized infection mechanism . There is an urgent need to understand how the commensal intestinal microbial community resists invasion by pathogenic species . Mathematical modeling and inference can help shed new light on this problem by disentangling the contribution of each factor at play . The combination of increasingly accurate and affordable sequencing methods with solidly grounded mathematical theory can help advance our understanding of the relationship between the human host and its microbial inhabitants .
A general approach for a deterministic model of time-dependent ecological dynamics is given by the following system of autonomous coupled first-order ordinary differential equations , in which each time course represents the time-variation in abundance , , of an ecological unit in a certain environment , ( 4 ) with unknown parameters , for . A requirement of ecological models for closed systems is that a unit that once goes extinct cannot return . Thus , for unit which is extinct at time , we require and at any time independent of any variation of the remaining , . In the framework of ( 4 ) , this necessitates , for and for such that , if we restrict to only pairwise interactions , we obtain for each unit , ( 5 ) where and for . This system of equations is also known as the Lotka–Volterra model [39] . The denotes the unlimited growth rate of unit in absence of any competition . The interaction term characterizes the effect of unit on . In particular , stands for activation and for repression . ( No interaction is accordingly indicated by ) . In this form , the model , which is governed by the absolute abundances of units and their physical , order-dependent interactions , also captures non-linear dynamics such as Monod-type/Michaelis–Menten kinetics in a first-order approximation . In addition to growth and interactions we introduce the effect of the application of external time-dependent stimuli , , on each ecological unit such that the full model writes , ( 6 ) where represents an external , time-variable stimulus of a perturbation whose relative susceptibility for each unit is represented by . In the framework of metagenomic data , one faces large magnitudes of total numbers of bacteria . A common approach to identify scale-dependencies of the system and to circumvent numerical problems associated with this is to use non-dimensional variables which allow to treat the model relative to changes on typical system scales [67] . For this purpose , we introduce the following representation of the dynamical variables , ( 7 ) where the dimensionless forms are denoted with asterisks and the barred variables denote the typical scales of the variables . For the measurements of the intestinal microbiota used in our analysis , we find typical scales for abundance and time of and . Equation ( 1 ) then reads in dimensionless form as , ( 8 ) We choose the scale for the perturbation signal such that it is scaled to 1 , i . e . . Thus , we obtain the rescaled growth rates , interaction parameters and susceptibilities as , , and and recover the original equation ( 1 ) by dropping the asterisks . Given this choice , the ( rescaled ) parameters of growth and susceptibility are found to be scale-invariant of changes in the typical abundance , in contrast to the interaction parameter . Input variable is one longitudinal data-set in time points with abundances of taxonomic units ( in the following analysis , genera ) , , and time-dependent perturbations represented by their signal . The parameters of interest are the growth , interaction and susceptibility parameters , , and . The operational taxonomic units counts per sample and relative phylogenetic profile as presented in [19] were used as input data for our analysis . As described in the Results section , we considered the ten most abundant genera ( including the pathogen C . difficile ) and a group “Other” containing the remaining lower abundance genera . The particular grouping was used to reduce sparsity in the data matrix and to avoid spurious , presumably noise-driven contributions . The choice of using the genus level for phylogenetic resolution was dictated by the fact that 1 ) it is consistent with the original published paper [19] and 2 ) it represents the most specific phylogenetic level for which we have classification data . In our grouping , we denote a microbial genus “undefined” ( abbreviated with “und . ” ) when the phylogenetic classification was non-ambiguous up to a certain phylogenetic level . In contrast to Buffie et al . [19] in which the data of the three replicates are presented by their average , we use the individual nine time courses from the cecum ( three from each colony ) and concatenate their compositions spanning 86 time points into the data matrices and . In case of non-detection of an otherwise present genus , we assign a uniformly distributed random value between zero and the detection limit of the corresponding sample . Whenever a genus is completely absent from all considered samples in a particular inference , its corresponding row in the data matrix of above is set to zero . The perturbation signal for clindamycin is modeled by a unit pulse of length day centered on the time of antibiotic administration . Subsequently , the inference was performed as described above with , i . e . in every round of cross-validation , six of the nine time courses were used as training and the remaining three as test set . Ten rounds of cross-validation yielded the minimizing regularization parameter . The result for using all the data of nine time courses is presented in Figure 2 . In the next step , we predicted the behavior of known trajectories only using their initial compositions and clindamycin application and/or C . difficile inoculation and compared it to the measured values . We used from above to infer , and on six out of the nine trajectories , two from each population . These parameters were used to solve eq . ( 2 ) numerically for the remaining three trajectories only providing initial compositions and perturbation profiles and/or C . difficile inoculation . Figure 3 shows the predicted trajectories of the second replicate of each of the three populations using parameters inferred on the remaining six . Moreover , the same parameters were used to assess the stability of the three steady states by linear stability analysis ( see above ) . In Figure 4 , we compared these to the final composition of the corresponding measured time courses .
|
Recent advances in DNA sequencing and metagenomics are opening a window into the human microbiome revealing novel associations between certain microbial consortia and disease . However , most of these studies are cross-sectional and lack a mechanistic understanding of this ecosystem's structure and its response to external perturbations , therefore not allowing accurate temporal predictions . In this article , we develop a method to analyze temporal community data accounting also for time-dependent external perturbations . In particular , this method combines the classical Lotka–Volterra model of population dynamics with regression techniques to obtain mechanistically descriptive coefficients which can be further used to construct predictive models of ecosystem dynamics . Using then data from a mouse experiment under antibiotic perturbations , we are able to predict and recover the microbiota temporal dynamics and study the concept of alternative stable states and antibiotic-induced transitions . As a result , our method reveals a group of commensal microbes that potentially protect against infection by the pathogen Clostridium difficile and proposes a possible mechanism how the antibiotic makes the host more susceptible to infection .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2013
|
Ecological Modeling from Time-Series Inference: Insight into Dynamics and Stability of Intestinal Microbiota
|
Studies of neuron-behaviour correlation and causal manipulation have long been used separately to understand the neural basis of perception . Yet these approaches sometimes lead to drastically conflicting conclusions about the functional role of brain areas . Theories that focus only on choice-related neuronal activity cannot reconcile those findings without additional experiments involving large-scale recordings to measure interneuronal correlations . By expanding current theories of neural coding and incorporating results from inactivation experiments , we demonstrate here that it is possible to infer decoding weights of different brain areas at a coarse scale without precise knowledge of the correlation structure . We apply this technique to neural data collected from two different cortical areas in macaque monkeys trained to perform a heading discrimination task . We identify two opposing decoding schemes , each consistent with data depending on the nature of correlated noise . Our theory makes specific testable predictions to distinguish these scenarios experimentally without requiring measurement of the underlying noise correlations .
Although much is known about how single neurons encode information about stimuli , how neurons contribute to reported percepts is less well understood[1] . The latter , called the “decoding problem” , seeks to identify how the brain uses the information contained in neuronal activity . Although some studies have sought to understand principled ways to decode population responses in the presence of correlated noise [2–12] , the rules by which the brain actually integrates information across noisy neurons remain unclear . Neuroscientists have traditionally investigated this question using two distinct approaches: causal or correlational . In causal approaches , experimenters selectively activate or inactivate brain regions of interest , and measure resulting perceptual or behavioural changes . In correlational approaches , experimenters measure correlations between behavioural choices and neuronal activity , typically quantified by ‘choice probability’ ( reviewed in Ref . [13] ) or , more straightforwardly , by ‘choice correlation’ ( CC ) [14 , 15] . If CCs reflect a functional link between neurons and behaviour , one would expect brain areas with greater CCs to contribute more strongly to behaviour . This naïve view is contradicted by recent results that reveal a striking dissociation between the magnitude of CCs and the effects of inactivation across brain systems in rodents[16 , 17] and primates[18 , 19] . In hindsight , this apparent disagreement is not all that surprising because the two techniques , on their own , yield results whose interpretation is fraught with major difficulties . For instance , the CC of a neuron depends not only on its direct influence on behaviour but also on the influence of all the other neurons with which it is correlated . As an extreme example , a neuron that is not decoded at all could be correlated with one that is , and thus exhibit choice-related activity[9] . Recent theoretical results show that it is possible , in principle , to use knowledge of noise correlations to extract decoding weights from CCs[14] . However , directly measuring the correlational structures that matter for decoding may be extremely difficult[20] . This problem is compounded by the fact that behaviourally relevant information may be distributed across neurons in multiple brain areas , so neuronal CCs in one area may depend on activity in other areas . Moreover , in causal approaches , inactivation of one brain area could lead to a dynamic recalibration of decoding weights from other areas . Therefore , changes in behavioural thresholds following inactivation may not be commensurate with the contribution of the area . When analysed in conjunction , however , results from correlational and causal studies may together provide constraints that can be used to precisely determine the relative contributions of the brain areas involved . In this work , we extend recent theories[14 , 15 , 20] and propose a general framework for inferring decoding weights of neurons across multiple brain areas using CCs and changes in behavioural threshold following inactivation . The two quantities together provide a direct estimate of the relative contributions of different areas without needing to precisely measure the correlation structure . This analysis is based on coarse-grained models of decoded neural noise that is correlated across populations . We demonstrate our technique by applying it to data from macaque monkeys trained to perform a heading discrimination task . In this task , there is a known discrepancy[18 , 21–23] between CCs and the effects of inactivating two brain areas: although neurons in the ventral intraparietal ( VIP ) area were found to be substantially better predictors of the animal’s choices than dorsal medial superior temporal ( MSTd ) neurons , performance is impaired by inactivating MSTd but not VIP . We use our framework to extract key properties of the decoder that can account for these counter-intuitive results . To our surprise , we find that , depending on the structure of correlated noise , experimental data are consistent with two opposing schemes that attribute either too much or too little weight to VIP . We use our theory to make specific testable predictions to distinguish these schemes using CCs measured during inactivation , again without measuring the detailed noise correlations .
We consider a linear feedforward network in which the firing rates r = [r1 , … , rN] of the N neurons are tuned to the stimulus s as f ( s ) = 〈r|s〉 , where the angle brackets denote an average over trials conditioned on the stimulus . The responses on a single trial differ from their averages by some noise with variance σk2 for neuron k , and exhibit a covariance Σ = 〈rrT|s〉 − f ( s ) f ( s ) T that we assume is stimulus-independent . These neural responses are combined linearly using weights w to yield a locally unbiased estimate s^ of the stimulus according to s^=wT ( r−f ( s0 ) ) +s0 . Here local means that the stimulus is near a reference s0 , which we will now take to be 0 without loss of generality , and f ( s0 ) is the mean population response to that reference . Unbiased estimation means that the estimate is accurate on average , so that 〈s^|s〉=s . In the experiments we model , the animals indeed are unbiased after training . The performance of a decoder is often characterized by the variance ε of its estimate: ε=〈s^2〉−〈s^〉2=〈 ( wTr ) 2〉− ( wTf ) 2=wTΣw ( 1 ) Other common measures of performance are the discrimination threshold ϑ , sensitivity index , d′ , and Fisher information J . These measures are all closely related . We will often refer to the discrimination threshold ϑ , which is the stimulus difference , Δs , required for reliable binary discrimination between two categories when discrimination is based on an estimator with finite variance . When 'reliable' is 68% correct , then this threshold is just the estimate's standard deviation , ϑ=ε . This definition coincides with the sensitivity index d′=Δμ/σs^=1 , when the mean difference , Δμ , between estimates for the two stimuli is the same size as the standard deviation , σs^ , of those estimates . When the neural response mean f ( s ) is tuned to the stimulus , but other statistics do not provide additional information ( i . e . for responses drawn from the exponential family ) , then the Fisher information , J , is exactly equal to the inverse variance of an unbiased , locally optimal linear estimator: J = 1/ε ( also assuming differentiable tuning curves and non-singular noise covariance ) . Many experiments assess performance using a two-alternative forced-choice experiment ( 2AFC ) . They quantify performance by the discrimination threshold , ϑ , which is the stimulus difference required for reliable binary discrimination ( 68% correct ) ( see Methods ) , and assess neural decoding based on choice probabilities[24] . However , theoretical results about decoding are much simpler when applied to continuous estimation ( which we will consider to be a continuous ‘choice’ ) . Conveniently , local continuous estimation and fine discrimination are closely related . For example , as mentioned above , the discrimination threshold ϑ is equal to the standard deviation of an unbiased local estimator , σs^ , if the output variability is Gaussian . Under the same assumptions , choice correlation has a simple near-affine relation to choice probability ( see Methods , [15] ) . We thus first describe the theory in terms of a local estimation task , and later apply the suitable transformations when we analyze data from binary discrimination tasks . If the brain decodes signals linearly from multiple populations of neurons , its overall estimate s^ can always be expressed as a linear combination of unbiased estimates from each population separately: s^=aTs^ ( 2 ) where s^=[s^1 , … , s^Z] is a vector of separate estimates from each of Z populations , and a is a vector of scaling factors for each estimate to create one overall estimate . We call these ‘scaling factors’ to distinguish them from the weights given to individual neurons . Thus the problem of decoding multiple populations can be viewed as one of scaling and combining estimates from individual populations . Note that this is equivalent to a single linear decoder of all populations together using w = [a1w1 ⋯ aZwZ] . For locally linear decoding , the assumption of no bias implies a normalization constraint on the weights and scaling factors . An unbiased estimate should match the stimulus , on average; and so a change in the estimate should match the change in the stimulus , on average: ∂s〈s^|s〉=∂swT ( f ( s ) −f ( 0 ) ) ≈wTf′ ( s ) =∂ss=1 . Analogously , unbiased scaling factors of individually unbiased estimates s^z satisfy aT∂s〈s^|s〉=aT1=1 , where 1 is a vector of all ones and where each population estimate s^x=wxT ( fx ( s ) −fx ( 0 ) ) obeys the normalization wxTfx′ ( s ) =1 . Using this decomposition into populations , we can dissociate how the weight patterns within each subpopulation ( wx ) and their scaling factors ( ax ) affect the output of the decoder . This mathematical separation is also appealing because it provides a common framework to synthesize results from experiments conducted at two fundamentally different levels of granularity . One class of experiments involves making fine measurements such as the correlation between trial-by-trial fluctuations in the activity rk of an individual neuron k and the animal’s decision ( Fig 1A ) . The second class of experiments studies causation by measuring behavioural effects of inactivating certain candidate brain areas . For perceptual discrimination tasks , this is done by comparing coarse measures such as the animal’s behavioural performance before ( ϑ ) and after ( ϑ−x ) inactivating population x ( Fig 1B ) . We would like to use these experimental measurements to identify the relative behavioural contributions of various brain areas . Therefore we will present a technique to infer neuronal readout weights in multiple brain areas , focusing primarily on how to extract the scaling factors , ax , of the brain areas rather than the fine structures , wx , of their decoding weights . Choice correlation of a neuron k is defined as the correlation coefficient between its response rk and the animal’s estimate of the stimulus s^ , Ck=Corr ( rk , s^|s ) , across repeated trials with the same stimulus s . Substituting the estimate into this correlation , we find: Ck=Cov ( rk , rTw|s ) Var ( rk|s ) Var ( rTw|s ) =〈rkrTw〉−〈rk〉〈rTw〉σk2 ( 〈wTrrTw〉−〈wTr〉〈rTw〉 ) = ( Σw ) kσk2wTΣw ( 3 ) where the noise variance for neuron k is Var ( rks ) =σk2=Σkk . All neurons' choice correlations can then be expressed together in vector form as C=S−1ΣwwTΣw , where S is a diagonal matrix of the standard deviations . These choice correlations follow a particularly simple pattern if readout weights are locally optimal [15] as obtained from linear regression as w ∝ Σ−1f′ . If we substitute these optimal weights into Eq ( 3 ) , the inverse covariance from the weights cancels the covariance driving the choice correlations: Ck , opt= ( ΣΣ−1f′ ) k ( Σ−1f′ ) TΣ ( Σ−1f′ ) σk2=fk′σk1f′TΣ−1f′=ϑϑk ( 4 ) where Ck , opt is the choice correlation of neuron k expected from optimal decoding , ϑk=fk′/σk is the discrimination threshold of neuron k ( or , equivalently , the standard deviation of an unbiased estimator based only on that neuron’s response ) , and ϑ is the behavioural discrimination threshold . If decoding were optimal , then this behavioural threshold will match the standard deviation of a locally optimal unbiased estimator based on the whole population , ϑ = ( f′TΣ−1f′ ) −1/2 . By itself , such a match would be strong evidence for optimal decoding , but testing this would require recording from all relevant neurons in the brain . The relationship in Eq ( 4 ) is thus a far more practical test for optimal decoding . If all neurons from multiple populations satisfy the above equation , this gives us strong evidence that the neuronal weights — and consequently also the relative scaling factors a of different populations — are optimal . As we will see later , the exact values of a can then be directly extracted from the behavioural thresholds following inactivation of those areas . The pattern of choice correlations generated by any generic suboptimal decoder is more complicated , as it depends explicitly on the structure of noise covariance and the readout weights [14] . For a population of N neurons , the noise covariance Σ describes , for a fixed stimulus , the power along N orthogonal modes of variation . Each of these modes could contribute to the overall choice correlation , depending on how strongly that mode is decoded . We express the decoding weights of a suboptimal decoder in terms of the covariance , as w = ( Σ−1g ) /f′TΣ−1g where g could be any vector in RN . The normalization ensures that this decoder is locally unbiased , satisfying wTf′ = 1 . Note that this recovers the optimal expression given by equation ( 4 ) if g is replaced by f′ . We now rewrite g in the basis of the eigenmodes ui of the covariance Σ , using g=∑i=1NuiuiTg . By multiplying and dividing by uiTf′ , we can decompose the choice correlations for a suboptimal decoder into a weighted combination of optimal choice correlations patterns Copti arising from each eigenmode: C=S−1ϑf′TΣ−1g∑i=1NuiuiTg=S−1ϑf′TΣ−1g∑i=1Nui ( uiTf′ ) ( uiTg ) ( uiTf′ ) =∑i=1NβiCopti ( 6 ) where Copti=ϑS−1ui ( uiTf′ ) ( 7 ) Copti is essentially the i'th noise mode ui rescaled by the individual neural sensitivity , and βi=1ϑ2 ( f′TΣ−1g ) ( gTui ) ( f′Tui ) . These multipliers βi reflect the extent of suboptimality . When decoding weights are optimal , then the readout direction ( again in units of the covariance ) is g = f′ , leading to βi = 1 for all i . Thus , for optimal decoding the above equation reduces to Eq ( 4 ) . In principle , elements of βi , and thus properties of the decoding weights , can be estimated by regressing measured choice correlations against individual columns of the matrix of choice correlations Copt predicted by optimal decoding . In practice , it is very difficult to estimate all of the multipliers βi because the components Ck , opti depend on the individual noise modes of Σ ( Eq ( 7 ) ) . Directly measuring Σ is a notoriously challenging task [20] that involves simultaneously recording the activity of a large population of neurons , and is nearly impossible for certain areas due to the geometry of the brain . Even if such recordings could be performed , it would be challenging to get an accurate assessment of the fine structure of the covariance with limited data , since the number of parameters to measure increases with population size faster than the number of measurements . Fortunately , since neuronal choice correlations are measurably large , it follows that one can infer the animal’s decoding weights with reasonable precision by estimating the few leading multipliers that depend only on the most dominant modes of covariance . This is because if the correlated noise modes with small variance were to dominate the decoder , then only a tiny fraction of each neuron’s variations would propagate to the decision , leading to immeasurably small choice correlations[15] ( S1 Fig ) . It is possible to model properties of the leading modes of covariance without large-scale recordings , and we will consider two different noise models: extensive information and limited information . In this section we describe these two noise correlation models coarsely , at the population level , so that we can use the shared fluctuations between populations to reveal the decoder's scaling factors . To attribute scaling factors to each of Z decoded populations , one must consider at least Z modes of the noise covariance , one per population . We will restrict our attention to decoders inhabiting only these leading modes . If there are Z dominant noise modes and they are correlated across populations , then we can approximate Σ with a rank-Z noise covariance matrix composed of both independent and correlated noise between the populations . These coarse-grained representations of population variability reflect the dominant decoded mode in each population . This level of description allows us to focus on how information is combined between populations . If the brain indeed combines activity from different areas suboptimally , then simplifying Eq ( 6 ) in the presence of information-limiting correlations gives choice correlations within each area that are not equal to the optimal choice correlations , but are still proportional to them . C=S−1ΣwwTΣw≈S−1FEFTwwTΣw≈S−1FEaaTEa=EaaTEaaTEa ( S−1F ) =EaaTEaϑϑk=βϑϑk ( 17 ) where βx= ( Ea ) xaTEa . Under conditions of suboptimality , choice correlations in different brain areas x may have different multipliers βx which depend on the scaling of the brain areas and on the covariance between the estimates s^x that can be derived from them . These multipliers βx can be directly identified by regressing measured choice correlations against ϑ/ϑk , the choice correlations predicted for optimal decoding . S4 Text shows that a similar relation holds for the extensive information model when only the leading mode of each population is decoded ( S4 Text – Eqn ( S4 . 1 ) ) . In the previous section , we showed how to reduce the fine structure of choice correlations down to one number for each population , the slope βx of its choice correlation . We will now show how these multipliers can be used , together with the behavioural thresholds ϑ following inactivations of different brain areas , to infer the relative scaling of their weights a . First we describe the main approach in the general setting with multiple populations , and then we specialize to the particular case of two populations and apply it to our data . Previous work has shown how one can combine knowledge of choice correlations and neural noise correlations to estimate the decoding weights of individual neurons[14] . If decoded neural responses in each population are dominated by a single mode , then we can extend this concept to the population level . The population-level analog of a neural response rk is an estimate s^x derived from population x . The analog of choice correlations Ck are the slopes βx that relate observed and optimal choice correlations , and the analog of noise covariance Σij between neurons i and j is the covariance εxy ( Eqs ( 11 ) & ( 14 ) ) between estimates s^x and s^y derived from distinct populations . Unlike neural noise correlations , we cannot directly measure the noise correlations E at the population level . Nonetheless , we can infer those population-level noise correlations indirectly from inactivation experiments , in which behavioral thresholds are measured after altering the decoder scaling afforded to different brain areas by a factor ρxϕ for inactivation experiment number ϕ . In our feedforward linear model , it is mathematically equivalent to reduce the activity by ρxϕ , or to alter a decoder's scaling ax by the same factor . Totally inactivating an area is equivalent to setting its scaling to zero , but here we permit partial inactivation of multiple brain areas . For now , we assume these inactivation factors are controlled by the experimenter , and thus known , although later we will incorporate some uncertainty about these inactivations . Each such experiment provides one constraint on the unknown population properties , according to θϕ2≈aϕ⋅E⋅aϕ|aϕ|l12=1 ( ∑xaxρxϕ ) 2∑xyaxρxϕExyρyϕay ( 18 ) where θϕ is the behavioural threshold during the ϕ’th inactivation experiment , aϕ is the vector of decoder scaling factors for the different populations with components axϕ = axρxϕ , and where the l1-normalization |aϕ|l1=∑xaxρxϕ ensures that the decoder remains unbiased after inactivation ( as observed experimentally[18 , 22] ) . In such experiments one could also measure the slopes βxϕ of the choice correlations for multiple different populations to provide additional measurement constraints βxϕ≈δx⋅E⋅aϕ|aϕ|l1=1∑xaxρxϕ∑yExyρyϕay ( 19 ) Notice that Eqs ( 18 ) and ( 19 ) can be written as multivariate polynomials up to cubic order jointly in the unknowns E and a . Altogether there are Z ( Z+1 ) /2 unknowns for the covariance matrix E , and another Z unknowns for the intact brain's decoder scaling factors a . As long as the number of independent threshold and slope measurements is at least as large as the number of unknowns , then Eq ( 19 ) can be solved numerically ( S2 Fig ) , revealing the correct decoder scaling for multiple populations . Slopes of choice correlations during inactivation experiments provides a larger number of data points from a given set of inactivation experiments than measuring the thresholds alone . We now use the techniques developed so far to infer the relative contributions of two brain areas in macaque monkeys to heading discrimination . Data were collected from monkeys trained to discriminate their direction of self-motion in the horizontal plane ( Fig 2A ) using vestibular ( inertial motion ) and/or visual ( optic flow ) cues ( see Methods; see also refs . [21 , 23] ) . At the end of each trial , the animal reported whether their perceived heading s^ was leftward ( s^<0° ) or rightward ( s^>0° ) relative to straight ahead .
Although both models were suboptimal to some degree , the overwhelming distinction between them is the efficiency they imply for neural computation , where efficiency is the ratio of decoded information to available information . The efficiency of the limited information model is around 80% , independent of population size N . In contrast , the extensive information model encodes information that grows with N , while decoding is restricted to the least informative dimensions of neural responses . These decoders extract only a tiny fraction of the available information , resulting in an efficiency that falls inversely with N . For a modest-sized population of 1000 neurons , the efficiency is already less than 1% . Thus , the conventional model of correlated noise ( with extensive information ) is radically suboptimal , whereas the limited information model extracts an impressive fraction of what is possible , limited largely by noise . It has previously been argued that the key factor that limits behavioural performance in complex tasks is suboptimal processing , not noise[39] . However , in simple tasks involving binary choices , and in areas in which most of the available information can be linearly decoded , it is unclear why the behaviour of highly trained animals should be so severely undermined by suboptimality . Moreover , radical suboptimality of the kind described here for the extensive information model implies tremendous potential for learning , as the neural circuits can continually optimize the computation by tuning the readout to more informative dimensions . This is hard to reconcile with the observation that behavioural thresholds in a variety of perceptual tasks typically saturate within a few weeks of training in both humans and monkeys[29 , 40–42] . In the presence of information-limiting noise , however , learning can only do so much , and performance must saturate at or below the ideal performance . Therefore , we regard the limited information model as a much more likely explanation of our data , for otherwise one would need to posit that cortical computations discard the vast majority of available information . Note that suboptimal cortical computation might still account for information loss in the limited information model , as opposed to neural noise[39] , but this information loss is now much more modest , probably around 20% . A direct way to tell the two models apart would be to measure the structure of noise correlations . Unfortunately , this is not straightforward , because the differences between noise models giving extensive or limited information can be quite subtle[20] . In fact , there can be a whole spectrum of subtly different noise models with different information contents , lying between the two models that we have considered here . Therefore , a more accurate technique to determine the information content ( which , after all , is a major reason why we care about noise correlations ) is simply to record from hundreds of neurons simultaneously , and then decode the stimulus . This will provide a lower bound on the information available in the neural population . One can then compare the resultant population thresholds with the behavioural threshold to determine how suboptimal the decoding needs to be to account for behaviour . Eventually , we expect this strategy will be successful , but it will require advances in recording technology to be viable in the target brain areas . Meanwhile , by examining the key properties of the decoding strategy implied by the two models , we identified distinct predictions that are testable without large-scale simultaneous recordings . Specifically , they involve fairly simple experiments such as graded inactivation of VIP , and measurement of CCs in either VIP or MSTd while the other area is inactivated ( Fig 8 ) . Future experiments will test each of these predictions to provide novel evidence about the information content and decoding strategy used by the brain . Similar efforts to deal with outcomes of correlational and causal studies using a coherent framework are rarely undertaken , despite their significance . To our knowledge , there is only one instance where this has been attempted before[43] . In that work , the authors used a recurrent network model with mutual inhibition between populations[44 , 45] to reconcile choice-related activity and the effect of silencing neurons . Although their study was similar to ours in spirit , their goal was different . They showed that inactivation just before a decision , when activity was highly correlated with the choice , had less impact on the behaviour than inactivation near the stimulus onset . This addresses a temporal , as opposed to a spatial , dissociation between correlation and causation , so a model with recurrent connectivity was essential to explain their findings . In contrast , we wanted to account for the discrepancies between measures of correlation and causation across brain areas . This latter phenomenon is entirely within the realm of standard feedforward network models in which both populations causally contribute , rather than compete to drive behaviour , and differ only in terms of the relative strength of their contributions . Time-varying weights have been shown to better predict animals’ choice in certain tasks[46] , and psychophysical kernels are sometimes skewed towards one end of the trial[47 , 48] , suggesting that decoding could also be suboptimal in time . Consistent with suboptimal integration , choice correlations in our task peak before the end of the trial , even though new evidence is still available ( Fig 7B ) . Such temporal weighting of information would naturally arise from recurrent connectivity , which is beyond the scope of this work . But it can also originate in feedforward networks , possibly through a gating mechanism that blocks the integration of neural responses beyond a certain time . [32] Other studies have considered that choice-related activity might arise from decision feedback[47 , 49 , 50] . Indeed , pure decision feedback to an area would create apparent sensitivity to sensory signals , even in the absence of direct feedforward input to the target neurons[47 , 49 , 50] . In such a case , neural sensitivity to the stimulus would then be precisely equal to the animal’s sensitivity . In the absence of other sources of variability , response fluctuations would be perfectly correlated with fluctuations in the fed-back choice , producing choice correlations of 1 . Of course there would be additional variability in the neural responses , and this would dilute both the choice correlations and neural tuning by equal amounts , giving rise to measured CCs that should match the optimal CCs ( Eq ( 4 ) ) . Even if there are other feedforward sensory components to the neural responses , direct decision feedback will pull the choice correlations toward this optimal prediction . Thus , simple decision feedback cannot account for the pattern of CCs observed in our VIP data , which are two to three times larger than predicted from optimal inference or direct decision feedback ( Fig 3 ) . Conversely , as we demonstrated through supplementary modeling , adding feedback or recurrent connections may not affect the suboptimal readout weights inferred using our scheme , even when those connections modulate responses along the decoded dimensions ( S16 Fig ) . Nevertheless , future expansions of our work should account for more general recurrent connectivity to study how neural circuits simultaneously integrate information across space and time . In particular , recurrent networks also include decision feedback as a special case , and might help test alternative theories on the origins of choice correlations[1 , 47] . Finally , while VIP inactivation did not impair heading discrimination , MSTd inactivation partially impaired the animal’s ability to perform the task . The fact that MSTd inactivation did not completely abolish performance cannot be accounted for by our two-population models unless the inactivation was only partial and/or VIP is read out to some degree . Additionally , we cannot exclude the possibility that VIP is merely correlated with behaviour and that a third brain area besides MSTd contributes some task-relevant information . In fact , both of our models actually predict a somewhat bigger deficit following MSTd inactivation ( Figs 5C and 6A ) than is observed experimentally ( Fig 1B ) . This highlights the importance of ultimately extending coding models to include more than two brain areas . As neuroscience moves towards ‘big data’ , there is a greater need for theoretical frameworks that can help discern simple rules from complex multi-neuronal activity[51] . We believe our work responds to this challenge and , despite its limitations , takes us closer to bridging the brain-behaviour gap for binary-decision tasks .
All surgical and experimental procedures were approved by the Institutional Animal Care and Use Committees at Washington University and Baylor College of Medicine , and were performed in accordance with institutional and National Institutes of Health ( NIH ) guidelines . Behavioural threshold ϑ is proportional to the square root of the decoder variance ( with proportionality of 1 for threshold of 68% correct ) , so ϑ2 = wTΣw . If decoding is confined to the subspace of leading eigenmodes ux of Σ spanned by neurons within each population x , then wx=ux/ ( fx′Tux ) where the constant of proportionality ensures unbiased decoding from that population . In this case , the behavioural threshold can be expressed purely in terms of weight scaling factors and the variance originating from noise within the noise modes as ( S3 Text ) : ϑ2=aTEa=ax2εxx+ay2εyy+2axayεxy ( 22 ) where E = εxy is the covariance matrix of the noise decoded from populations x and y . Thresholds following inactivation can be determined by setting the weight scaling factor for the inactivated areas to zero . In the case of two populations , this yields ϑ−x2=εyy and ϑ−y2=εxx . Six adult rhesus monkeys ( A , B , C , J , S , U , and X ) took part in various aspects of the experiments . Three animals were employed in each of the MSTd ( C , J and S ) and VIP ( X , B and J ) inactivation experiments . Two animals provided the neural data from each brain area ( A and C for MSTd; C and U for VIP ) . All animals were trained to perform a heading discrimination task around psychophysical threshold . In each trial , the subject experienced a real or simulated forward motion with a small leftward or rightward component ( angle s , Fig 1A ) . Subjects were required to maintain fixation within a 2x2˚ electronic window around a head-fixed visual target located at the center of the display screen . At the end of each 2-s trial , the fixation spot disappeared , two choice targets appeared and the subject made a saccade to one of the targets to report his perceived heading relative to straight ahead . Nine logarithmically spaced heading angles were tested ( 0˚ , ±0 . 5˚ , ±1 . 3˚ , ±3 . 5˚ , and ±9˚ for monkeys A and J , 0˚ , ±1˚ , ±2 . 5˚ , ±6 . 4˚ , and ±16˚ for monkeys B , C , S and U ) , including the ambiguous case of straight ahead motion ( s = 0˚ ) . These values were chosen to obtain near-maximal psychophysical performance while allowing neuronal sensitivity to be estimated reliably for most neurons[21 , 23] . Subjects received a juice reward for indicating the correct choice . For trials in which the ambiguous heading was presented , rewards were delivered randomly on half of the trials . The experiment consisted of three randomly-interleaved stimulus conditions ( vestibular , visual , and combined ) . In the vestibular condition , the monkey was translated by a motion platform while fixating a head-fixed target on a blank screen . In the visual condition , the motion platform remained stationary while optic flow simulated the same range of headings . Under the combined condition , both inertial motion and optic flow were provided . Each of the 27 unique stimulus conditions ( 9 heading directions × 3 cue conditions ) was repeated at least 20 times , for a total of 540 discrimination trials per recording session . Identical stimuli and trial structure were employed during both neural recordings and inactivation experiments . Activity of single neurons in areas MSTd and VIP was recorded extracellularly using epoxy-coated tungsten microelectrodes ( impedance of 1–2 MΩ ) . Area MSTd was located using a combination of magnetic resonance imaging ( MRI ) scans , stereotaxic coordinates ( ~15 mm lateral and ~3–6 mm posterior to AP-0 ) , white/gray matter transitions , and physiological response properties . In some penetrations , electrodes were further advanced into the retinotopically organized area MT[23] . Most recordings concentrated on the posterior/medial portions of MSTd , corresponding to more eccentric , lower hemifield receptive fields in the underlying area MT . To localize area VIP , we first identified the medial tip of the intraparietal sulcus and then moved laterally until there was no longer directionally selective visual response in the multiunit activity , as described in detail previously[21] . Behavioural performance was quantified by plotting the proportion of 'rightward' choices as a function of heading ( the azimuth angle of translation relative to straight ahead ) . Psychometric data were fit with a cumulative Gaussian function with mean μ and standard deviation ϑ , and this standard deviation defined the psychophysical threshold , corresponding to 68% correct performance ( d′ = 1 , assuming no bias , i . e . μ = 0 ) . For the analysis of neuronal responses , we used the linear Fisher information J which is simply a measure of the signal-to-noise ratio: signal power divided by noise power . The linear Fisher Information captures all of the Fisher information in responses generated from the exponential family with linear sufficient statistics . Its inverse is exactly equal to the variance of an unbiased , locally optimal linear estimator ( for differentiable tuning curves and nonsingular noise covariance ) . We defined the square root of this variance ( i . e . the standard deviation of the estimator ) to be the neuronal discrimination threshold , which corresponds to 68% accuracy in binary discrimination . This threshold can be obtained directly from the neuron’s tuning curve and noise variance as follows: ϑk=1Jk=σkfk′ ( 23 ) where ϑk and Jk are the threshold and linear Fisher information[52] for neuron k , fk′ is the derivative of the neuron’s tuning curve at the reference stimulus ( 0˚ ) , and σk2 is the variance of the neuronal response for that stimulus . Neuronal thresholds computed using the above definition were very similar to those computed using a traditional approach based on neurometric functions constructed from the responses of the recorded neuron and a presumed 'antineuron' with opposite tuning[53] ( S4 Fig ) . To quantify the relationship between neural responses and the monkey’s perceptual decisions , we first computed choice probabilities ( CP ) using ROC analysis[54] . For each heading , neural responses were sorted into two groups based on the choice that the animal made at the end of each trial . In previous studies , the two choice groups were typically related to the preferred and non-preferred stimuli for a given neuron[21 , 23] . In this study , in order to appropriately compare different neurons in a population code , the two choice groups were simply rightward and leftward choices; hence , CPs may be greater than or less than 1/2 . ROC values were calculated from these response distributions , yielding a CP for each heading , as long as the monkey made at least 3 choices in favor of each direction . To combine across different headings , we computed a grand CP for each neuron by balanced z-scoring of responses in different conditions , which combines z-scored response distributions in an unbiased manner across conditions , and then performed ROC analysis on that combined distribution[55] . The CPs were then converted to choice correlations according to Ck≈π2 ( CPk−12 ) ( refs . [14 , 15] ) where CPk and Ck are the choice probability and choice correlation of neuron k respectively ( S1 Text ) . Due to the convention we chose for computing CPs , the resulting choice correlation could be positive or negative depending whether a neuron predicted rightward choices by increasing or decreasing its response relative to reference stimulus . For an optimal decoder , the sign of a neuron’s choice correlation should match the sign of the derivative of its tuning curve , so we modified the definition of ref . [15] ( Eq ( 4 ) ) to accommodate our sign convention , yielding Ck , opt=sgn ( fk′ ) ϑ/ϑk where sgn denotes the signum function . There were neurons in both MSTd and VIP whose choice-related activity during the visual condition is anticorrelated with their signal-related activity[21 , 23] . Further analysis showed that heading preferences of these neurons during visual and vestibular conditions differed . Therefore the analysis of data collected during the visual condition presented in the supporting material included only the subset of recorded neurons that had similar heading preferences as in the vestibular condition[23] ( MSTd: 66/129 neurons; VIP: 63/88 neurons ) . Pairwise neuronal recordings carried out separately in areas VIP and MSTd were used to estimate noise correlations between pairs of neurons , Rij = Corr ( ri , rj|s = 0 ) , where ri and rj are the responses of neurons i and j , and correlation coefficients were computed by averaging over trials with headings near 0° . The same recordings were used to compute signal correlations , Rijsig=Corr ( fi , fj ) , where fi and fj are the tuning curves of neurons i and j , and the correlation coefficients were computed by averaging over a uniform distribution of headings in the horizontal plane . The typical noise correlations , R¯ , were then modeled as linearly proportional to the signal correlations ( Eq ( 8 ) ) . The slope of the relation was much steeper in VIP than MSTd[21] . For the vestibular condition , slopes were found to be mM = 0 . 19±0 . 08 and mV = 0 . 70±0 . 16 within MSTd and VIP respectively , and for the visual condition they were mM = 0 . 12±0 . 09 and mV = 0 . 50±0 . 14 . The above fits determined the average relationship between noise and signal correlations , but there was considerable diversity around this trend . To emulate this diversity , we used a technique similar to the one proposed in ref . [31] . Specifically , we sampled correlation coefficient matrices R from a Wishart distribution with a mean matrix R¯ given by Eq ( 8 ) and the fitted slope m , and rescaled them to ensure Rii = 1 . The number of degrees of freedom for the Wishart distribution was adjusted so sampled matrices had the same uncertainty in slope m as the data when subjected to the same fitting procedure . Covariance matrices were generated by scaling the correlation coefficients by the standard deviations for each neuron . Model variances were set equal to the mean responses , so the standard deviation of neuron i is fi1/2 . Thus the covariance Σ is related to correlation coefficients R by Σij=Rijfifj . Correlations between responses of MSTd and VIP neurons were not measured experimentally , so the slope mMV of any linear trend relating noise and signal correlations between the two areas was not known . We explored different possibilities by varying mMV according to: mMV=kmMmV ( 24 ) where k ∈ [0 , 1 ) . Each value of k produced correlation between areas with magnitude εMV which was expressed as εMV = γεMM . If the information reaching MSTd ( M ) and VIP ( V ) is not perfectly redundant across the populations , then the resulting covariance matrix will be of the form given by Eq ( 13 ) where M and V take the places of x and y . The resultant covariances εMM , εVV , and εMV are difficult to determine even with large-scale recordings since their magnitudes may be very small compared to the magnitude of noise in Σ . Nevertheless , we know that for large populations , the behavioural threshold will be dominated by the magnitude of information-limiting correlations . Specifically , they are related through the relative scaling of decoding weights in Eq ( 22 ) . Consequently , we can determine εMM and εVV from behavioural thresholds following inactivation using εMM=ϑ−V2 and εVV=ϑ−M2 . We can then use Eq ( 22 ) in conjunction with Eq ( 21 ) to determine both the ratio aM/aV of scaling factors and the magnitude of correlation between populations εMV = γεMM . Complete inactivation of one of the areas will affect neuronal choice correlations in the non-inactivated area . If Cx and C˜x denote the choice correlations of neurons in area x before and after inactivation of y , then it can be shown that C˜x=ζxCx and similarly C˜y=ζyCy where scalars ζy and ζy are ( S9 Text ) : ζx=1βxϑ−yϑ;ζy=1βyϑ−xϑ ( 25 ) where βx and βy are the multipliers that relate the observed and optimal patterns of neuronal choice correlations in areas x and y . The above equation implies that choice correlations in the active area will increase by a factor proportional to the behavioural effect of inactivating the other area . Intuitively , this is because inactivating an area that was very important for behaviour will dramatically increase the burden on the active area , leading to an increase in the magnitude of choice-related activity .
|
The neocortex is structurally organized into distinct brain areas . The role of specific brain areas in sensory perception is typically studied using two kinds of laboratory experiments: those that measure correlations between neural activity and reported percepts , and those that inactivate a brain region and measure the resulting changes in percepts . The two types of experiments have generally been interpreted in isolation , in part because no theory has been able combine their outcomes . Here , we describe a mathematical framework that synthesizes both kinds of results , giving us a new way to assess how different brain areas contribute to perception . When we apply our framework to experiments on behaving monkeys , we discover two models that can explain the perplexing finding that one brain area can predict an animal’s reported percepts , even though the percepts are not affected when that brain area is inactivated . The two models ascribe dramatically different efficiencies to brain computation . We show that these two models could be distinguished by a proposed experiment that measures correlations while inactivating different brain areas .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"decision",
"making",
"engineering",
"and",
"technology",
"electronics",
"social",
"sciences",
"random",
"variables",
"vertebrates",
"neuroscience",
"covariance",
"animals",
"mammals",
"primates",
"cognitive",
"psychology",
"mathematics",
"cognition",
"vision",
"neuronal",
"tuning",
"monkeys",
"animal",
"cells",
"signal",
"decoders",
"probability",
"theory",
"cellular",
"neuroscience",
"psychology",
"eukaryota",
"cell",
"biology",
"neurons",
"psychophysics",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"physical",
"sciences",
"sensory",
"perception",
"cognitive",
"science",
"amniotes",
"organisms"
] |
2018
|
Inferring decoding strategies for multiple correlated neural populations
|
The frontal cortex controls behavioral adaptation in environments governed by complex rules . Many studies have established the relevance of firing rate modulation after informative events signaling whether and how to update the behavioral policy . However , whether the spatiotemporal features of these neuronal activities contribute to encoding imminent behavioral updates remains unclear . We investigated this issue in the dorsal anterior cingulate cortex ( dACC ) of monkeys while they adapted their behavior based on their memory of feedback from past choices . We analyzed spike trains of both single units and pairs of simultaneously recorded neurons using an algorithm that emulates different biologically plausible decoding circuits . This method permits the assessment of the performance of both spike-count and spike-timing sensitive decoders . In response to the feedback , single neurons emitted stereotypical spike trains whose temporal structure identified informative events with higher accuracy than mere spike count . The optimal decoding time scale was in the range of 70–200 ms , which is significantly shorter than the memory time scale required by the behavioral task . Importantly , the temporal spiking patterns of single units were predictive of the monkeys’ behavioral response time . Furthermore , some features of these spiking patterns often varied between jointly recorded neurons . All together , our results suggest that dACC drives behavioral adaptation through complex spatiotemporal spike coding . They also indicate that downstream networks , which decode dACC feedback signals , are unlikely to act as mere neural integrators .
Behavioral adaptation is the process by which animals extract the rules of their environment and learn to respond to cues to increase their chances of survival . The frontal areas of the brain , including the dorsal anterior cingulate cortex ( dACC ) , are involved in driving this process , although the underlying neuronal mechanisms are not well understood [1] . Most studies have focused on the number of spikes discharged by single dACC units after informative events occur . Other potentially informative features of the neural response , such as reproducibility of spike timing across trials , have typically been ignored . The reason for that may be the apparent unreliability of spike timing when observing frontal activity , which is consistent with theoretical studies [2] . Accordingly , most models of cognitive processing [3–5] rely on stepwise firing rate inputs , therefore disregarding the potential impact of the temporal structure of the driving signals . In the specific case of dACC , a recent theory [1] suggests that this area transmits a graded signal: the expected value of engaging cognitive resources to adapt the behavior . This signal has to be remembered from the moment when the current behavioral policy appears to be improper until the moment when a more appropriate strategy can be implemented . Hence , a simple neural integrator [6–9] , which by construction is insensitive to spike timing , would be well suited to decode and memorize this signal . This neural integrator could be implemented by the lateral prefrontal cortex [10] , which is a plausible dACC target during behavioral adaptation [11] . Nevertheless , some brain regions are known to be sensitive to both the timing [12] and the spatial distribution [13] of spikes within their inputs . These features may improve information transfer between neurons through , for instance , coincidence detection [14] . Some studies reported the presence of a temporal structure in frontal activity , including in dACC [15–23] . However , these observations are not sufficient to make conclusions about the relevance of this temporal structure for the downstream network’s dynamics and for the decision about future behavior . Indeed , to the best of our knowledge , there exists no study comparing the reliability and correlation with behavior of spike count and spike timing in individual frontal neurons during a cognitive task . Comparing spike count versus spike timing sensitive decoders is central to the general view of temporal coding [24] . In this framework , temporal coding can be defined as the improvement of information transmission based on sensitivity to spike timing within an encoding time window [24] . In fact , the temporal structure can be present but still not improve decoding , because spike timing and spike count can carry redundant information [25 , 26] . Furthermore , the temporal structure can be informative but still fail to correlate with behavior , suggesting that downstream processes disregard it and rely solely on neural integration [27 , 28] . Here , we address the issue of temporal coding in dACC . We use recordings from monkeys engaged in a trial-and-error learning task [29] , in which performance relied on reward-based decision making and behavioral adaptation ( Fig 1a and Materials and Methods ) . The task consisted of finding by trial and error which one of four targets was rewarded . Each trial led to the touch of a target and feedback: a reward if the touch was correct , nothing otherwise . In each block of trials ( i . e . , a problem ) , monkeys first explored the different targets in successive trials . The first reward indicated discovery of the correct response . This was followed a period in which they could repeatedly touch the correct target in three successive trials to exploit and receive additional rewards . The firing rate of single dACC units was previously shown to increase at feedback time during either exploration , repetition , or when switching between those two states [29] . Hence , dACC neurons may signal whether and/or how behavior should be adapted . In this context , we probe the putative structure and function of a downstream neuronal network decoding dACC feedback-driven signals . To do so , we investigate to what extent the temporal structure of dACC spike trains , during post-feedback firing , could improve information transmission and predict behavior ( Fig 1b ) . Assuming a neural integrator decoding scheme , the downstream network would compute and maintain the memory of the need for behavioral adaptation on the basis of the number of spikes emitted by dACC ( Fig 1c , middle ) . Alternatively , the downstream network could be sensitive to the spatiotemporal structure of dACC activity ( Fig 1c , bottom ) . For instance , temporal coincidences in the afferent dACC signals could favor the switch to , and maintenance of , a high-activity state in the downstream network to encode behavioral adaptation ( [32] , see also [33 , 34] ) . We bring forth evidence for a spatiotemporal decoding of dACC activity . First , we find that there are informative temporal patterns in single units leading to a decoding more efficient than with spike integration . The optimal decoding time scale is in the range of 70–200 ms . Second , some spike coincidences across jointly recorded neurons are advantageous for decoding . Furthermore , the data suggest that downstream neurons could benefit from a non-linear spatiotemporal integration of inputs . Finally , we develop a new method to evaluate to what extent dACC temporal patterns can predict the behavior of monkeys comparatively to spike count . Importantly , we find that temporal patterns are significantly and sizably predictive of the upcoming response time of monkeys .
We first tested how single-trial , single-unit dACC activity could send signals that could drive behavioral adaptation after feedback . Behavioral adaptation occurred either after the first reward ( thus switching from exploration to repetition ) or after any error during exploration ( Fig 1a ) . Signaling the need for adaptation requires that spike trains emitted during either first reward or errors can be discriminated from those emitted during repetitions ( referred to as first reward and error discrimination analyses , respectively ) . Neurons in dACC could show early post-feedback responses specific to behavioral adaptation [29] . Therefore , we analyzed spike trains starting at the onset of the feedback delivered 600 ms after target touch . We will refer to any post-feedback time interval ( i . e . , following either an error , first reward , or repetition ) as a “task epoch . ” We quantified to what extent spike trains emitted during different task epochs were discriminable by a downstream decoder by classifying them based on a spike train dissimilarity measure [35] . This dissimilarity measure computed the minimal cost to transform the first spike train into the second one through two possible recursive operations: ( i ) adding or removing a spike , for a cost of 1; and ( ii ) changing the timing of a spike by dt , for a cost of q dt ≤ 2 . Note that the maximum cost allowing two spikes to be temporally matched ( coincidence detection ) is 2 because it corresponds to the cost of removing and adding one spike ( Fig 2a and Materials and Methods ) . This measure allows different temporal sensitivities of a downstream decoder to be evaluated by varying the parameter q . A value of q = 0 s-1 describes a decoder sensitive to pure spike count . On the other hand , a larger q value corresponds to a decoder sensitive to precise spike times . The larger the q value , the smaller the maximum interspike interval leading to coincidence detection , and the more the decoder disregards spike count . We stress that even when the neural activity is temporally structured , sensitivity to spike timing does not necessarily improve decoding . For instance , spike timing and spike count might provide redundant information , and then a neural integrator could be more robust ( S1 Text , Sec . 4 ) . We quantified the classification performance ( i . e . , how well , on average , a spike train was correctly associated to the task epoch with the most similar activity ) by computing the mutual information between the predicted distribution of spike trains across task epochs and the true distribution ( Materials and Methods ) . Throughout this article , mutual information values are expressed as percentage of the maximum value corresponding to perfect discrimination . Information values were computed for different analysis windows , all starting 1 ms after feedback time and with increasing duration . In this way , the state of a putative decoder of dACC feedback-related discharges could be evaluated at different delays after the start of the decoding process . We investigated the nature of dACC firing statistics determining the advantage of temporal decoding . Spike-timing reliability might mainly reflect differences in the temporal variations of firing rates between task epochs . Alternatively , beyond this time-dependent firing rate , temporal correlations between spikes within one trial may impact the spike time reproducibility . Indeed , cellular processes ( such as spike-triggered hyperpolarizing currents ) may affect future spiking probability depending on past spike times [4 , 36] , especially when the synaptic current received by the neuron is not very variable . Similarly , recurrent neural network dynamics—within dACC or upstream—may create correlations in spike times [37 , 38] . Here , we tested whether or not , beyond their existence , spike-timing correlations sizably and consistently ( over neurons ) impacted information transmission . Multiple neurons were often simultaneously recorded ( median = 2 ) . Thus , we also decoded the activity of pairs of neurons ( Monkey M , n = 122 pairs; Monkey P , n = 271 ) while varying both the temporal sensitivity q and the degree of distinction k ( Fig 2b , Materials and Methods ) . For the computation of the dissimilarity measure , the parameter k represents the cost of transforming a spike from neuron 1 into a spike from neuron 2 . Therefore , during classification of spike trains from pairs of units , the dissimilarity between spikes from different neurons increases with k . The parameter k permits testing of whether the informative spikes are neuron specific or if they tend to be emitted synchronously by two neurons . In the former case , the amount of information would increase if the decoder were accounting for neural identity ( k > 0 ) , as compared to a decoder blind to neural identity and sensitive to interferences between neurons ( k = 0 ) . In the latter case , k = 0 could be optimal for decoding because it makes the discharge of either one of the neurons sufficient to have reliable joint spiking . The presence of information in single-unit spike timing does not necessarily imply that the downstream networks do actually use it [27 , 28] . In particular , if dACC spike timing were not used , then different temporal patterns would be rather unlikely to reliably correlate with different behavioral outputs . Here we examined whether first reward single-unit activity could predict upcoming behavior . We focused on the behavioral response time , i . e . , the time between the “go” signal and touch on target ( Materials and Methods ) . The response time was measured during the trial following the first reward . Thus , several seconds separated the analyzed neural activity and the behavioral measure . Interestingly , the response times of both monkeys consistently increased on the touch following the first reward compared to the touch leading to first reward ( S11c Fig ) . This was in agreement with a behavioral switch from exploration to repetition . We separated trials into two groups: one group with response times higher than the median , and the other with response times below the median . The probability of switching to repetition was very high in both groups and statistically equivalent between them ( S2 Table ) . We tested the hypothesis that longer response times may reflect a longer decision-making process , when monkeys might act more carefully to avoid mistakes .
Post-feedback spike counts in dACC neurons were shown to depend on whether behavioral adaptation was required [29] . Given the absence of stimulus-driven temporal fluctuations in input and high noise in spike timing , a plausible hypothesis would be that only spike count is relevant to the transmission of the need to adapt behavior by dACC firing [2] . By contrast , we provide evidence for an efficient spatiotemporal spike coding of behavioral adaptation signals . Our analysis accounts for the temporal sensitivity of a biologically plausible neural decoder that would receive post-feedback dACC discharges . Adjusting the temporal sensitivity of the decoder ( within the interval of one second after the feedback ) can enhance the readout of single-unit spike trains relevant to behavioral adaptation . Beyond the existence of a temporal patterning of dACC activity , these results indicate that spike-timing reliability supplements spike-count reliability . Interestingly , in frontal areas single-unit spike generation mechanisms or network dynamics , rather than external stimuli or motor feedback , are probably responsible for spike timing reliability and spike-count variability [37 , 40 , 43 , 44] . We found strong temporal correlations , stronger-than-Poisson spike count variability , and heterogeneous spike times across the dACC population . This feedback-type-specific dynamics is thus unlikely to arise through neuronal assemblies connected by balanced excitatory and inhibitory inputs with uniform wiring probability and with stationary weak-to-moderate strengths [37 , 40] , as these features tend to create Poisson-like spike trains . Spike-triggered hyperpolarizing currents or short-term plasticity could also plausibly favor the presence of informative temporal correlations in dACC activity and could participate to shaping the lower-than-Poisson spike count variability occurring shortly after the feedback [36 , 39 , 44] . Note that the optimal range of decoding time scale that we found ( τ ≈ 70–200 ms ) is larger than those found when decoding responses to stimuli with relevant temporal patterning or contrast at onset time ( e . g . , auditory stimuli , τ ≈ 5 ms [45]; visual stimuli , τ ≈ 10–100 ms [13 , 35] ) . This is consistent with the idea of a hierarchy of increasing time scales from sensory to higher-order areas [46] . However , there are also exceptions to this rule , for instance in the gustatory modality ( for which the timing of the stimulus is less relevant ) . Indeed , the optimal time scales were found to be close to the one we found in dACC ( τ ≈ 50–500 ms [47] ) . Given that during a gustatory stimulation , the motor behavior of the animals and/or some sensorial input transients were probably participating in shaping the temporal code [47] , it is quite remarkable that we found equivalent time scales in our data for which internal neuronal dynamics was probably the major contributor to spike timing reliability . Also , the optimal spike coincidence timescale we observed loosely matches the period of local field potential ( LFP ) oscillations in the delta and theta range , on which frontal neurons can phase lock during cognitive tasks [16 , 19 , 48] . LFPs partially reflect the synaptic input of the local population [49] , which could both shape and be influenced by the temporal spiking patterns of dACC . The optimal temporal sensitivity range for decoding identified in this study remains an approximation . First , different methods ( S1 and S4 Texts ) , or different analysis windows ( Fig 4 ) , might give slightly different optimal values . Yet , although it is not feasible to extensively test all possible decoders , our analysis accounts for biophysically reasonable assumptions on the downstream decoder . In this framework , we provide strong evidence for the plausibility of decoding through spike coincidences ( up to a few hundred ms ) , compared to a neural integrator decoder . Second , spike trains were referenced to feedback time , but the internal reference of the brain could be different and more or less accurate [50] ( e . g . , coincidence detection during a population onset [24] , or precise spike timing relations in a neuronal population [51] ) . Aligning to feedback times was very relevant for behavioral-adaptation task epochs where monkeys could not predict the outcome and were thus reacting to feedback . However , anticipation of rewards during repetition periods may have promoted internal references dissociated or jittered from actual juice delivery , decreasing the apparent temporal reliability ( as suggested by the data , see S4 Fig ) . The spike-time sensitive decoder can be understood as a downstream network that , through synaptic plasticity [52] , becomes differentially selective to coincident spiking patterns that are specific to task epochs . The optimal temporal sensitivity range is compatible with the time constant of NMDA-mediated currents . Indeed , the efficiency of the spike coincidence mechanism decreased with interspike intervals up to approximately 200 ms , which relates to an exponential decay time-constant of 100 ms . Within the decoding by coincidence framework , decoding relies on the convergence of excitatory neurons that transmit similar temporal patterns to a post-synaptic compartment ( triggering summation of depolarizations ) . Yet , informative neurons with distinct and potentially antagonistic temporal patterns may improve information transfer , for instance , if they were decoded by different specialized post-synaptic neurons . We showed that paired decoding generally enhanced information transmission relative to the pair’s most discriminative unit . This suggests that highly informative activity can be advantageously combined with less informative inputs that do not act as contaminating noise . The information increase was achieved by varying the degree of distinction between the two units ( parameter k ) . This mechanism may be implemented by different spatial organizations of synapses , which could modulate , through non-linear summation , the temporal precision of spike coincidence detection . Other mechanisms , such as different synaptic weights or synaptic timescales ( i . e . , two weak/shorter depolarizations that require more precise coincidence to efficiently sum ) , or targeted inhibition , may also induce a similar effect . In addition , we showed that in a smaller proportion of pairs the activity of both units did not need to be distinguished to achieve optimal discrimination . Thus , if these two units were excitatory , direct summation of their post-synaptic potentials would be advantageous . The partial spatial specificity of reliable spikes may be advantageous during realistic decision-making when quick choices should be made between many strategies . Indeed , the combination of spatial and temporal information can increase the number of possible specific activity patterns compared to simultaneous firing of all neurons . Artifacts of spike sorting were unlikely to affect our conclusions ( S2 Text ) . The removal of ( rare ) coincident spikes during sorting decreased the reliability of both spike count and temporal patterns . Noisy spikes erroneously assigned to a given neuron can make spike count more reliable whenever two mixed single units have the same firing preference ( with respect to task epoch ) . On the other hand , temporal coding may only become more reliable if the two mixed units are temporally coherent . Our results showed that different dACC units did not often share coherent spike times , which suggests that possible erroneous spike assignments would not favor temporal decoding . Furthermore , pairs recorded on the same electrode did not always differ from pairs recorded on different electrodes ( S3 Table ) . Indeed , while no differences occurred when discriminating first reward versus repetition task epochs , differences could exist for error versus repetition decoding . When present , such differences were more likely to reflect a topological organization of some inputs . Indeed , a by-product of a cross-talk between sorting templates would rather have a constant impact across all task epochs . We further probed dACC function by testing how it could affect future behavior . We found a significant correlation between neural activity at feedback time and the monkeys’ response time during the following trial . This finding is functionally different from the correlation previously reported between pre-movement dACC activity ( which often resembles an integration to threshold [53 , 54] , in contrast to feedback-driven dACC responses ) and immediate motor response [53–55] . This motor correlation could become apparent through the comparison between trials with high versus low firing rates ( or , equivalently , spike-counts in a given window ) . In particular , Michelet et al . showed that the quicker the increase of firing rate to threshold , the quicker the movement [53] . This implied high versus low spike-count correlation when aligning spike trains with respect to movement . In contrast , we observed a correlation between dACC activity and behavior in terms of deviation from prototypical activity patterns , while we did not observe a robust link between large versus small number of spikes emitted during first-reward–triggered discharges and different behaviors . This result can be well understood when considering that dACC can signal a given behavioral strategy when its activity lies close to a given prototypical state . Hence , this interpretation can be consistent with a report of increased spike count variability ( and hence , of absence of defined state of activity ) in dACC during periods of behavioral uncertainty [56] . It can also be related to the sudden reorganization of dACC activity in a new “rule-encoding network state” when animals switch to a new rule [57] . Within this framework , first reward feedback triggers specific dACC activity patterns [58] that shape the response of downstream areas such that the appropriate decision ( here , switching to repetition ) is made . Deviation from these “prototypical patterns” would lead to a slower behavioral response . In addition , if the deviation of dACC discharges from their usual pattern were triggered by increased uncertainty or difficulty , slowing the behavioral response may prevent incorrect choices ( as suggested by the similar error rates between trials with fast versus slow responses ) . Interestingly , these results also suggest that the information transmitted to downstream areas cannot be mapped onto a mere intensity value ( i . e . , a single dimension ) , such as the magnitude of the required cognitive control , as in the case of the integrator model . Rather , the deviation from a prototypical pattern , which relates to behavioral modulation , appeared to occur in many different ways ( through either an increase or a decrease of spike count , or through spike timing deviations within the heterogeneous temporal patterns of dACC neurons ) . This hints to the transmission of a high-dimensional representation by dACC , possibly linked to the embedding of the cognitive control signal into a specific context , or behavioral strategy [1 , 29 , 30] . One limitation of our study is that we only characterized the dimensionality of the representation transmitted by dACC through the large differentiation , at the population level , between measures based on firing rate and measures based on ( absolute ) deviation from prototype . A full evaluation of this dimensionality will need future studies to evaluate the space of neuronal variability and its relation to behavioral variability in each single neuron . Importantly , beyond the deviations of dACC spike trains from prototypical spike count , our findings indicate that deviations from prototypical temporal patterns were predictive of the monkeys’ upcoming response time . This was consistent and significant in both monkeys . Furthermore , compared to the prediction based on spike count deviations , the prediction power of adapted temporal sensitivity was either equivalent ( monkey P ) or significantly stronger ( for monkey M , which showed the most reliable relation between neural activity and behavior ) . This strongly suggests that the temporal patterning of single unit activity is not an epiphenomenon irrelevant to downstream network dynamics . Interestingly , dACC differs from other decision-making related areas such as middle temporal ( MT ) or orbitofrontal cortex ( OFC ) regarding the nature of the relation between neuronal variability and future response time variability . Indeed , in MT and OFC , the firing rate of specific neuronal populations predicts behavioral modulation [59 , 60] . In addition , evidence suggests that neurons in MT are decoded through integration , a process that could be reflected in LIP ( lateral intraparietal cortex ) activity [61] and that appears to have one-dimensional dynamics [62] . Altogether , our results appear hard to reconcile with the hypothesis of a decoding of post-feedback dACC activity by a neural integrator . Other types of decoders could be compatible with both an increase in information through spatiotemporal coincidences and a correlation of deviation from prototypical temporal patterns to behavior . For instance , as we illustrate in Fig 1c , a recurrently connected neuronal population , which maintains memory through a high-activity state , can be modulated by the temporal structure of its input [32] . Alternatively , a downstream network maintaining a memory through repetitions of sequential activations of NMDA-connected neurons would also be sensitive to spatiotemporal patterns [63] . Our findings , therefore , call for a better understanding of how models of short-term memory and decision-making could reliably be modulated by a temporal input at the timescale of hundreds of ms .
Two male rhesus monkeys were implanted with a head-restraining device , and neuronal activity was recorded by one to four epoxy-coated tungsten electrodes ( horizontal separation: 150 μm ) placed in guide tubes and independently advanced in the dorsal bank of the rostral region of the cingulate sulcus . Recording sites were confirmed through anatomical MRI and histology [11 , 29] . Extracellular activity was sampled at 13 kHz and unitary discharges were identified using online spike sorting based on template matching ( MSD , AlphaOmega ) . All experimental procedures were in agreement with European , national , and local directives on animal research . For monkey P , all recorded units were used . For monkey M , only units showing a significant response to at least one event ( either error , or first reward , or repetition reward , fixation breaks ) were used ( TEST 1 in [29] ) . The mean and standard deviation of the baseline firing rate ( taken from -600 to -200 ms before feedback onset ) were computed . Units with a change of firing rate of magnitude higher than 5 standard deviation of the baseline within more than six 10 ms bins between +60 and +800 ms of at least one event were selected . Note that this test cannot favor temporal coding in any way . Monkeys had to find , by trial-and-error , the rewarded target among 4 targets presented on a touch screen ( Fig 1a ) . To begin a trial , the animal had to touch a central item ( “lever” ) , which triggered the appearance of a fixation point . After 2 s of gaze fixation , the four targets appeared simultaneously . At fixation point offset , the animal had to select a target by making a saccade toward it , fixate it for 0 . 5 s , and touch it following a “go” signal ( i . e . , all targets bright ) . All targets dimed at the touch , and switched off after 0 . 6 s . Reward ( fruit juice ) was delivered if the correct target was selected , otherwise no reward occurred . Throughout this article , we define a “trial” as the period of time between the touch of the lever and 1 s after the reception of a feedback ( either error , or first reward , or repetition reward ) . In addition , we call “task epoch” the time interval between 1 ms and 1 s after the reception of a given feedback . After feedback , a time break of 2 s was imposed before starting a new trial . Any break in gaze fixation or touch within a trial led to resuming the sequence at the lever touch . Note that we did not consider that this started a new trial . In case of an incorrect choice , the animal could select another target in the following trial , and so on until the discovery of the rewarded target ( i . e . , exploration ) . The correct target remained the same in the following trials , allowing the animal to exploit the rewarded action ( i . e . , repetition ) . We define a “problem” as the block of trials that are associated with one rewarded target location and that terminate with a “signal to change” . This signal to change was a flashing signal indicated the end of repetition and the beginning of a new problem ( the new rewarded target had a 90% probability to be different from the target rewarded in the preceding problem ) . In 90% of problems , the repetition period lasted three trials after the first reward , whereas in 10% of problems , 7–11 repetitions could occur . Repetition trials beyond the third one were excluded from analysis to avoid possible surprise effects . At the time of recordings , the task was well known: monkeys omitted to repeat the correct touch in one of the trials following the discovery of the rewarded target in only around 1% of problems . Then , both the incorrect touch and the following trials were discarded from analysis , but previous trials were kept . As previously reported [41] , monkeys might be able to infer the rewarded target after three non-redundant errors , i . e . , the third error would systematically trigger a switch to repetition . Therefore , only first and second erroneous touches as well as first rewards preceded by less than three errors were included in the analysis . For the repetition period , we selected all correct trials that followed a search with up to three preceding search errors . We assessed to what extent single-trial spike trains encoded enough information to discriminate between the following types of post-feedback task epochs: ( a ) after first reward versus after second , third , and fourth rewards , i . e . , “first reward discrimination;” ( b ) after no reward versus after second , third , and fourth rewards , i . e . , “error discrimination . ” To compare different decoding schemes ( i . e . , spike-count and timing-sensitive decoders ) , the spike-train metrics framework [13 , 35] was considered . Neuronal decoding relied on the assessment of the Victor and Purpura distance dq , k ( s , s’ ) between two spike trains ( s , s’ ) , both within and between task epochs . One spike train s contained the spikes of one ( Fig 2a ) or two ( Fig 2b ) neurons during one task-epoch of a given trial . The distance dq , k ( also named “dissimilarity” throughout the article ) depended on two parameters , namely , the spike timing sensitivity q and , in the case of multiunit activity , the degree of distinction k between different neurons ( Fig 2 ) . Algorithmically , the distance dq , k between two spike trains was computed as the minimal cost to transform the first spike train into the second one , using three possible operations: adding or removing a spike , for a cost ( Dmax/2 ) = 1; changing the timing of a spike by an amount dt , for a cost ( q dt ) ≤ Dmax , where Dmax = 2 is the maximum cost corresponding to removing a spike and re-inserting it at the right time; changing the identity of the neuron which fired the spike , for a cost k . Each spike train s was classified as belonging to the task epoch E that contained the most similar neuronal responses to s . The similarity between s and all discharges produced during E was quantified by computing a global dissimilarity to the ensemble of spike trains in task epoch E . The results presented in the main body were obtained by taking this global dissimilarity as the median of the pairwise distances between s and any other spike train of E ( i . e . , median ( dq ( s , s' ) ) s'∊E , s' ≠ s ) . Comparable results were found by biasing the global dissimilarity measure towards small spike train distances , i . e . , by “ignoring” more the very dissimilar spike trains during classification ( “nearest-neighbor” classification , S5 Fig ) . The accuracy of the classification process was measured as the mutual information between the actual probabilities of dACC spike trains to belong to each task epoch and the probabilities predicted by the clustering method . More precisely , if we define a confusion matrix N , whose entries N ( i , j ) are the number of spike trains belonging to task epoch i and classified in task epoch j , the raw information was 1∑I , JNI , J∑i , jNi , j ln ( Ni , j ∑I , JNI , J ( ∑kNi , k ) ( ∑lNl , j ) ) . This measure accounts for the differences in number of trials between task epochs , it tends toward zero for chance-like predictions ( for large data samples ) , and its maximum ( for perfect prediction ) depends on the relative proportion of trials between task epochs . As the number of trials was neuron dependent , each information value was normalized with respect to its maximum value . Full details on the use of spike-train metrics are in S1 Text . To test whether classification was above chance , trials were randomly permuted between task epochs , and two groups were recreated ( with the same number of trials as the original task epoch clusters ) . The information content associated to the shuffled groups was then computed . The process was repeated 1 , 000 times , leading for each q or [q , k] , to 1 , 000 values of information under the null hypothesis that the discrimination between groups is due to random similarities between any two spike trains . The information analysis was done on increasing time windows , starting 1 ms after the onset of the feedback ( to avoid pump-driven artifacts ) . The first window lasted until 50 ms post-feedback , and was incrementally increased to 600 ms by 50 ms steps , and then up to 1 s by 100 ms steps . The higher resolution for smaller windows allowed the time course of the fast initial transient to be evaluated . We computed the maximum ( over q or [q , k] ) number Nw of consecutive windows for which the information was strictly larger than the 95th percentile of the 1 , 000 sets of permuted data . The same process was repeated for each set of permuted data , relative to the remaining 999 permuted sets . A neuron ( or a pair ) was considered as significant if Nw was strictly larger in the actual data than in 95% of permuted data . This process did not favor a given q or k , and could select neurons/pairs of neurons with different information time courses . Also , it allowed us to exclude neurons with very unreliable activity , which would act as “noise” during the subsequent analyses . The sample information bias was empirically estimated ( for each q or [q , k] and each analysis window ) as the mean information in the 1 , 000 permuted datasets [64] , and was subtracted from the information in the original data ( slightly negative corrected information were clipped at 0 ) . Note that , throughout the paper , the term “information” refers to a bias-corrected and normalized-to-maximum value . After bias correction , for each q or [q , k] and for each significant neuron , the temporal evolution of information values was summarized by taking the mean information over 10 analysis windows of increasing lengths ( ending from 100 ms to 1 s post-feedback onset , by steps of 100 ms , favoring neither early nor late information ) . We refer to this quantity as “time-averaged information” in the article , <I>t . Computing the time-averaged information is equivalent to averaging over delays before a decision is made by the animal . Finally , a non-parametric Friedman ANOVA was used to compare the time-averaged bias-corrected normalized information as a function of different q or [q , k] , with Tukey's honestly significant difference criterion correction for multiple comparisons . Note that our procedure is more powerful than simply counting the number of neurons that would reach a certain significance level separately for each q or [q , k] . Indeed , the latter method cannot be sensitive to differences in classification ability between two parameter sets that reach a certain significance level , and it is restricted to the use of large enough p-values ( which can be evaluated with a reasonable number of permutations ) . We used spike-time shuffling to investigate to what extent random samples from a time-varying trial-averaged rate density ( as in “Poisson” neurons with time-varying rate ) could underlie the advantage of the temporal structure for decoding [35] . For each cell and each task epoch separately , we grouped all spikes emitted in the interval [0 . 001 , 1] s post-feedback and randomly assigned each of them to a trial ( repeated 1 , 000 times , Fig 5a ) . This procedure is equivalent to drawing the number of spikes in each trial from a Binomial distribution with parameters n = Nspikes and p=1Ntrials , which—following the common approximation—is close to a Poisson variable . Indeed , p was rather small ( the trial number was usually big: 25th quantiles were 14 . 25 and 51 . 25 for first reward and repetition respectively ) ; and the total number of spikes n was big ( 25th quantiles were 53 . 75 and 175 for first reward and repetition , respectively ) . Under the Poisson approximation , spike counts restricted to sub-analysis windows are also Poisson ( Raikov's theorem ) . This allowed us to build the spike-shuffled data for smaller analysis windows by simple pruning of the 1 , 000 shuffled data of the largest window . We used a second shuffling procedure to test to what extent information transmission could be determined by time-varying firing rates and spike-count variability as in the original data ( Fig 5d ) . In contrast to the previous shuffling method , this procedure considered that time-varying firing rate was modulated by a multiplicative factor . This factor constrained the spike-count variability to fit the original data and it was specific to each trial and time independent . This shuffling procedure not only conserved the PETH but also the number of spikes present in each trial . To do so , for each cell , each task epoch , each analysis window , and all 1 , 000 shufflings independently , we randomly permuted all available spikes before reassigning to each trial the exact same number of spikes as in original data ( without replacement ) . Because both shuffling methods produced spike-shuffled data with the same number of trials as in the original data , the finite-sample information bias should be similar in both cases and should cancel when looking at the information difference , which was the relevant quantity . The bias was therefore not re-evaluated for this analysis . The response time was defined as the time between the “go” signal ( for the hand movement ) following the first reward , and the subsequent target touch . For this analysis , only neurons with significant first reward classification and with at least five available trials were used ( or subgroups of this ensemble , see S12 Fig ) . For a given neuron , the dissimilarity between a recorded spike train s and the prototypical spike train sprototype emitted at first reward was estimated as the median of all pairwise distances d*q ( s , s’ ) , where s’ corresponds to all the recorded first reward spike trains that are different from s . We defined a new metrics d*q by dividing the Victor and Purpura distance dq by either the number of coincident spike pairs between s and s’ , or 1 if no spikes were coincident . Note that we define two spikes to be coincident when their dissimilarity was smaller than Dmax = 2 ( see Fig 2a ) . For q > 0 , the distance d*q quantified the mean jitter of coincident spikes , plus a cost for unmatched spikes . For q = 0 , d*q estimated the normalized absolute spike-count difference between the two spike trains . Importantly , d*q did not scale with the number of emitted spikes: a trial with too little spikes , as well as a trial with many spikes at inaccurate times , will have a large normalized distance to the prototype . We further discuss the rationale for designing d*q in S4 Text . Let r˜ denote the median value of observed response times , T+ be the set of first reward trials followed by a response time larger than r˜ , and T- the set of trials followed by a response time lower than r˜ . For each spike train s , we calculated the dissimilarity between s and prototypical first reward activity ( i . e . median ( dq* ( s , s′ ) ) s′∈1st reward , s′≠s , similar to the spike train classification analysis ) . We then defined D¯T+ ( D¯T− ) as the mean over all s∊T+ ( T- ) of the dissimilarity between s and prototypical first reward activity . We finally computed the overall difference of deviation from the prototypical discharge at first reward as D¯=D¯T+−D¯T− . D¯ was computed for multiple time window lengths: from 100 ms to 1 s post-feedback time , by increments of 100 ms . Finally , a bias score b=∑positive biaswindows−log10 ( psigned−rank ( D¯ ) ) +∑negative biaswindowslog10 ( psigned−rank ( D¯ ) ) was computed , where the set of “positive bias windows” contained those analysis windows for which the sum of ranks for positive values was larger than the sum of ranks for negative values . Similarly , the “negative bias windows” were those with a sum of ranks for negative values larger than the sum of ranks for positive values . A positive ( negative ) bias in a given window would cause a corresponding increase ( decrease ) in b . To assess the significance of the bias score b , 1 , 000 surrogate datasets , in which the difference between high and low response time groups was eliminated , were compared to the real data . For each surrogate , and independently for each neuron , the sign of all D¯ values ( for all analysis windows ) had a 0 . 5 probability to be changed . The p-value was computed as the proportion of surrogate datasets leading to higher or equal |b|as the real data . The temporal sensitivity q leading to best first reward discrimination in the population and q = 0 were compared ( signed-rank test ) . To do so , we computed the mean values of D¯ over analysis windows ending from 100 ms to 1 s post-feedback time , by increments of 100 ms . Similar results were found when assessing the optimal q value by using either the original Victor and Purpura distance dq ( main text ) , or the normalized distance d*q ( S12 Fig and S4 Text ) . A similar analysis was done to test whether firing rates could also relate to response time . To do so , D¯ was replaced by the difference in mean firing rate between high and low response time trials , D¯rate . Table 1 summarizes an additional set of employed statistical measures . The latter were often non-normal; therefore , non-parametric tests were considered ( p ≤ 0 . 05 was considered as statistically significant ) : ( 1 ) correlations were assessed with Spearman coefficient with a permutation test ( or big sample approximation ) ; ( 2 ) distributions were compared with the 2-sided Kolmogorov-Smirnov test; ( 3 ) central tendencies were compared between 2 unpaired ( resp . paired ) distributions with the 2-sided ranked-sum ( signed-rank ) test; ( 4 ) deviation of distributions from 0-centered-symmetry was also tested with the 2-sided signed-rank test . When testing pairs of units , one limitation was that some pairs happened to share a neuron , and hence were correlated ( in particular if non-shared neurons were discharging significantly less than the shared one ) . This was problematic for analyzing the optimal temporal sensitivity , which is not a parameter accounting for the interaction between neurons , and which can be impacted more by the neuron which fires the most . We therefore verified that the significance of the advantage of the temporal sensitivity during paired decoding could be reached without overlapping pairs ( positivity of maxk ( 〈I ( q=10s−1 ) 〉t ) −maxk ( 〈I ( q=0s−1 ) 〉t ) , signed-rank test , p ≤ 0 . 05 in 1 , 000/1 , 000 random down-samplings to non-overlapping pairs ) . Note that , in contrast , interaction parameters such as the information gain or kopt are truly pair specific , implying that it was reasonable to keep overlapping pairs for the analysis . Note that although most statistical tests presented in the main results were carried out by pooling data from both monkeys , consistent trends were observed for both individuals .
|
In classical views of how information is processed in the brain , cognitive areas are often thought to encode incoming signals by a simple summation of spikes—action potentials fired by neurons and transmitted along the nerves—elicited by different cues . It is through this summation of spikes that cognitive areas are hypothesized to combine information from different cues and build memories . We investigated whether summation is relevant during the processing of signals emitted by the dorsal anterior cingulate cortex , a brain area which is thought to control behavioral adaptation in response to feedback cues that indicate the animal’s performance during a task . We found that a mere summation of spikes emitted in response to feedback cues actually extracts significantly less information compared to a code that also takes into account the timing of the spikes . Furthermore , we discovered that this temporal structure of the spike discharges predicts well the future behavior of monkeys . Overall , our findings suggest that the brain areas processing the signals emitted by the dorsal anterior cingulate cortex are sensitive to spike times and , thus , are unlikely to implement a mere approximate summation of inputs .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2015
|
Spatiotemporal Spike Coding of Behavioral Adaptation in the Dorsal Anterior Cingulate Cortex
|
Known protein coding gene exons compose less than 3% of the human genome . The remaining 97% is largely uncharted territory , with only a small fraction characterized . The recent observation of transcription in this intergenic territory has stimulated debate about the extent of intergenic transcription and whether these intergenic RNAs are functional . Here we directly observed with a large set of RNA-seq data covering a wide array of human tissue types that the majority of the genome is indeed transcribed , corroborating recent observations by the ENCODE project . Furthermore , using de novo transcriptome assembly of this RNA-seq data , we found that intergenic regions encode far more long intergenic noncoding RNAs ( lincRNAs ) than previously described , helping to resolve the discrepancy between the vast amount of observed intergenic transcription and the limited number of previously known lincRNAs . In total , we identified tens of thousands of putative lincRNAs expressed at a minimum of one copy per cell , significantly expanding upon prior lincRNA annotation sets . These lincRNAs are specifically regulated and conserved rather than being the product of transcriptional noise . In addition , lincRNAs are strongly enriched for trait-associated SNPs suggesting a new mechanism by which intergenic trait-associated regions may function . These findings will enable the discovery and interrogation of novel intergenic functional elements .
A large fraction of the human genome consists of intergenic sequence . Once referred to as “junk DNA” , it is now clear that functional elements exist in intergenic regions . In fact , genome wide association studies have revealed that approximately half of all disease and trait-associated genomic regions are intergenic [1] . While some of these regions may function solely as DNA elements , it is now known that intergenic regions can be transcribed [2]–[7] , and a growing list of functional noncoding RNA genes within intergenic regions has emerged [8] . Despite this progress , a complete understanding of the extent of intergenic transcription and the identity of these transcripts has remained elusive . The first attempts to analyze the extent and nature of intergenic transcription utilized tiling array technology [2]–[5] . These studies suggested that intergenic transcription is pervasive , but concerns about cross-hybridization have fueled a debate about the data [9]–[12] . Furthermore , in order to avoid technical difficulties associated with analyzing repeat sequence using tiling arrays , the studies were restricted to evaluating less than half of the genome . More recently , a few studies have focused on evaluating the extent of intergenic transcription using sequencing-based approaches , but with the exception of the recently published ENCODE project results [13] , [14] , these studies have thus far been limited to very narrow preselected regions of the genome and a small number of tissues [6] , [7] . Overcoming these prior shortcomings , the ENCODE project used RNA-seq analysis in combination with other technologies to profile 15 human cell lines , providing evidence for transcription across 83 . 7% of the human genome and firmly establishing the reality of pervasive transcription [14] . Long intergenic noncoding RNAs ( lincRNAs ) are defined as intergenic ( relative to current gene annotations ) transcripts longer than 200 nucleotides in length that lack protein coding capacity . LincRNAs are known to perform myriad functions through diverse mechanisms ranging from the regulation of epigenetic modifications and gene expression to acting as scaffolds for protein signaling complexes [8] , [15] . The first attempts to generate lincRNA annotation sets either profiled lincRNAs specific to a small number of tissues or required that transcripts harbor specific structural features such as splicing and polyadenylation [16]–[18] . The GENCODE consortium ( GENCODE v7 ) has manually curated approximately five thousand lincRNAs that are not restricted to particular tissues or structural features , however this annotation set contains only a small fraction of all lincRNAs because it does not take advantage of RNA-seq data to identify novel transcripts [19] , [20] . The limited scale of current lincRNA annotations , including GENCODE , is clearly incompatible with the massive amount of intergenic transcription observed by the ENCODE project . It should therefore be expected that the genome encodes far more lincRNAs than are currently known . In order to bridge the gap between the observation of pervasive intergenic transcription by the ENCODE project and the currently limited set of annotated lincRNAs , we performed an analysis of a unique set of RNA-seq data derived from both novel and published datasets that complements and significantly expands prior efforts [14] , [16] , [19] . This analysis resulted in a clear corroboration of the observations of pervasive transcription across the human genome by the ENCODE project [14] . Furthermore , analysis of previously annotated putative lincRNAs , including those of the ENCODE project [19] , in addition to de novo discovery of novel lincRNAs from RNA-seq data has resulted in the compilation of the most comprehensive catalog of human lincRNAs . Owing to the extended breadth of tissues sampled and relaxed constraints on transcript structure , we find significantly more lincRNAs than all previous lincRNA annotation sets combined . Our analyses revealed that these lincRNAs display many features consistent with functionality , contrasting prior claims that intergenic transcription is primarily the product of transcriptional noise [12] . In sum , our findings corroborate recent reports of pervasive transcription across the human genome and demonstrate that intergenic transcription results in the production of a large number of previously unknown lincRNAs . We provide this vastly expanded lincRNA annotation set as an important resource for the study of intergenic functional elements in human health and disease .
We have analyzed six novel RNA-seq datasets generated as part of the Human Epigenome Atlas ( http://www . genboree . org/epigenomeatlas/index . rhtml ) and 121 previously published RNA-seq datasets representing 23 human tissues under multiple conditions and consisting of over 4 . 5 billion uniquely mapped reads ( Table S1 ) . This set of RNA-seq data allowed for detection of both rare and tissue-specific transcription events that would otherwise be undetectable . In contrast to the limited reach of prior tiling array studies [2]–[5] , we analyzed the much larger portion ( 83 . 4% ) of the genome to which RNA-seq reads can be uniquely mapped thus providing a broader view of the transcriptome . At a threshold of one RNA-seq read , we observed reads mapping to 78 . 9% of the genome and , if additional evidence of transcription is taken into account including the full structures of known genes , spliced ESTs and cDNAs , we found evidence that 85 . 2% of the genome is transcribed ( Figure 1A ) . This result closely agrees with the recently published findings from the ENCODE project in which evidence for transcription of 83 . 7% of the genome was uncovered [14] . Interestingly , even with 4 . 5 billion mapped reads , we observe an increase in genomic coverage at each lower read threshold implying that even more read depth may reveal yet higher genomic coverage . ( Figure S1 ) . As expected , protein coding gene exons contain the largest fraction of highly expressed bases ( Figure 1B ) as well as a disproportionately large fraction of total reads relative to their small ( <3% ) amount of genomic sequence ( Figure S2 ) . However , many regions of high expression do exist within intergenic regions , far more than are accounted for by current noncoding RNA gene annotations ( Figure 1C ) . We reasoned that this unaccounted for intergenic transcription must derive from novel intergenic transcripts , and we next directed our efforts toward identification and analysis of these transcripts . We hypothesized that much of the intergenic transcription not accounted for by previously annotated transcripts is derived from novel lincRNAs . We reasoned that because lincRNA expression is known to be highly tissue-specific [16] , the breadth of tissues and conditions sampled in the RNA-seq datasets analyzed here would aid lincRNA discovery . We used this large set of RNA-seq data in combination with previous noncoding RNA annotation sets to generate the most comprehensive catalog of lincRNAs ( Figure 2A ) . In order to generate this lincRNA catalog , we first compiled known and putative annotated lincRNAs . We collected noncoding RNAs present in public databases , including GENCODE v6 , and from literature sources [16] , [18] resulting in a set of 351 , 940 transcripts . In addition , we performed de novo transcriptome assembly on each of the RNA-seq datasets ( Table S2 ) to generate 6 , 833 , 809 de novo assembled transcripts . Both previously annotated and de novo assembled transcripts were filtered to remove transcripts overlapping protein coding genes , known non-lincRNA noncoding RNA genes , and pseudogenes . Transcripts longer than 200 nucleotides were further filtered to remove any transcripts containing ( or overlapping any other transcript containing ) an open reading frame ( ORF ) longer than 100 amino acids . Out of concern that some de novo assembled transcripts may be unannotated extensions of neighboring protein coding genes , as was recently observed for a fraction of GENCODE long noncoding RNAs [19] , we created an additional filter to remove transcripts linked to neighboring genes by RNA-seq reads . To do this , we extended protein coding gene reference annotations using de novo transcriptome assembly and removed transcripts overlapping these extended gene structures ( see Methods , Dataset S1 ) . In a final step , we removed transcripts expressed at fragments per kilobase of transcript per million mapped reads ( FPKM ) <1 , a threshold approximately equivalent to one copy per cell [21] ( Table S1 ) . To decrease redundancy , and with the goal of identifying lincRNA “genes” rather than potentially redundant overlapping “transcripts” , the remaining transcripts were merged if they shared at least one exon ( see Methods ) resulting in 53 , 864 distinct putative lincRNAs at FPKM>1 , 3 , 676 lincRNAs at FPKM>10 , and 925 lincRNAs at FPKM>30 ( Dataset S2 and Figure S3 ) . Surprisingly , greater than 94% of the final set of merged lincRNAs at each expression level consists exclusively of novel de novo assembled transcripts discovered from the RNA-seq data in this study ( Table S3 and Dataset S2 ) . Rather than being clustered near currently annotated genes , these lincRNAs are spread throughout intergenic sequence . 58 . 1% of FPKM>1 lincRNAs , 61 . 9% of FPKM>10 lincRNAs , and 67 . 7% of FPKM>30 lincRNAs are greater than 30 kilobases from the nearest protein coding gene on either strand . We annotated the lincRNAs as belonging to the same “group” ( see Methods ) if they are within 1 kilobase of each other to account for the possibility that some proximal lincRNA annotations may be partial structures of larger transcripts ( see Discussion ) . This grouping resulted in 35 , 585 distinct lincRNA groups at FPKM>1 , 2 , 970 at FPKM>10 , and 764 at FPKM>30 , and the lincRNAs in the catalog are named according to these groups ( Dataset S2 ) . These annotations are likely to be incomplete due to limitations in transcript assembly from RNA-seq data; indeed , some annotations may be fragments of larger overlapping lincRNA transcripts . Therefore , the actual number of independent lincRNAs may differ from the above numbers , and future work is needed to more fully define complete , independent lincRNA transcript annotations ( see Discussion ) . We evaluated the stringency with which our filtering process removed protein coding transcripts by analyzing ribosomal profiling data from HeLa cells ( Figure 2B ) [22] . As expected , lincRNAs resemble the 3′ untranslated region exons of protein coding genes , with very few transcripts showing significant engagement with the ribosome . This finding is in agreement with the recent observation that GENCODE long noncoding RNAs ( a subset of our catalog ) generally lack mass spectrometry based evidence for translation [23] . In contrast , a recent study found that many previously annotated mouse lincRNAs bind the ribosome [24] . While the biological significance of this discrepancy is unknown , it may be the result of differences in the stringency of the filtering approach employed in the generation of the lincRNA annotations under consideration . Further confirming the stringency of our filters , a computational analysis of protein coding potential using the program PhyloCSF revealed that our set of filtered lincRNAs lack predicted protein coding capacity ( Figure 2C ) . From these analyses we conclude that our filtering approach effectively removed protein coding transcripts from the catalog . While the remainder of this study focuses on this catalog of putative lincRNAs ( Dataset S2 ) , we have provided multiple alternative lincRNA catalogs . These include a combined catalog of the lincRNAs identified in this study merged ( see Methods ) with a set of additional lincRNAs identified in Cabili , et al . [16] which passed all of our filters except were not expressed at FPKM>1 in any of the RNA-seq datasets analyzed here . The added lincRNAs are expressed at FPKM>1 in one or more of the RNA-seq datasets analyzed in Cabili et al . [16] , which are entirely distinct from the datasets analyzed here , and are therefore likely to be genuine lincRNAs by our criteria . This catalog ( Dataset S3 ) includes 54 , 784 lincRNAs at FPKM>1 ( 920 additional lincRNAs compared to Dataset S2 ) , 3 , 764 lincRNAs at FPKM>10 ( 88 additional lincRNAs ) , and 942 lincRNAs at FPKM>30 ( 17 additional lincRNAs ) . In addition , we have included a catalog of spliced lincRNAs that are expressed at FPKM>1 in at least one dataset ( 4 , 576 lincRNAs , Dataset S4 ) , of which 61% are exclusively composed of de novo assembled transcripts discovered in this study . We have also compiled a catalog of lincRNAs expressed at FPKM>1 in at least two datasets ( 26 , 455 lincRNAs , Dataset S5 ) , of which 97% are exclusively de novo assembled transcripts discovered here . Additionally , an alternative lincRNA catalog containing only those lincRNAs expressed significantly higher than randomly sampled intergenic regions ( see Methods ) were included ( 5 , 267 lincRNAs , Datasets S6 , S7 ) . Furthermore , as an additional resource we provide the expression level ( FPKM and raw RNA-seq read counts ) of all lincRNAs ( in Dataset S2 ) and RefSeq protein coding genes across all 127 RNA-seq datasets ( Dataset S8 ) . The degree to which intergenic transcription is functional remains uncertain and controversial [9]–[12] , [25] . In order to evaluate whether the lincRNAs identified in the present study are specifically regulated as opposed to transcriptional noise , we determined if the lincRNA genes harbor canonical epigenetic marks for activation and repression with the reasoning that noise transcripts should lack coherent epigenetic modification patterns . Consistent with observations based on earlier long noncoding RNA annotations [18] , [19] , [26] , [27] , analysis of ChIP-seq and RNA-seq data [28] , [29] revealed that the catalog of lincRNAs shows patterns of epigenetic modification similar to protein coding genes ( Figure 3A ) . Activating histone marks , H3K4me3 and H3K36me3 , are both significantly enriched within highly expressed lincRNAs . Similarly , the repressive mark H3K27me3 is significantly enriched within lowly expressed lincRNAs . Thus , the expression of lincRNAs appears to be specifically regulated . If lincRNAs are specifically regulated at the level of transcription , it is expected that their expression levels are specific to their tissue source . Indeed , prior studies of lincRNAs have shown that lincRNAs display very strong tissue-specific expression [16] , [19] . To test whether this remains true with our expanded set of lincRNAs we performed unsupervised hierarchical clustering using lincRNA expression levels in replicate RNA-seq datasets from various tissues ( Figure S4 ) . Replicates of each tissue type strongly clustered together , indicating that lincRNA differential expression is indeed reproducibly tissue-specific , supporting specific regulation of lincRNA expression . LincRNAs do not need to be polyadenylated to be functional [30] . Because of this , we included in our analysis many RNA-seq libraries that were not polyA+ selected . In fact , earlier tiling array studies revealed that intergenic transcripts tend to be bimorphic; that is , they appear in both polyA+ and polyA− fractions , as opposed to protein coding transcripts that are primarily polyA+ [3] . The recently published ENCODE results corroborate this finding [14] , [19] . In agreement with these studies , we found that the polyadenylation status of lincRNAs in our catalog is reproducibly bimorphic across multiple cell types while protein coding transcripts are strongly enriched in the polyA+ sample . The reproducibility of this lincRNA bimorphic state suggests that lincRNA polyadenylation is regulated and that many lincRNAs exist at least partially as nonpolyadenylated transcripts ( Figure 3B and Figure S5 ) . This finding indicates that future studies of lincRNAs should not ignore the nonpolyadenylated RNA fraction . We next evaluated whether lincRNAs are conserved . It has been observed that lincRNAs can contain conserved motifs tethered together by nonconserved sequence [25] , [31] , [32] . Therefore , we evaluated lincRNA conservation using a scanning 50 bp window ( Figure 3C , Figure S6 , and Table S4 ) . Consistent with prior studies , lincRNAs display detectable but modest conservation [16] , [19] . We applied this same method to known functional human lincRNAs and found that the majority of the lincRNAs identified in this study display a level of conservation consistent with known functional lincRNAs ( Figure 3C ) . Almost half of all trait-associated SNPs ( TASs ) identified in genome-wide association studies are located in intergenic sequence while only a small portion are in protein coding gene exons [1] . This curious observation points to an abundance of functional elements in intergenic sequence . While some of these regions may function at the DNA level alone , it is possible that many function by encoding RNA . In fact , TASs have already been identified within or proximal to noncoding RNAs including some lincRNAs [16] , [33]–[36] . We reasoned that if lincRNAs are functional , they should be enriched for TASs compared to nonexpressed intergenic regions . Indeed , we find that lincRNAs are more than 5-fold enriched for TASs compared to nonexpressed intergenic regions ( Figure 4 ) despite an approximately equal distribution of SNPs between these regions ( Figure S7 ) . Therefore , many trait-associated intergenic regions may function by encoding lincRNAs .
There has been a recent debate about whether there is pervasive transcription of the human genome and what the number and abundance of intergenic transcripts is [9]–[12] . Until recently , a key missing component to this debate has been an analysis of ultra deep RNA-seq data sampling a wide array of tissue types . Without this , insufficient read depth can result in a failure to identify low abundance intergenic transcripts , and limited tissue sampling results in missed tissue specific expression . During the course of this study , the ENCODE project released a large scale analysis of RNA-seq data that provided clear evidence that the human genome is pervasively transcribed [14] . We analyzed a distinct , complementary set of RNA-seq data that also fulfills these requirements of read depth and tissue breadth , covering both polyadenylated and nonpolyadenylated RNA fractions . In strong agreement with the ENCODE results , we observed that approximately 85% of the genome is transcribed , supporting prior observations of pervasive transcription based on tiling arrays that have been recently questioned [2]–[5] . There is an apparent discrepancy between this observed pervasive transcription and the relative paucity of annotated lincRNAs , the most numerous intergenic RNAs . It should be expected that intergenic regions encode far more lincRNAs than are currently annotated . Indeed , here we found that there are many more lincRNAs than previously known , even after aggressive filtering that removed the vast majority of previously annotated long noncoding RNAs and newly discovered intergenic transcripts ( Dataset S2 ) . These observations clearly demonstrate that the human genome is pervasively transcribed , and that lincRNAs make up an extremely common class of intergenic transcripts . In agreement with prior observations of smaller lincRNA annotation sets , our analyses of the expanded lincRNA catalog presented here revealed that most lincRNAs are expressed at lower levels than protein coding genes [16] , [19] . Though most lincRNAs are expressed at only a few copies per cell , we found that many lincRNAs are highly expressed with nearly 4 , 000 expressed at >FPKM 10 and nearly 1 , 000 expressed at >FPKM 30 , rivaling the expression of many messenger RNAs . We chose to apply an expression cutoff to remove very lowly expressed transcripts from the catalog of lincRNAs . However , it may be the case that there exist many functional lincRNAs with very low expression levels , below our expression filter cutoff . For example , the functional human lincRNA HOTTIP is expressed in approximately one out of three cells [37] . Furthermore , recent findings have shown that the intergenic transcriptome may be vastly more complex than currently appreciated when very lowly expressed transcripts are considered [7] . It is possible that some of these are functional transcripts despite their apparent low expression , perhaps having brief bursts of expression during stages of the cell cycle or functioning in single cells in a heterogeneous population as has been previously observed [14] . Therefore , while we have provided the most complete lincRNA catalog to date , there may be additional lowly expressed , yet potentially functional lincRNAs that were excluded here . In order to minimize any potential contamination of the lincRNA catalog with protein coding transcripts , the filtering approach used was very aggressive . In fact , most previously annotated noncoding RNAs failed to pass our filters and were therefore excluded from the lincRNA catalog ( Table S3 and Dataset S9 ) . The vast majority of these transcripts ( including most GENCODEv6 “lincRNAs” and “processed transcripts” ) overlap known or predicted protein coding genes , pseudogenes , or non-lincRNA noncoding RNAs ( e . g . microRNAs ) ( Table S3 ) . Some of these removed transcripts may be functional long noncoding RNAs , such as GAS5 ( removed because it contains 10 snoRNA genes within its introns ) . However , in order to most confidently identify only lincRNAs , rather than potential unannotated extensions of known genes , these were removed . Of those previously annotated noncoding RNAs that are intergenic , more than half contain predicted ORFs longer than 100 amino acids . For example , two previously characterized functional human lincRNAs were found to contain ORFs longer than 100 amino acids , Xist and HOTAIR . These results demonstrate that our filtering approach , which eliminates all transcripts with ORFs larger than 100 amino acids , may have removed some lincRNAs with large , nonfunctional ORFs . However , the use of a 100 amino acid ORF cutoff , a commonly used threshold to define potential protein coding genes , is justifiable because ORFs of this size infrequently occur by chance and instead indicate potential for protein coding capacity [38] , [39] . Rather than discard all transcripts with large ORFs , as we did here , one option to discriminate between transcripts that are coding versus noncoding is to analyze the frequency of synonymous codon substitutions ( PhyloCSF ) [40] . However , this approach is limited to ORFs that can be aligned across species , potentially missing recently evolved or otherwise nonconserved novel protein coding genes . Importantly , our approach of removing all transcripts with large open reading frames effectively removed transcripts with significant predicted coding potential ( Figure 2C ) , indicating that using an ORF size cutoff is at least as conservative as filtering based on PhyloCSF analysis . The lack of engagement of the ribosome , observed with ribosomal profiling data , confirms the stringency of the ORF cutoff filter ( Figure 2B ) . Further analysis of these removed large ORF-containing intergenic transcripts is outside the scope of this study , but we have included these annotations for investigators interested in further analyzing their coding potential in search of novel protein coding genes ( Dataset S10 ) . Despite the fact that most previously annotated noncoding RNAs failed to pass our filters , our lincRNA catalog contains significantly more lincRNAs than previously known ( >94% of lincRNAs are entirely novel at each expression level ) . This is the result of two unique features of our study . First , the RNA-seq read depth and diversity of tissues surveyed allowed for the detection of rare and tissue specific transcripts that were previously unknown . Many of these novel transcripts passed all filters and are annotated as novel lincRNAs in our catalog . Second , in contrast to prior lincRNA annotation efforts that were restricted to identification of only spliced or polyadenylated lincRNAs [16] , [19] , [41] , we sought to generate annotations of a more complete set of human lincRNAs regardless of splicing or polyadenylation status . The reasons for taking this approach are manifold . Two of the most well known and abundant functional human lincRNAs , NEAT1 and MALAT1 , are single exon genes ( as are approximately 5% of protein coding genes ) [42] , suggesting that non-spliced transcripts may make up an important class of lincRNA . Additionally , numerous functional nonpolyadenylated noncoding RNAs have been described [30] , [43] . Even long noncoding RNAs which can be spliced are often found in their unprocessed forms [44] , a distinct property of long noncoding RNAs that would result in missed lincRNAs if splicing were a required attribute . Therefore , we chose not to exclude any lincRNAs from this catalog due to lack of splicing or polyadenylation . Importantly , because nonspliced , nonpolyadenylated transcripts could theoretically be erroneously de novo assembled from reads derived from contaminating genomic DNA in RNA-seq data , we took multiple measures to mitigate any contributions of genomic DNA contaminant reads ( see Methods ) . Due to inherent limitations of de novo transcriptome assembly using short reads of finite depth , it is not always possible to unequivocally determine the complete structure of a transcript . This is particularly true for lowly expressed transcripts where the number of reads available is limited , and for genomic regions to which reads cannot be uniquely mapped . In the case of shallow read depth , exons of multi-exonic transcripts may lack reads connecting the exons , and de novo assembly could result in separate annotation of each exon as a distinct transcript . In support of this , we found that lower expressed lincRNAs discovered from de novo transcript assembly were less likely to have multi-exonic structures ( Table S5 ) . Additionally , the annotated 5′ and 3′ ends of the lincRNAs may represent truncations of the full length transcripts . Indeed , our analysis of PET tag data revealed that while the majority of our lincRNA catalog is overlapped by at least one PET tag , in most cases there is minimal PET tag support for the annotated 5′ and 3′ ends of the lincRNAs ( Table S6 ) . It is therefore the case that some lincRNA annotations in the catalog we provide ( Dataset S2 ) , particularly single exon lincRNA annotations , may represent fragments of larger transcripts . Furthermore , considering the reported prevalence of low level overlapping transcripts throughout intergenic sequence [7] , it is not clear that full lincRNA structures can be unequivocally deconvoluted using short read RNA-seq technology . The determination of full lincRNA structures will be an important future effort in the field and may rely upon new datasets of longer read length and greater read depth , use of multiple orthogonal data types in the same tissue , new technologies such as ultra long read next generation sequencing , and further improvements in software for de novo transcript assembly . In addition , the majority of RNA-seq data we analyzed lacks strand information and as a result most of the lincRNAs in our catalog are of ambiguous strandedness . Prior annotations have relied upon splice site orientation to infer the strandedness of the transcript [16] . While this is a reasonable approach that we too have adopted when applicable in the present lincRNA catalog , stranded RNA-seq data is needed to most confidently assign strandedness to de novo assembled transcripts . While determining the isoforms and full structures of all lincRNAs is clearly desirable , these incomplete lincRNA structure annotations are nonetheless of tremendous practical value . Knowledge of the structure of a portion of a transcript is often sufficient to test for differential expression or perform RNAi knockdown experiments , and facilitates the cloning and sequencing of the full length transcript . Because of this , instead of placing additional restrictions upon lincRNA annotations , our filtering strategy was aimed toward identification of as many transcripts as possible that fit within the definition of a lincRNA . However , for investigators interested in more refined lincRNA annotations , we have provided multiple more restrictive lincRNA catalogs ( Datasets S4 , S5 , S6 ) . A key question in the field is whether the transcripts resulting from pervasive transcription of intergenic regions are functional or the result of noisy transcription . The lincRNAs we describe are specifically regulated and contain conserved sequence , attributes inconsistent with transcriptional noise ( Figure 3 ) . Furthermore , lincRNAs were found to be strongly enriched for intergenic TASs compared to nonexpressed intergenic regions ( Figure 4 ) . This striking finding supports the possibility that many intergenic SNPs mark regions that function as lincRNAs rather than DNA elements . Because nearly half of all TASs are intergenic , it is possible that lincRNAs play a significant role in the majority of human traits and diseases thus far analyzed in GWASs . One functional lincRNA ( MIAT ) was first identified during the experimental interrogation of an intergenic TAS [35] , and another lincRNA PTCSC3 , was identified nearby a TAS found from a papillary thyroid carcinoma GWAS , perhaps representing the first of many such discoveries to come from intergenic TASs . The finding that lincRNAs are strongly enriched for TASs provides a new opportunity to revisit intergenic trait-associated regions with unknown functional mechanisms by testing whether the overlapping lincRNA is involved in the observed phenotype . This noncoding RNA catalog represents a major step toward achieving a more complete understanding of this exciting frontier . We have identified a large number of putative lincRNAs with characteristics suggesting functionality . However , many of these lincRNAs are low expressed and definitive proof of functionality for a lincRNA requires functional experiments . High throughput functional genomic approaches , such as RNAi and cDNA overexpression screens , will serve as crucial tools for future efforts to uncover the roles of lincRNAs in diverse biological systems . With the requisite technology now available for these next generation experimental approaches , the time is ripe for this dark matter of the human genome to step further into the spotlight .
127 RNA-seq sequence files ( 5 novel and 122 publicly available datasets , Table S1 ) were aligned to hg18 with TopHat v1 . 1 . 4 allowing only uniquely mapped reads using the option -g 1 ( all other parameters were default , see the TopHat manual http://tophat . cbcb . umd . edu/manual . html ) . Detailed information pertaining to each dataset , including novel datasets , is available in the sources provided in Table S1 . These RNA-seq datasets were chosen because they sampled a wide breadth of human tissues and cell types , have well documented experimental methods used for their generation , and were publicly available . While datasets with longer reads and deeper read depth were preferred because they allow for more complete de novo transcript assembly , some datasets with short reads and shallow read depths were included in order to sample as many tissue types as possible . Datasets derived from tissues with mutated genomes , such as cancers , were included to capture tissue specific expression even though some reads from mutated genomic positions would fail to map to the reference hg18 genome . SAMtools v0 . 1 . 7 and BEDTools v2 . 12 . 0 were used to process aligned read files . The uniquely mappable human genome , defined here as the portions of the genome to which RNA-seq reads can be uniquely mapped , was derived for hg18 from http://www . imagenix . com/uniqueome/downloads/hg18_uniqueome . unique_starts . base-space . 50 . 2 . positive . BED . gz [45] . It contains 2 , 570 , 174 , 327 bp or 83 . 4% of the total human genomic sequence . To determine the genomic coverage of RNA-seq data , all aligned RNA-seq reads were combined and read coverage at each genomic base position was determined with the BEDTools function genomeCoverageBed . Split reads ( i . e . exon-exon junction spanning reads ) were counted such that intronic sequence was included as part of the reads . In Figure 1A , “All genes , ESTs , cDNAs” includes GENCODE v10 genes ( excluding pseudogenes ) , RefSeq NM and NR genes , UCSC Known Genes , spliced H-Invitational cDNAs , spliced ESTs ( UCSC Genome Browser “Spliced EST” track ) , and previously annotated spliced lincRNAs [16] . In all cases , intronic sequences of genes , cDNAs and ESTs were included . An additional set of annotated lincRNA transcripts from Cabili et al . [16] passed all our filters except were not expressed at FPKM>1 in any of the datasets analyzed here and were therefore removed from the lincRNA catalog in Dataset S2 . However , some of these transcripts were reported as expressed at FPKM>1 in at least one of the datasets analyzed in Cabili et al . [16] , all of which are distinct from the datasets analyzed here . These additional lincRNAs were merged with the lincRNAs in the catalog in Dataset S2 resulting in an additional 920 lincRNAs in 741 groups at FPKM>1 , 88 lincRNAs in 82 groups at FPKM>10 , and 17 lincRNAs in 17 groups at FPKM>30 . This expanded lincRNA catalog is in BED format for genome build hg18 in Dataset S3 and was not used further for any analyses in this study . Genomic DNA contamination is a potential source of false positive expression signal in RNA-seq data that may contribute to de novo assembly of erroneous transcripts . In principle , only exon-exon junction spanning reads can be unequivocally determined as derived from RNA . Proper de novo assembly of both nonspliced and spliced ( aside from the exon-exon junction spanning reads ) transcripts may therefore suffer if significant genomic DNA contamination is present . Because our analysis utilized a wide range of novel and previously existing RNA-seq datasets of unknown genomic DNA contamination content , we took multiple steps to mitigate this possibility . First , for all RNA-seq datasets , we analyzed the distribution of reads between protein coding exons compared to other regions with the expectation that read distributions should be similar between RNA-seq datasets generated from libraries of the same type ( e . g . polyA+ selected ) . A dataset with an unusually high percentage of intronic and intergenic reads could contain significant genomic DNA contamination . Our analysis of the datasets used in this study revealed that , as expected , polyA+ specific RNA-seq datasets have a higher fraction of reads mapping to protein coding gene exons than rRNA-depleted or polyA− specific datasets . Furthermore , no obvious outlier datasets were found for any of the library types . The results of this analysis ensured that no datasets with high genomic DNA contamination were used in this study ( Figure S2 ) . Next , as described in Figure 2A and in the Methods , we applied both size and expression cutoffs for all lincRNAs . The size cutoff prevents miscalling errant single reads , either from genomic DNA contamination or from read mapping artifacts , as lincRNAs while the expression cutoff removes lincRNAs that are assembled from rare genomic DNA-derived reads . The combination of these approaches served to minimize the contribution of genomic DNA to the lincRNA catalog . H9 ESC and HeLa RNA-seq data from fractions exclusively containing polyA− or polyA+ transcripts were analyzed [46] . Transcripts with RNA-seq reads in all four datasets and with FPKM>1 in at least one of the two fractions for each cell type were analyzed for Figure 3B ( 16 , 819 NM genes and 127 lincRNAs ) . For Figure S5 , transcripts with reads in both fractions and FPKM>1 in at least one of the two fractions for a specific cell type were included in the analysis of that cell type ( 20 , 470 NM genes and 849 lincRNAs in H9 ESCs; 18 , 294 NM genes and 1 , 009 lincRNAs in HeLa ) . The whiskers of the box and whisker plot extend to +/−1 . 5 times the interquartile range or the most extreme datapoint . Publicly available paired-end ditag ( PET ) cluster annotations derived from 7 cell lines or tissues , generated by the ENCODE project , were downloaded from http://genome . ucsc . edu/cgi-bin/hgFileUi ? db=hg19&g=wgEncodeGisRnaPet . The PET cluster annotation files used were ( by cell or tissue type ) : A549 ( wgEncodeGisRnaPetA549CellPapClusters . bedCluster ) , H1_hESC ( wgEncodeGisRnaPetH1hescCellPapClustersRep1 . bed ) , HeLa-S3 ( wgEncodeGisRnaPetHelas3CellPapClustersRep1 . bed ) , IMR90 ( wgEncodeGisRnaPetImr90CellPapClusters . bedCluster ) , MCF-7 ( wgEncodeGisRnaPetMcf7CellPapClusters . bedCluster ) , Prostate ( wgEncodeGisRnaPetProstateCellPapClustersRep1 . bed ) , SK-N-SH ( wgEncodeGisRnaPetSknshCellPapClusters . bedCluster ) . Further description of these PET clusters , including how the annotations were generated , is available at the UCSC Genome Browser site here http://genome . ucsc . edu/cgi-bin/hgTrackUi ? hgsid=321010719&c=chr21&g=wgEncodeGisRnaPet . BEDTools was employed to compute overlap between lincRNA and RefSeq NM gene 5′ and 3′ ends and PET cluster 5′ and 3′ end ‘blocks’ . In the case of ambiguous stranded lincRNAs , both potential orientations were allowed for determining overlap with the 5′ and 3′ ends of PET clusters . Ribosome profiling data and matched mRNA-seq data from HeLa cells corresponding to the experiments ( mock transfected 12 hr time point ) presented in Guo et al . [22] were downloaded from the NCBI GEO ( GSE22004 ) . The expression level of the filtered set of lincRNAs and of RefSeq NM transcripts was evaluated as above . The 803 lincRNAs expressed at an FPKM>1 and a sample of 1292 RefSeq NM transcripts expressed at an FPKM>1 ( divided into their constituent CDS and 3′ UTR regions ) were broken up into 30 bp windows with a 1 bp offset . A modified version of HTSeq ( described above ) was used to count reads aligning to each window for both RNA-seq and ribosomal profiling data . The ratio of ribosome-associated reads over mRNA-seq reads was evaluated for each window and the maximum ratio for a given transcript was taken as a measure of ribosome engagement . The whiskers of the box and whisker plot in Figure 2B extend to +/−1 . 5 times the interquartile range with outliers depicted as dots . Wilcoxon rank sum test was used to calculate P values . The program PhyloCSF ( 9/16/2010 release ) [40] was used to computationally evaluate the coding potential of the filtered lincRNA transcripts . A BED file describing these lincRNA transcripts as well as a random sample of 8310 RefSeq NM transcripts was loaded onto the Galaxy webserver ( https://main . g2 . bx . psu . edu/ ) and the tool ‘Stitch Gene Blocks’ was used to retrieve multiple alignment files with sequence entries for the following genome builds based on the 44 way Multiz alignment to hg18: hg18 panTro2 rheMac2 tarSyr1 micMur1 otoGar1 tupBel1 mm9 rn4 dipOrd1 cavPor3 speTri1 oryCun1 ochPri2 vicPac1 turTru1 bosTau4 equCab2 felCat3 canFam2 myoLuc1 pteVam1 eriEur1 sorAra1 loxAfr2 proCap1 echTel1 dasNov2 choHof1 . Genome build names were converted to common names and PhyloCSF was run using the options –orf = StopStop3 and –frames = 6 . ChIP-seq data from IMR90 cells [28] was retrieved from the NCBI SRA ( Table 1 ) and aligned to hg18 using Bowtie v0 . 12 . 7 allowing only uniquely mapped reads ( -k 1 ) . A modified version of HTSeq v0 . 5 . 3p ( described above ) was used to count reads mapping to lincRNAs and RefSeq NM genes . The ratio of IP reads to matched input control reads was used as the measure of ChIP signal . RNA-seq data from IMR90 cells [29] was also analyzed to obtain FPKM values for lincRNAs and RefSeq NM genes using the same procedure used for lincRNA discovery . The whiskers of the box and whisker plot extend to +/−1 . 5 times the interquartile range or the most extreme data point . RNA-seq datasets from B cells , H1 ESCs , and brain ( see Table S1 ) were clustered by lincRNA expression levels . LincRNAs with FPKM>10 in one or more of the 7 RNA-seq datasets analyzed in Figure 3B were used to generate the heatmap and dendrogram . These 7 datasets were chosen for this analysis because they have replicates from each tissue and have deep read counts for all replicates ( Table S1 ) , important features for accurate measurement of differential expression . Using Gene Cluster 3 . 0 , FPKM values were log2 transformed and the genes ( rows ) and samples ( columns ) were normalized by multiplying each log2 transformed FPKM value by a scale factor such that the sum of the squares of the values in each row and column are 1 . 0 . Euclidean distance using centroid linkage was calculated for all samples and the heatmap and dendrogram was generated with Java TreeView . Red corresponds to fully induced expression and blue corresponds to fully repressed expression . Base-wise conservation scores ( PhyloP score calculated with PHAST ) , based on the multiple alignment of placental mammal genomes , were downloaded from the UCSC Genome Browser . The 50 bp window in each lincRNA transcript with the highest average PhyloP score was identified . The process was repeated for RefSeq NM genes and a set of size-matched ( to lincRNAs ) repetitive elements from RepeatMasker ( UCSC Genome Browser ) . PhyloP scores for the maximally conserved 50 bp windows of each lincRNA are listed in Table S4 .
|
Much of the human genome is composed of intergenic sequence , the regions between genes . Intergenic sequence was once thought to be transcriptionally silent “junk DNA , ” but it has recently become apparent that intergenic regions can be transcribed . However , the scope , nature , and identity of this intergenic transcription remain unknown . Here , by analyzing a large set of RNA-seq data , we found that >85% of the genome is transcribed , allowing us to generate a comprehensive catalog of an important class of intergenic transcripts: long intergenic noncoding RNAs ( lincRNAs ) . We found that the genome encodes far more lincRNAs than previously known . A key question in the field is whether these intergenic transcripts are functional or transcriptional noise . We found that the lincRNAs we identified have many characteristics that are inconsistent with noise , including specific regulation of their expression , the presence of conserved sequence and evidence for regulated processing . Furthermore , these lincRNAs are strongly enriched with intergenic sequences that were previously known to be functional in human traits and diseases . This study provides an essential framework from which the functional elements in intergenic regions can be identified and characterized , facilitating future efforts toward understanding the roles of intergenic transcription in human health and disease .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"genomics",
"genome",
"expression",
"analysis",
"biology",
"computational",
"biology"
] |
2013
|
Pervasive Transcription of the Human Genome Produces Thousands of Previously Unidentified Long Intergenic Noncoding RNAs
|
Colonization and disruption of the epithelium is a major infection mechanism of mucosal pathogens . The epithelium counteracts infection by exfoliating damaged cells while maintaining the mucosal barrier function . The sexually transmitted bacterium Neisseria gonorrhoeae ( GC ) infects the female reproductive tract primarily from the endocervix , causing gonorrhea . However , the mechanism by which GC overcome the mucosal barrier remains elusive . Using a new human tissue model , we demonstrate that GC can penetrate into the human endocervix by inducing the exfoliation of columnar epithelial cells . We found that GC colonization causes endocervical epithelial cells to shed . The shedding results from the disassembly of the apical junctions that seal the epithelial barrier . Apical junction disruption and epithelial exfoliation increase GC penetration into the endocervical epithelium without reducing bacterial adherence to and invasion into epithelial cells . Both epithelial exfoliation and junction disruption require the activation and accumulation of non-muscle myosin II ( NMII ) at the apical surface and GC adherent sites . GC inoculation activates NMII by elevating the levels of the cytoplasmic Ca2+ and NMII regulatory light chain phosphorylation . Piliation of GC promotes , but the expression of a GC opacity-associated protein variant , OpaH that binds to the host surface proteins CEACAMs , inhibits GC-induced NMII activation and reorganization and Ca2+ flux . The inhibitory effects of OpaH lead to reductions in junction disruption , epithelial exfoliation , and GC penetration . Therefore , GC phase variation can modulate infection in the human endocervix by manipulating the activity of NMII and epithelial exfoliation .
Microbial pathogens establish infection at the mucosal surface by colonization , disruption , and penetration of the epithelium [1] . The epithelium is the first line of the host defense against microbial pathogens , providing a physical barrier and a sensor of invading pathogens [2 , 3] . In the female reproductive tract ( FRT ) , this mucosal surface is composed of multilayered non-polarized squamous epithelial cells at the ectocervix and vagina , or monolayered polarized columnar epithelial cells at the endocervix and uterus . Different from multilayered squamous epithelial cells that are held together by adherent junctions , the monolayer epithelium is sealed by the apical junction , which prevents the entry of pathogens through the paracellular space ( gate function ) and maintains the polarity of the apical and basolateral surfaces ( fence function ) [4–7] . The apical junction is formed by the integral proteins , claudin , occludin , junctional adhesion molecules , E-cadherin , and the associated proteins zonula occludens-1 ( ZO1 ) and β-catenin [7] . ZO1 and β-catenin link the apical junction to the actin cytoskeleton and signaling networks [7–11] . The actin cytoskeleton and non-muscle myosin II ( NMII ) form a supporting ring at the apical junction [12–14] . The contraction of the actomyosin ring can transiently open the “gate” of the apical junction , regulating the permeability of the epithelium [15–17] . Over activation of the actomyosin ring can lead to the disassembly of the apical junction by inducing the endocytosis of junctional proteins [18 , 19] . As a strategy of protecting the epithelium from pathogens , infected cells with associated microbes are shed [20] . While the exfoliation of multilayered squamous epithelium is mediated by weakening cell-cell adhesion , exfoliation of polarized epithelial monolayers requires collaboration between NMII and apical junctional complexes . Actomyosin and apical junctional proteins are recruited to the plasma membrane of epithelial cells in contact with an exfoliating cell . NMII-generated forces in neighboring cells “squeeze” the exfoliating cell out while apical junctional complexes ensure that the epithelial barrier remains uncompromised [21–25] . How bacteria break the epithelial barrier and escape from epithelial shedding to achieve infection remains elusive . Neisseria gonorrhoeae ( GC ) , a Gram-negative bacterium , infects the mucosal surface of human genital tissues in men and women and causes one of the most common sexually transmitted infections , gonorrhea [26] . In the FRT , the endocervix has been suggested as a primary site for GC to initiate infection that may lead to pelvic inflammatory disease [27 , 28] . Previous studies , using epithelial cells , fallopian tube organ culture , and mouse vaginal infection models , have shown that GC can adhere to , invade into , and transmigrate across epithelial cells [29–31] . However , how GC infect the polarized human columnar endocervical epithelial cells has not been well studied . GC major surface molecules , including pili , lipooligosaccharide ( LOS ) , porin , and opacity-associate protein ( Opa ) , function concertedly for infection . Opa has been suggested to be involved in GC adherence to , invasion into , and transmigration across polarized epithelial cells [32–36] , as well as GC-GC interaction by binding to LOS [37–39] . Opa , pili and LOS undergo phase variation . This phase variation has been implicated for the capability of GC to infect various locations of the FRT and generate different pathological conditions and complications [32 , 40 , 41] . Most GC isolated from patients [42] and infected mice are Opa positive [43] , underscoring the importance of Opa in infections . Opa has been shown to inhibit GC-induced exfoliation of squamous epithelial cells from the lower genital tract of mice by engaging carcinoembryonic antigen-related cell adhesion molecules ( CEACAMs ) and activating integrin , which enhances GC colonization [44 , 45] . These data indicate that Opa phase variation is a major way for GC to modify their pathogenicity . GC establishes infection by interacting with various receptors on epithelial cells , such as the binding of Opa to CEACAMs or heparin sulfate proteoglycans ( HSPG ) [34 , 46–48] . These interactions alter signaling cascades in epithelial cells , such as phosphatidylinositol 3-kinase , phospholipase C , and Ca2+ flux . The signaling leads to actin reorganization , which can drive microvillus elongation and the subsequent engulfment of GC [49 , 50] . We have shown that GC-induced transactivation of epidermal growth factor receptor ( EGFR ) is critical for the optimal level of GC invasion into non-polarized epithelial cells and transmigration across polarized epithelial cells [51 , 52] . GC interaction with polarized epithelial cells weakens the apical junction by inducing the disassociation of ZO1 and β-catenin from the junctional complex , consequently facilitating GC transmigration [51] . Our recently published studies found a surprising role for Opa in inhibiting GC transmigration across polarized epithelial cells [38] . How GC manipulate columnar endocervical epithelial cells through Opa for infection is unknown . A major obstacle against addressing this question has been a lack of infection models that mimic all aspects of human infection . In this study , we established a new ex vivo infection model , human endocervical tissue explants . Using this model and polarized epithelial cells , as well as isogenic strains of GC expressing invariable Opa , we revealed the mechanistic links between GC infectivity , GC-induced exfoliation , apical junction disassembly , and signaling in polarized columnar endocervical epithelial cells , and novel roles of Opa phase variation in these events . GC induce the exfoliation of polarized endocervical epithelial cells by disrupting the apical junction . Opposite to GC-induced shedding of squamous epithelial cells , the exfoliation of columnar epithelial cells does not reduce GC adherence and invasion; instead , it increases GC penetration into the subepithelium . Both GC-induced epithelial exfoliation and apical junction weakening require Ca2+-dependent redistribution of active NMII . The expression of CEACAM-binding OpaH but not HSPG-binding OpaC inhibits GC-induced exfoliation and junctional disruption by interfering with NMII activation and reorganization as well as Ca2+ flux , while GC piliation promotes these events . Our results suggest that GC modify the exfoliation process for infection by activating Ca+ flux and NMII redistribution in endocervical epithelial cells and change the magnitude of this process through regulating the levels of NMII activation and redistribution by Opa and pili phase variation .
We utilized human endocervical tissue explants and the polarized human colonic epithelial cell line T84 to determine whether GC-infected polarized epithelial cells undergo exfoliation . Tissue explants that were cultured with the mucosal side up and T84 cells that were polarized on transwells were inoculated apically with a GC strain , MS11 that express phase variable Opa and pili ( MS11Pil+Opa+ ) at a MOI of ~10 for 6 or 24 h . Thin sections of cryo-preserved endocervical tissues and T84 cells were stained with a DNA dye and GC-specific polyclonal antibodies and analyzed using and three-dimensional confocal fluorescence microscopy ( 3D-CFM ) . Images showing both the mucosal and subepithelial sides of the endocervix and T84 monolayers were analyzed . Epithelial cells at the top of the endocervical epithelium of tissue explants or T84 monolayers , indicated by white lines , were counted as exfoliating cells ( Fig 1A ) and quantified as the percentage of total epithelial cells . After 24 h incubation , the exfoliation of GC-inoculated epithelial cells was significantly increased in both the endocervical epithelium ( Fig 1B and 1C ) and the T84 monolayer ( Fig 1D and 1E ) , compared to uninfected controls . This indicates that polarized T84 monolayers behave similarly to the endocervical epithelium upon GC infection . There was no significant increase in the percentage of GC-inoculated epithelial cells exfoliated from T84 monolayer after 6-h inoculation , compared to uninfected cells ( Fig 1E , left panel ) . To determine if Opa has a role in the exfoliation of columnar epithelial cells , we inoculated endocervical tissue explants and polarized epithelial cells with MS11Pil+ΔOpa , a GC strain where all 11 opa genes were deleted [39] . MS11Pil+ΔOpa increased the percentage of epithelial exfoliation from 32 . 3% to 66 . 3% in tissue explants ( Fig 1B and 1C ) and from 31 . 2% to 55 . 8% in T84 monolayers ( Fig 1D and 1E ) . Even at 6 h , MS11Pil+ΔOpa-infected T84 cells exfoliated significantly more than the uninfected control ( Fig 1D and 1E ) . To determine whether different Opa variants have similar effects on the exfoliation of endocervical epithelial cells , we utilized MS11Pil+ΔOpa strains that express invariant OpaH ( binding to CEACAMs ) or OpaC ( binding to HSPG ) . We found that the exfoliation level of MS11Pil+OpaH-inoculated endocervical tissue explants was as low as that of MS11Pil+Opa+ infected explants , while the exfoliation level of MS11Pil+OpaC-inoculated explants was as high as that of MS11Pil+ΔOpa-infected explants ( Fig 1C ) . These results indicate that GC induces the epithelial exfoliation from the endocervix and cell line-formed polarized monolayers , and the expression of CEACAM-binding OpaH but not HSPG-binding OpaC inhibits the exfoliation . The similar inhibitory effect of MS11Opa+ and MS11OpaH on epithelial exfoliation suggests that MS11Opa+ expresses primarily CEACAM-binding Opa proteins . To determine if GC-induced exfoliation of endocervical epithelial cells depends on NMII , we inhibited the activation of NMII using inhibitors specific for Rho-associated kinase ( ROCK ) , Y27632 , and myosin light chain kinase ( MLCK ) , ML-7 and PIK . Polarized T84 cells and human endocervical tissue explants were treated with individual inhibitors for 1 h before and during incubation with GC . We found that both the small chemical inhibitor ( ML-7 ) [53] and the catalytic site-targeted peptide inhibitor ( PIK ) [54] of MLCK reduced the exfoliation of MS11Pil+ΔOpa-infected ( Fig 2A and 2B ) but not MS11Pil+Opa+-infected epithelial cells from T84 monolayers ( S1A and S1B Fig ) . In contrast , treatment with the ROCK inhibitor did not significantly change the percentage of epithelial exfoliation , no matter if epithelial cells were infected with MS11Pil+Opa+ ( S1A and S1B Fig ) or MS11Pil+ΔOpa ( Fig 2A and 2B ) . Importantly , the treatment of MLCK inhibitor , ML-7 or PIK , also decreased the epithelial exfoliation of human endocervical tissue explants to the basal level no matter if it was based on the total number of epithelial cells ( Fig 2C and 2D ) or GC-associated epithelial cells ( S1C Fig ) . As MLCK activation requires Ca2+-bound calmodulin [55 , 56] and the MLCK inhibitor PIK blocks the calmodulin-binding site in MLCK [54] , we investigated if GC-induced exfoliation of polarized epithelial cells depends on Ca2+ flux . We utilized 2APB , an inhibitor that blocks Ca2+ release from intracellular stores [57 , 58] . Treatment with 2APB also reduced the exfoliation of polarized T84 cells to the level similar to ML-7 and PIK ( Fig 2A and 2B ) . As controls , we treated polarized T84 cells with the inhibitors alone , and found that ML-7 , PIK , and 2APB did not affect epithelial exfoliation , but the NMII motor inhibitor blebbistatin increased epithelial exfoliation without GC inoculation ( S2 Fig ) . These results suggest that GC induce exfoliation of polarized epithelial cells via Ca2+- and MLCK- but not ROCK-dependent activation of NMII . We have previously shown that GC can transmigrate across polarized epithelial cells , and Opa expression inhibits the transmigration [38] . To determine whether such transmigration occurs in the endocervical epithelium and which Opa variant inhibits GC transmigration , we utilized the tissue explants and GC strains expressing single invariable Opa . After incubating with piliated GC for 24 h , we examined GC transmigration across the endocervical epithelium by quantifying the percentage of GC-associated endocervical epithelial cells with GC staining in the basal side ( Fig 3A ) . In infected endocervical tissue explants , the percentages of epithelial cells associated with penetrated MS11ΔOpa and MS11OpaC were significantly higher than those with penetrated MS11Opa+ and MS11OpaH ( Fig 3B ) . However , there was no significant difference between the percentages of epithelial cells with penetrated MS11ΔOpa and MS11OpaC and between those with penetrated MS11Opa+ and MS11OpaH ( Fig 3B ) . These results indicate that GC can penetrate into the subepithelium of the human endocervix in the tissue explant model , and the expression of CEACAM-binding OpaH , which reduces GC-induced epithelial exfoliation , but not HSPG-binding OpaC , which does not affect the exfoliation , inhibits GC penetration . We next asked whether Opa-mediated inhibition of penetration is related to the efficiency of GC adherence using polarized human endometrial epithelial cells , HEC-1-B ( Fig 3C and 3D ) , and T84 monolayers ( Fig 3E and 3F ) . Similar to what we observed in the endocervical tissue , the numbers of MS11Pil+ΔOpa transmigrating across polarized HEC-1-B ( Fig 3C ) and T84 monolayers ( Fig 3E ) were significantly higher than those of MS11Pil+Opa+ . However , the expression of Opa had no significant effect on GC adherence to the apical surface of HEC-1-B ( Fig 3D ) and T84 cells ( Fig 3F ) . Pili have been shown to be involved in GC transmigration across polarized epithelial cells [59 , 60] . To examine the relationship between pili and Opa , we compare the transmigration and adherence efficiencies of piliated and non-piliated MS11Opa+ and MS11ΔOpa in polarized T84 cells ( Fig 3E and 3F ) . We found that the numbers of non-piliated MS11Opa+ and MS11ΔOpa that adhered to and transmigrated across epithelial monolayers were significantly lower than their piliated strains . However , Opa expression only reduced the transmigration of piliated but not non-piliated GC ( Fig 3E ) . These results suggest that pili and Opa play opposing roles in GC transmigration , with pili promote GC transmigration , probably by enhancing adherence , and Opa inhibiting GC transmigration without affecting GC adherence . The inhibitory effects of CEACAM-binding Opa on both GC-induced epithelial exfoliation and GC penetration in the endocervical tissue explants implicate a relationship between these two events . To investigate this relationship , we determined whether inhibiting GC-induced exfoliation would affect the ability of GC to adhere to , invade into , and transmigrate across polarized epithelial cells . Inhibition of GC-induced exfoliation by the Ca2+ ( 2APB ) and MLCK inhibitors ( ML-7 and PIK ) significantly reduced the transmigration of MS11Pil+ΔOpa across the polarized T84 monolayer ( Fig 4A ) . However , none of these inhibitors had any significant effect on the adherence and invasion of MS11Pil+ΔOpa ( Fig 4B and 4C ) . The ROCK inhibitor that did not affect GC-induced exfoliation also had no impact on GC adherence , invasion and transmigration ( Fig 4A–4C ) . Treatment with the inhibitors alone did not significantly affect the barrier function of the epithelium and GC growth except that treatment of PIK longer than 6 h reduced the overall yield of gonococci to one-half ( S3 Fig ) . Similar to the results obtained from polarized T84 cells , treatment with either ML-7 or PIK decreased the percentage of epithelial cells with basally associated GC among the total GC-associated epithelial cells from 27% to 7 . 1% ( Fig 4D ) , significantly inhibiting GC penetration into the endocervical epithelium . Our results suggest that Ca2+ flux and the activation of NMII by MLCK in polarized epithelial cells , which are required for GC-induced epithelial exfoliation , also are critical for GC transmigration across and penetration into the human endocervical epithelium , but not for GC adherence and invasion . The linkage between the efficiency of GC penetration into the epithelium with GC-induced epithelial exfoliation and apical junction disruption shown here and previously [51] implicate GC-induced junction disruption as an underlying cause of epithelial exfoliation . To test this hypothesis , we determined whether Opa expression and the MLCK and Ca2+ inhibitors , which all inhibited GC-induced epithelial exfoliation , also prevent GC from disrupting the apical junction . The structural integrity of the apical junction was evaluated by analyzing the distribution of E-cadherin using immunofluorescence ( IFM ) and 3D-CFM and quantifying the fluorescence intensity ratio ( FIR ) of E-cadherin at the cytoplasm to that at the apical junction . In polarized T84 cells that were not inoculated with GC , E-cadherin staining was primarily localized at the apical junction ( Fig 5A , top panels ) . Incubation with GC changed the continuous E-cadherin staining at the apical junction into puncta in the cytoplasm , indicating endocytosis of E-cadherin ( Fig 5A ) . This led to a significant increase in the cytoplasm: junction FIR of E-cadherin in both MS11Pil+Opa+ and MS11Pil+ΔOpa-inoculated epithelial cells , compared to non-inoculated controls ( Fig 5B ) . In particular , the magnitude of the increase in the FIR was significantly greater in MS11Pil+ΔOpa-infected than MS11Pil+Opa+-infected epithelial cells ( Fig 5B ) . Our Western blot analysis did not find any significant changes in the protein level of the apical junctional protein ZO1 between epithelial cells inoculated with piliated MS11Opa+ , MS11ΔOpa , and no GC ( Fig 5C ) . These results suggest that Opa expression suppresses GC-induced apical junction disassembly by inhibiting E-cadherin endocytosis . We used inhibitors to determine the role of NMII and Ca2+ flux in GC-induced junction disassembly . Treatment with the MLCK inhibitor ML-7 and the Ca2+ inhibitor 2APB , but not the ROCK inhibitor Y27632 , decreased the punctate staining of E-cadherin in the cytoplasm and the cytoplasm: junction FIR of E-cadherin to or below the control level in epithelial cells without GC inoculation ( Fig 5A and 5B ) . Thus , Ca2+/MLCK inhibitors suppress GC-induced junction disassembly . Our previous studies show that GC-induced junctional disassembly leads to a significant increase in the lateral diffusion between the apical and basolateral membrane but not in the permeability of epithelial monolayers [51] . To determine whether Opa , MLCK and Ca2+ flux are involved in this functional alteration of the apical junction , we stained the basolateral surface of polarized T84 epithelial cells exclusively with CellMask dye for 15 min , after apical incubation with fluorescently labeled piliated GC for 6 h . The appearance of basolaterally stained CellMask dye in the apical membrane indicates a decrease in the fence function of the apical junction ( Fig 5D ) . In control cells where no GC were added , <10% of the cells showed the CellMask staining at the apical surface . The percentage of cells with basolaterally labeled CellMask reaching the apical surface increased to 19 . 4% when MS11Pil+Opa+ was inoculated and to 62 . 2% when MS11Pil+ΔOpa was inoculated ( Fig 5E and 5F ) . These results indicate that while both Opa+ and ΔOpa GC decrease the fence function of the apical junction , MS11Pil+ΔOpa caused a greater reduction than MS11Pil+Opa+ . Moreover , the treatment with the MLCK inhibitor ML-7 or the Ca2+ inhibitor 2APB significantly lowered the percentage of epithelial cells with the CellMark staining leaked to the apical surface ( Fig 5E and 5F ) , thereby inhibiting the GC-induced fence function reduction . We determined if MS11Pil+ΔOpa can induce the disruption of the apical junction in human endocervical tissue . Sections of uninfected and infected tissue explants were stained for the junctional protein ZO1 , GC , and DNA and analyzed by 3D-CFM ( Fig 5G ) . We quantified junction disruption by determining the percentage of GC-associated epithelial cells that lost continuous apical staining of ZO1 , using 3D reconstituted confocal images ( Fig 5G , right panels ) . After a 24-h incubation with MS11Pil+ΔOpa , ZO1 staining at the apical junction of GC-associated epithelial cells appeared to be reduced ( Fig 5G , left panels ) , and 93 . 2% of GC-associated epithelial cells showed defective ZO1 staining ( Fig 5G , right panels , arrows ) , compared to 16 . 7% of uninfected cells ( Fig 5H ) . In contrast to the recruitment of ZO1 to epithelial cells neighboring exfoliating cells in uninfected monolayers , no accumulation of ZO1 staining was observed around GC-infected exfoliating cells ( Fig 5I , arrows ) . Furthermore , GC inoculation significantly changed the morphology of endocervical epithelial cells , with the cells losing their tall and columnar shape ( Fig 5G , left panels ) . Treatment with the MLCK inhibitor ML-7 restored both the morphology ( Fig 5G left panels ) and apical distribution of ZO1 ( Fig 5H ) . These data confirm the ability of GC to compromise the apical junction of the endocervical epithelial cells in a NMII-dependent manner in the human tissue explants . These results together show that both Opa expression and Ca2+/MLCK inhibitors suppress GC-induced disruption of the apical junction , indicating that similar to GC-induced epithelial exfoliation , Ca2+ signal and MLCK-mediated NMII activation are required for GC-induced apical junction disruption while Opa expression inhibits the junction disruption . Our finding of that GC induce both epithelial exfoliation and apical junction disassembly in a NMII-dependent manner suggests that GC regulate the activity of NMII in polarized epithelial cells . We examined the cellular distribution of active NMII after 6-h incubation with GC , using antibody specific for phosphorylated myosin light chain ( pMLC ) and 3D-CFM . In uninfected polarized T84 ( Fig 6A ) and HEC-1-B cells ( S4A Fig ) , pMLC was primarily localized at the apical junction . The polarized distribution of pMLC at the apical surface was quantified by the FIR of pMLC at the apical to the lateral ( Apical: Lateral ) membrane areas in individual cells using CFM images scanning across the apical and basolateral surfaces ( Fig 6A–6C ) . The polarized distribution pMLC at the apical junction was quantified by the FIR of pMLC at the junction to non-junction ( Junction: Non-junction ) areas of the apical region using CFM images scanning through the apical junction ( Fig 6D and 6E ) . The apical inoculation of piliated MS11Opa+ and MS11ΔOpa caused significant increases in apical: lateral FIR in both polarized T84 ( Fig 6A–6C ) and HEC-1-B cells ( S4 Fig ) , compared to the no GC control . There were also significant increases in the junction: non-junction FIR in infected polarized T84 cells , compared to non-infected cells ( Fig 6D and 6E ) . Moreover , both the apical: lateral and junction: non-junction FIRs were significantly higher in MS11Pil+ΔOpa-infected than those in MS11Pil+Opa+-infected T84 cells ( Fig 6C and 6E ) , but this difference was not detected in HEC-1-B cells that do not express CEACAMs [61] ( S4 Fig ) . In contrast , the apical: lateral FIR in epithelial cells infected by non-piliated MS11 , no matter if GC expressed Opa or no , were all significantly reduced to a similar level , compared to those infected by piliated MS11 ( Fig 6B ) . We further noticed that NMII at the apical surface appeared to accumulate at GC adherent sites ( Fig 6A , middle panels , arrows ) . To determine if GC inoculation changes the activation level of NMII , we quantified the amount of pMLC and MLC by Western blot . Polarized T84 cells were incubated with or without piliated MS11Opa+ or MS11ΔOpa apically for 6 h before lysis and Western blot analysis . The antibody staining density ratios of pMLC to MLC in MS11Pil+ΔOpa- but not MS11Pil+Opa+-inoculated cells were significantly higher than that in uninoculated epithelial cells ( Fig 6F , top panels , and Fig 6G ) . However , GC inoculation did not significantly change the staining density ratio of MLC to tubulin ( Fig 6H ) . Thus , MS11Pil+ΔOpa , but not MS11Pil+Opa+ , increases the activation level of NMII . To explore the possibility of that GC-induced NMII redistribution occurs in vivo and the role of Opa phase variation , we incubated human endocervical tissue explants with piliated MS11Opa+ , ΔOpa , OpaH , or OpaC for 24 h . Cryo-sections of the endocervical tissue were stained for pMLC , GC and nuclei . In addition to its apical junction localization , pMLC was concentrated at the basal surface of the endocervical epithelial cells contacting with the basal membrane ( Fig 7A , upper panels ) . When incubated with MS11Pil+Opa+ , there was a redistribution of pMLC from the basal to apical surface , resulting in a significant higher apical: lateral FIR in GC-inoculated tissue explants than that in no GC control ( Fig 7 ) . Inoculation of MS11Pil+ΔOpa further increased the apical: later FIR of pMLC , similar to what we observed in polarized T84 ( Figs 7B and 6A–6C ) . Expression of OpaH , but not OpaC , in MS11Pil+ΔOpa reduced the apical: lateral FIR back to the level in MS11Pil+Opa+-infected cells ( Fig 7B ) . Furthermore , pMLC at the apical surface of the endocervical epithelial cells also concentrated at GC adherent sites ( Fig 7A , second row , white arrows ) , but not at the membrane of cells neighboring exfoliating cells ( Fig 7A , second row , orange arrows ) . These observations confirm that GC increase the relative amount of activated NMII at the apical surface of the endocervical epithelial cells in human tissue explants . Our results from both human endocervical tissue explants and polarized epithelial cell lines suggest that GC interactions via pili cause an accumulation of activated NMII at GC adherent sites and the apical membrane of columnar epithelial cells , and the expression of CEACAM-binding Opa suppresses the activation and redistribution of NMII . The activation of NMII is mediated by the phosphorylation of MLC by MLCK downstream of Ca2+-activated calmodulin [62–65] and/or by ROCK downstream of Rho GTPases [62 , 66] . Our findings that GC-induced epithelial exfoliation and apical junctional disruption , as well as GC transmigration , are inhibited by the MLCK and Ca2+ but not ROCK inhibitors suggest that MLCK mediates the activation and redistribution of NMII triggered by GC . We determined the effects of the MLCK and ROCK inhibitors on GC-induced MLC redistribution and phosphorylation using 3D-CFM and Western blot . Our 3D-CFM analysis found that treatment with the MLCK inhibitor ML-7 or PIK significantly reduced both the apical: lateral and junction: non-junction FIRs of pMLC in GC-infected epithelial cells ( Fig 6A and 6C–6E ) , as well as the accumulation of pMLC at GC adherent sites ( Fig 6A , bottom panels , arrows ) . However , treatment with the ROCK inhibitor Y27632 further increased the junction: non-junction FIR of pMLC in MS11Pil+ΔOpa-inoculated epithelial cells , while having similar inhibitory effects as the MLCK inhibitors on the apical: lateral FIR of pMLC ( Fig 6A and 6C–6E ) . Our Western blot analysis showed that treatment with either the MLCK or the ROCK inhibitor reduced the pMLC: MLC but not the MLC: tubulin density ration in MS11Pil+ΔOpa-inoculated epithelial cells to basal levels ( Fig 6F–6G ) . Moreover , the MLCK inhibitors ML-7 and PIK significantly reduced the apical: lateral FIR of pMLC ( Fig 7 ) and pMLC accumulation at GC adherent sites ( Fig 7A , white arrows ) in MS11Pil+ΔOpa-inoculated endocervical tissue explants . These results suggest that both MLCK and ROCK are involved in the activation of MLC phosphorylation induced by MS11Pil+ΔOpa , but MLCK and ROCK distinctly regulate the subcellular location of active NMII with MLCK promoting and ROCK inhibiting the accumulation of active NMII to the apical junction . A major upstream signaling molecule of MLCK is calmodulin that is activated by Ca2+ [62–65] . To investigate if Ca2+ is involved in GC-induced redistribution of active NMII , we determined if GC inoculation would induce Ca2+ flux in polarized epithelial cells . We used two Ca2+ indicators , FluoForte ( Fig 8A–8C ) and Fluo-4 ( S5 Fig ) to determine the cytoplasmic Ca2+ level . Polarized T84 ( Fig 8A–8C , S5A and S5B Fig ) and HEC-1-B ( S5C and S5D Fig ) were incubated apically with piliated or non-piliated MS11Opa+ or MS11ΔOpa for 4 h . The cells were then loaded with the fluorescent Ca2+ indicator , and the cell membrane marked by the membrane dye CellMask . Cells were imaged using 3D-CFM ( Fig 8A , S5A and S5C Fig ) . The mean fluorescence intensity ( MFI ) of the Ca2+ dyes in individual cells was measured to estimate the cytoplasmic level of Ca2+ ( Fig 8B and 8C , S5B and S5D Fig ) . Compared to uninoculated cells , polarized T84 cells and HEC-1-B inoculated with either MS11Pil+Opa+ or MS11Pil+ΔOpa exhibited significant increases in the MFI of both FluoForte ( Fig 8B ) and Fluo-4 ( S5B and S5D Fig ) . The MFIs of both the Ca2+ indicators in MS11Pil+ΔOpa-inoculated epithelial cells were significantly higher than those in MS11Pil+Opa+-inoculated cells ( Fig 8B , S5B and S5D Fig ) . However , the MFI of the Ca2+ indicator in epithelial cells inoculated with MS11Pil-ΔOpa , was significantly reduced compared to those infected by MS11Pil+ΔOpa , but similar to those infected by MS11Pil+Opa+ ( Fig 8B ) . Treatment with the inhibitor specific for Ca2+ release from intracellular storages 2APB or the intracellular Ca2+ chelator BAPTA brought the MFI of the Ca2+ indicators in both MS11Pil+Opa+ and MS11Pil+ΔOpa-inoculated polarized epithelial cells back to the basal level as seen in uninoculated cells ( Fig 8C , S5B and S5D Fig ) . These results suggest that GC interacting with the apical surface of polarized epithelial cells increases the cytoplasmic level of Ca2+ , by opening the intracellular Ca2+ storages . Opa inhibits and pili may facilitate GC-induced Ca2+ flux . To determine whether GC-induced redistribution of active NMII depends on Ca2+ , we treated polarized T84 cells with the Ca2+ inhibitor 2APB or BAPTA before and during incubation with piliated GC . Both inhibitors decreased the apical: lateral ( Fig 8E ) and the junction: non-junction FIRs ( Fig 8F ) of pMLC in both MS11Pil+Opa+- and MS11Pil+ΔOpa-inoculated polarized epithelial cells , as well as the accumulation of pMLC at GC adherent sites ( Fig 8D , right panels ) . The results in this and previous sections together suggest that Ca2+-dependent activation of MLCK is responsible for GC-induced accumulation of active NMII at the apical junction and GC adherent sites .
A primary challenge in our understanding of GC pathogenesis in the FRT is to mechanistically explain why a small percentage of GC infections lead to invasive diseases while the rest of the infections remain localized . A major research obstacle is a lack of infection models that mimic the anatomic environment and process of GC infection in vivo . This study utilized our newly established human endocervical tissue model with the support of the traditional polarized epithelial cells . Our results demonstrate that GC can penetrate into the subepithelium of the endocervix , and the efficiency of GC penetration is regulated by Opa phase variation . GC enter the subepithelium by disassembling the apical junction and inducing the exfoliation of polarized columnar epithelial cells . These events are caused by the elevation of the cytoplasmic Ca2+ level and the activation and reorganization of NMII in epithelial cells . The expression of CEACAM-binding Opa inhibits GC penetration by suppressing NMII activation and redistribution , as well as Ca2+ flux , thereby limiting GC-induced junction disruption and exfoliation of polarized endocervical epithelial cells . Epithelial exfoliation serves as a protective mechanism of the host as the process sheds off host cell-associated pathogens . Muenzner et al . have shown that GC-induced exfoliation of non-polarized human cervical epithelial cells and squamous epithelial cells in the lower genital tract of the female mice reduces GC colonization [44 , 45] . In contrast , we show here that the exfoliation of columnar endocervical epithelial cells does not affect GC adherence and invasion , rather it allows for an increase in GC penetration into the subepithelium of the human endocervix . These conflicting results , observed in different infection models , can be explained by differences between the types of epithelial cells and the mechanisms underlying GC- and apoptosis-induced exfoliation . A primary difference between multilayered squamous and monolayered columnar epithelial cells is the number of epithelial cell layers . Multilayered squamous epithelial cells can shed in layers with new layers growing underneath [67] , while monolayered columnar epithelial cells exfoliate individual cells to protect the integrity of the epithelial barrier [23 , 68 , 69] . In line with previous findings , epithelial cells in our human endocervical tissue model shed primarily as individuals , rarely as a layer . Based on these observations , we can reason that individual cell shedding from the monolayered endocervical epithelium has a much less impact on GC colonization than the shedding of epithelial layers from multilayered squamous cells . The second major difference between squamous and columnar epithelial cells is the cell-cell junction , the former by adherens junctions ( E-cadherin-E-cadherin interactions ) and focal adhesion ( integrin-extracellular matrix interactions ) [70–72] and the latter by apical junctions ( consisting of tight and adherens junctions ) and desmosomes [73 , 74] . Different junctional complexes suggest that GC require two distinct mechanisms to regulate exfoliation and infection in the two types of epithelial cells . In support of this hypothesis , GC-induced exfoliation of squamous epithelial cells has been shown to be regulated through CD105-dependent activation of integrin [44 , 45] . Here we show that GC induce the exfoliation of the columnar endocervical epithelial cells by disrupting the apical junction through reorganizing its actomyosin support . While integrins may also be involved in the exfoliation of columnar epithelial cells as they mediate the interaction between epithelial cells and the basal membrane , the apical junction plays the essential role in holding and sealing the columnar epithelial monolayer [75] . GC-induced apical junction disruption weakens the barrier function of the epithelium , thereby allowing the penetration of GC into the endocervical subepithelium . In contrast , the exfoliation of squamous epithelial cells may not significantly impact the barrier function of the epithelium due to the presence of additional cell layers . These data together provide explanations for clinical observations that GC rarely cause symptomatic vaginitis [76] but can be found in the subepithelium of the endocervical biopsies from GC-infected women [77] . Whether GC-induced junctional disruption is sufficient to allow for GC to penetrate through the paracellular space of the endocervical epithelium remains a challenging question . We have previously shown that GC-induced junction disruption does not significantly increase the diffusion of soluble fluorescent dyes from the apical to basolateral chambers of well-polarized T84 cells , but it causes an increase in the lateral mobility of the plasma membrane over the apical junction , leading to reductions in cell polarity [51] . Here we confirm that GC-induced junctional disassembly also leads to a decrease in the polarity of endocervical epithelial cells in human tissues , supporting the existence of such an event in human infection . Our studies demonstrate a causative relationship between apical junction disruption and epithelial exfoliation , as the pharmacological inhibition of Ca2+ flux and NMII activation and the natural expression of Opa suppress both GC-induced apical junction disassembly and exfoliation of polarized epithelial cells . This is in sharp contrast to the exfoliation process involved in the columnar epithelial renewal and wound repair , where the apical junction remains intact [23 , 68] . While the precise mechanism underlying the normal exfoliation of endocervical epithelial cells has not been examined , previous studies have shown that the barrier function of the epithelial monolayer is maintained by recruitment of apical junctional proteins and actomyosin to the plasma membrane of cells that neighbor the exfoliating cell [23 , 25 , 68 , 78] . Actomyosin-generated forces in neighboring cells probably push exfoliating cells out while junctional proteins maintain the epithelial barrier . However , in GC-infected cells , the junctional proteins ZO1 and E-cadherin are redistributed from the apical junction to the cytoplasm and intracellular vesicles respectively , and NMII is recruited to GC adherent sites and the apical junction , but not to the plasma membrane of cells neighboring exfoliating cells . As over-activation of NMII in perijunctional actomyosin rings can induce the disassembly and internalization of junctional proteins [15] , our data suggest that the exfoliation of GC-infected endocervical epithelial cells is induced by GC actively via the reorganization and activation of their NMII , modifying the normal exfoliation process to facilitate GC penetration into the subepithelium . A significant finding of this study is that NMII is a target of GC to induce the disassembly of the apical junction and the exfoliation of the endocervical epithelial cells . When the activation and redistribution of NMII are inhibited by pharmacological reagents or Opa expression , GC-induced disassembly of the apical junction and exfoliation are reduced . While the essential role of NMII in the apical junction prohibits us from using a knockdown approach , the catalytic site-specific peptide inhibitor of MLCK PIK [54] and the inhibitory effect of naturally expressed Opa have confirmed the results . Our analysis further suggests that GC-induced NMII activation that leads to exfoliation and GC transmigration primarily depends on MLCK- rather than ROCK-mediated phosphorylation of MLC . Even though both MLCK and ROCK contribute to the phosphorylation of MLC , the two differentially regulate the distribution of active NMII , as the ROCK inhibitor further increases while the MLCK inhibitors reduce the level of active NMII in the apical junction . This result suggests that the subcellular location rather than the level of NMII activation is important for GC infection . Using piliated and non-piliated MS11ΔOpa , we found that pili promote but Opa inhibits GC penetration into the endocervical epithelium and GC-induced columnar epithelial exfoliation by enhancing or suppressing the activation and redistribution of NMII and Ca2+ flux . While the promoting effects of pili are associated with an increase in GC adherence , the inhibitory effects of Opa is independent of GC adherence and invasion . Using MS11ΔOpa expressing a single Opa that cannot undergo phase variation , we showed that the inhibitory effects of MS11Opa+ are mediated by CEACAM-binding Opa but not HSPG-binding Opa . These suggest that CEACAM-binding Opa targets to NMII in columnar epithelial cells , different from its target reported in squamous epithelial cells . Thus , the same Opa may use different mechanisms to suppress the exfoliation of polarized endocervical and non-polarized ectocervical epithelial cells . The inhibitory effect of Opa on GC-induced Ca2+ flux supports that the cytoplasmic Ca2+ is a target of CEACAM-binding Opa proteins to regulate NMII [56] . However , the exact underlying mechanism is unknown . The inhibitory effects of OpaH may be mediated through engaging CEACAMs [79] , which function to enhance cell-cell adherence and suppress cell signaling [80 , 81] . This study found that MS11ΔOpa induced a significantly higher level of pMLC accumulation in the apical junction in CEACAM-expressing T84 cells than MS11Opa+ ( Fig 6E ) , but increased the pMLC accumulation to a level similar to MS11Opa+ in HEC-1-B cells that do not express CEACAMs [61] ( S4 Fig ) . These results support the hypothesis that CEACAMs are involved in suppressing GC-induced NMII redistribution . While potentially involved in the exfoliation of both squamous and columnar epithelial cells , CEACAMs may differentially modulate signaling induced by GC , due to distinct distributions of CEACAMs and organizations of signaling , cytoskeleton , and cell-cell junctions in the two types of epithelial cells . Our results also show that GC-induced NMII activation depends on Ca2+ flux that can activate MLCK via calmodulin [56] . In addition to Ca2+ flux , CEACAM-binding Opa proteins can potentially inhibit GC-induced NMII activation in a Ca2+-independent manner , such as by activating MLC phosphatase that dephosphorylates MLC [82] or through modulating GC-epithelial physical interactions . By interacting with CEACAMs on epithelial cells and LOS on neighboring GC , Opa potentially alters the physical tensions that GC exert onto the mucosal surface , consequently changing the organization of NMII beneath the plasma membrane of epithelial cells [83 , 84] . Since different isoforms of the 11 Opa proteins have different binding abilities to CEACAMs [85] , they can modulate GC-epithelial interactions and signaling distinctly when Opa undergoes phase variation . This study is the first to utilize human endocervical tissue explants to examine the mechanism by which GC establish infection in this in vivo location . Our results have extended our mechanistic understanding of GC pathogenesis in the context of human infection . Based on these results , we propose the follow working model for GC infection in the human endocervix ( Fig 9 ) . Pili-initiated interactions of GC with the apical surface of the human endocervical epithelial cells induce Ca2+- and MLCK-dependent activation and redistribution of NMII to GC adherent sites . NMII activation and remodeling cause the disassembly of the apical junction that holds the epithelium together and seals the paracellular space . Apical junction disruption leads to epithelial exfoliation and GC penetration into the endocervical epithelium . Phase variation of Opa from HSPG-binding to CEACAM-binding isoforms inhibits GC-induced Ca2+ flux and NMII activation and redistribution , consequently reducing GC-induced epithelial exfoliation and GC penetration . The accumulated data from previous and current studies indicate that GC manipulate the epithelial barrier by regulating host cell signaling and cytoskeletal systems for their infection . The nature and level of GC-mediated manipulation are modulated by phase variation of GC surface molecules and types of epithelial cells that GC interact with , which enable GC to infect various regions of the FRT and generate different infection outcomes .
N . gonorrhoeae strain MS11 that expressed both pili and Opa ( MS11Pil+Opa+ ) was obtained from Dr . Herman Schneider , Walter Reed Army Institute for Research . Derivatives of this strain , MS11ΔOpa , MS11OpaH ( CEACAM-binding ) , and MS11OpaC ( HSPG-binding ) have previously been described [38 , 39] . MS11Opa+ and MS11ΔOpa strains express similar LOS [39] . MS11 Pil+Opa+ and Pil-Opa+ colonies were identified based on their morphology using a dissecting light microscope . Our previous sequencing analysis showed that they expressed different pilE variants [38] . GC were grown on plates with GC media ( Difco , BD Bioscience , Franklin Lakes , NJ ) and 1% Kellogg’s supplement [86] for 16–18 h before inoculation . The concentration of GC in suspension was determined using a spectrophotometer and inoculated at MOI 10:1 . Human colorectal carcinoma cell line , T84 ( ATCC , Manassas , VA ) , was maintained in DMEM:Ham F12 ( 1:1 ) supplemented with 7% heat-inactivated fetal bovine serum ( FBS ) . Human endometrial adenocarcinoma cell line , HEC-1-B ( ATCC ) , was maintained in Eagles MEM alpha medium supplemented with 10% heat-inactivated FBS . Cells were seeded at 6x104 per transwell ( 6 . 5 mm diameter and 3 μm pore size , Corning , Corning , NY ) and cultured for ~10 days until transepithelial electrical resistance ( TEER ) reached >1000 Ω ( T84 ) or >300 Ω ( HEC-1-B ) . TEER was measured using a Millicell ERS volt-ohm meter ( EMD Millipore , Billerica , MA ) . The tissue explants were cultured as previously described [87] . Endocervical tissues were obtained from patients undergoing voluntary hysterectomies and received within 24 h post-surgery . Samples were cut into ~2 . 5 cm ( L ) X 0 . 6 cm ( W ) X 0 . 3 cm ( H ) pieces , incubated in CMRL-1066 ( GIBCO , Gaithersburg , MD ) plus antibiotics for 24 h , and then switched to antibiotic-free media for 24 h . Individual endocervical tissue pieces were incubated with GC at a MOI of ~10 ( based on the average number of endocervical epithelial cells in endocervical tissue pieces ) in the presence or absence of the MLCK inhibitor ML-7 ( 10 μM , EMD Millipore ) and PIK ( Dreverse , 100 μM ) [54] on transwells for 24 h . Unassociated GC were removed by extensive washes at 6 and 12 h . The tissue was fixed , embedded in gelatin , cryopreserved , sectioned crossing the apical and basolateral surfaces of the epithelium , stained for ZO1 ( BD Bioscience ) , pMLC ( Cell Signaling Technology , Beverly , MA ) , and GC [51] by specific antibodies , and nuclei by DAPI ( Life Technologies , Carlsbad , CA ) , and analyzed using CFM ( Zeiss LSM 710 , Carl Zeiss Microscopy LLC , Oberkochen , Germany ) . Images were acquired as Z-series of 0 . 37 μm slices , and 3D composites obtained using the NIH ImageJ software . To quantify epithelial exfoliation , the number of epithelial cells localized on the top of the monolayer ( exfoliated ) and the total number of epithelial cells were counted by visual inspection in each randomly acquired image , to determine the percentage of exfoliated cells . To estimate the level of GC penetration into the subepithelium , the number of non-exfoliated epithelial cells ( in clearly visible epithelial monolayers ) with basal GC staining and the total number of GC-associated epithelial cells were visually counted to calculate the percentage of infected epithelial cells with GC penetration into the basolateral side . To evaluate the disruption of the apical junction , the number of GC-associated epithelial cells that lost continuously apical staining of the junctional protein ZO1 and the total number of GC-associated epithelial cells were visually counted to calculate the percentage of GC-associated epithelial cells with apical junction disassembly . To quantify the redistribution of pMLC , the fluorescence intensity ratios ( FIR ) of pMLC staining at the apical to lateral in individual epithelial cells were determined using average FI as previously described [51] . Cells were pretreated with or without NMII kinase inhibitors , Y27632 ( 10 μM , EMD Millipore ) , ML-7 ( 10 μM , EMD Millipore ) and PIK ( 100 μM ) or Ca2+ inhibitors , 2APB ( 10 μM , EMD Millipore ) and BAPTA ( 50 μM , EMD Millipore ) for 1 h , and incubated with GC in the presence or absence of the inhibitors for 6 h . Cell were washed and fixed with 4% paraformaldehyde , permeabilized with 1% Triton X100 , and stained with anti-E-cadherin ( BD Bioscience ) , anti-pMLC ( Cell Signaling Technology , Carlsbad , CA ) , anti-GC antibodies , and DAPI for nuclei . Cells were analyzed by CFM . Images were acquired as Z-series of 0 . 37 μm slices , and 3D composites obtained . Epithelial exfoliation will be quantified using xz images as described for the tissue explants . The distribution of E-cadherin and pMLC was quantitatively analyzed by measuring the FIR at the apical junctional to the cytoplasmic area ( from xy images ) or at the apical to lateral surface area ( from xz images ) in individual cells . Polarized T84 cells were incubated with GC apically for 6h . Then the cells were incubated with the CellMask dye ( 5μg/ml , Life Technologies ) in the basolateral chamber only for 15 min , and xz images were acquired using Leica TCS SP5 X confocal microscope ( Leica Microsystems , Buffalo Grove , IL ) . The number of epithelial cells displaying CellMask staining at the apical membrane was countered visually as the percentage of the total number of epithelial cells in each randomly acquired image . Polarized T84 cells , apically incubated with GC for 6 h with or without inhibitors , were lysed by RIPA buffer [0 . 1% triton × 100 , 0 . 5% deoxycholate , 0 . 1% SDS , 50 mM Tris-HCl pH 7 . 4 , 150 mM NaCl , 1 mM EGTA , 2 mM EDTA , 1 mM Na3VO4 , 50 mM NaF , 10 mM Na2PO4 , and proteinase inhibitor cocktail ( Sigma-Aldrich , St . Louis , MO ) ] . Lysates were resolved using SDS-PAGE gels ( BioRad , Hercules , CA ) and analyzed by Western blot . Blots were stained for pMLC or MLC ( Cell Signaling Technology ) , stripped , and reprobed with anti-β-tubulin antibody ( Santa Cruz , Santa Cruz , CA ) . Blots were quantified using a Fujifilm LAS-3000 ( Fujifilm Medical Systems U . S . A . , Inc . , Stamford , CT ) . The assays were performed as previously described [51] . Briefly , polarized epithelial cells that were pretreated with or without the inhibitors for 1 h were incubated apically with GC at 37°C for 3 h for adherence and 6 h for invasion and transmigration assays with or without the inhibitors . GC in the basolateral media were cultured and counted as transmigrated bacteria . Cells were washed and lysed to count adherent GC . Cells were treated with gentamicin , washed , and lysed to count bacteria that were resistant to gentamicin treatment as invaded GC . T84 cells were seeded at 1×105 per transwell on the underside of transwells [88] and cultured for ~10 days until TEER reached >1000 Ω . Cells were pre-treated with or without the Ca2+ inhibitors , 2APB ( 10 μM , Sigma , Saint Louis , MO ) and BAPTA ( 50 μM , Sigma ) , for 1 h and incubated with GC ( MOI of 10 ) apically in the presence or absence of the inhibitors for 4 h . Then cells were incubated with the fluorescent Ca2+ indicator Fluoforte ( 100 μg/ml , Enzo Life Sciences , Farmingdale , NY ) or Fluo-4 ( 100 μM , Life Technologies ) for 1 h . Confocal xz images were acquired in the presence of the membrane dye CellMask ( 5 mg/ml , Life Technology ) using Leica TCS SP5X confocal microscope ( Leica Microsystems , Buffalo Grove , IL ) , based on the instruction by manufacturers . To quantify the intracellular Ca2+ level , the cytoplasmic region of individual cells was manually selected based on the CellMask staining in randomly acquired confocal images , and the mean fluorescent intensity ( MFI ) of Fluoforte and Fluo-4 in the cytoplasmic region was measured using the NIH ImageJ software . Statistical significance was assessed using the Student’s t-test and one-way ANOVA by Prism software ( GraphPad Software , La Jolla , CA ) . Human cervical tissue was obtained from National Disease Research Interchange ( NDRI , Philadelphia , PA ) . Human cervical tissues used were anonymized . The usage of human tissues has been approved by the Institution Review Board of the University of Maryland .
|
Neisseria gonorrhoeae ( GC ) infects human genital epithelium causing gonorrhea , a common sexually transmitted infection . Gonorrhea is a critical public health issue due to increased prevalence of antibiotic-resistant strains . Because humans are the only host for GC , a lack of a human infection model has been a major obstacle to our understanding of GC infection . Here we use a human tissue explant model to examine the mechanism by which GC infect the human endocervix , the primary site for GC infection in women . We show that GC penetrate into the human endocervix by activating the actin motor myosin and epithelial shedding . Myosin activation causes the disruption of the endocervical epithelial barrier by inducing apical junction disassembly and epithelial cell shedding , allowing GC penetration into the human endocervical tissue . GC activate myosin by inducing Ca2+-dependent phosphorylation of myosin light chain . We further show that GC can enhance and reduce the penetration by expressing pili and the opacity-associated protein that promotes and inhibits myosin activation , respectively . Our study is the first demonstration of GC penetration into the human endocervix . Our results provide new insights into the mechanism by which GC manipulate signaling and cytoskeletal apparatus in epithelial cells to achieve penetrating and non-penetrating infection .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"phosphorylation",
"medicine",
"and",
"health",
"sciences",
"membrane",
"staining",
"epithelial",
"cells",
"cytoplasmic",
"staining",
"molecular",
"motors",
"actin",
"motors",
"sexually",
"transmitted",
"diseases",
"motor",
"proteins",
"research",
"and",
"analysis",
"methods",
"specimen",
"preparation",
"and",
"treatment",
"staining",
"infectious",
"diseases",
"contractile",
"proteins",
"animal",
"cells",
"proteins",
"biological",
"tissue",
"biochemistry",
"cytoskeletal",
"proteins",
"cell",
"staining",
"cell",
"biology",
"anatomy",
"post-translational",
"modification",
"epithelium",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"myosins"
] |
2017
|
Neisseria gonorrhoeae infects the human endocervix by activating non-muscle myosin II-mediated epithelial exfoliation
|
Deciphering the mechanisms of regulation of metabolic networks subjected to perturbations , including disease states and drug-induced stress , relies on tracing metabolic fluxes . One of the most informative data to predict metabolic fluxes are 13C based metabolomics , which provide information about how carbons are redistributed along central carbon metabolism . Such data can be integrated using 13C Metabolic Flux Analysis ( 13C MFA ) to provide quantitative metabolic maps of flux distributions . However , 13C MFA might be unable to reduce the solution space towards a unique solution either in large metabolic networks or when small sets of measurements are integrated . Here we present parsimonious 13C MFA ( p13CMFA ) , an approach that runs a secondary optimization in the 13C MFA solution space to identify the solution that minimizes the total reaction flux . Furthermore , flux minimization can be weighted by gene expression measurements allowing seamless integration of gene expression data with 13C data . As proof of concept , we demonstrate how p13CMFA can be used to estimate intracellular flux distributions from 13C measurements and transcriptomics data . We have implemented p13CMFA in Iso2Flux , our in-house developed isotopic steady-state 13C MFA software . The source code is freely available on GitHub ( https://github . com/cfoguet/iso2flux/releases/tag/0 . 7 . 2 ) .
Fluxomics is the omics field that analyses metabolic fluxes ( i . e . , reaction and transport rates in living cells ) which are a close reflection of the metabolic phenotype . As such , quantitative tracking of metabolic fluxes is vital for deciphering the regulation mechanisms of metabolic networks subjected to perturbations , including disease states and drug-induced stress . However , unlike other omics data that can be quantified directly , the fluxome can only be estimated through an indirect interpretation of experimental data[1–3] . There are two main model-based approaches to quantifying metabolic fluxes , Flux Balance Analysis ( FBA ) and 13C Metabolic Flux Analysis ( 13C MFA ) . Both methods use stoichiometric , thermodynamic and experimental constraints to find the range of feasible fluxes across a metabolic network and then find the flux distributions within that space that optimize a given objective function . However , both techniques differ in the type of objective function optimized . In FBA , the objective function is a set of fluxes to be minimized or maximized . These fluxes must represent a biological objective deemed desirable in the conditions of study ( e . g . , synthesis of biomass components for proliferating systems ) [4] . A significant limitation of FBA is that the choice of objective ( s ) can significantly influence the predicted flux distributions . In 13C MFA , the objective function is to minimize the difference between simulated and measured 13C enrichment in metabolites [5 , 6] . 13C enrichment is quantified in metabolic products and intermediates after incubating samples with metabolic substrates labeled with 13C ( tracers ) and provides information about how carbons are redistributed along metabolic pathways[7] . Compared to FBA , 13C MFA has a greater capacity to elucidate the fluxes of central carbon metabolism . However , 13C MFA is more complex to solve than FBA due to the non-linear nature of the 13C MFA objective . A significant limitation of FBA is that there is generally a wide range of optimal flux distributions[8] . This is not usually the case with 13C MFA which can generally determine flux distributions with a high degree of accuracy . 13C MFA achieves this by integrating large sets of measured isotopologue fractions from parallel experiments with tracers optimized for different parts of the network[9–16] . However , when 13C MFA is used in large metabolic networks and with a limited set of measurements , it can also suffer from the same limitation as FBA and result on a wide interval of flux values for part of the metabolic network[5 , 17–19] . On FBA , an approach to reduce the range of optimal solutions consists in running a second optimization step on the optimal solution range . One of such methods is parsimonious FBA ( pFBA ) [20] . This approach , which follows the principle of parsimony or simplicity , consists on finding the optimal value of the primary objective function through FBA and then running a second optimization step where the sum of reaction fluxes is minimized while maintaining the optimal primary objective . The GIMME ( and its derivative GIM3E ) algorithms[21 , 22] are based on a similar principle as pFBA . Unlike standard pFBA , where all reactions fluxes are minimized with equal weight , GIMME integrates gene expression data to give greater weight to the minimization of fluxes through reactions catalyzed by lowly expressed enzymes . Different to FBA , for 13C MFA , there is currently no approach that relies on a second optimization to reduce the solution space when experimental data is insufficient to constrain the system towards a unique solution . In addition to model-based approaches ( e . g . , FBA or 13C MFA ) , metabolic fluxes can also be analyzed through the direct or semidirect interpretation of 13C data . This approach primarily consists of predicting the contribution of a labeled substrate to the synthesis of a given metabolite ( nutrient contribution ) and predicting the relative activity of pathways ( pathway activity analysis ) . Pathway activity analysis assumes that the isotopologue fractions used as a surrogate for the pathways of interest are primarily generated through them . This assumption is generally based on the assertion that the pathways of interest are the most direct way to generate such fractions from the labeled substrate used in the experiment[2 , 7 , 23–25] . Unlike 13C MFA , direct interpretation of 13C data is generally not able to quantify network-wide flux maps . Instead , it provides a series of qualitative or semiquantitative flux predictions around each analyzed metabolite . Strategies that couple direct interpretation of 13C data to regression and correlation analyses are widely applied to unveil the effect of an external perturbation , such as a therapeutic intervention , on central carbon metabolism[26–30] . Here we present parsimonious 13C MFA ( p13CMFA ) , a new model-based approach to flux estimation . p13CMFA first minimizes the difference between experimental and simulated 13C enrichment in metabolites ( 13C MFA ) and then applies the flux minimization principle to select the best solution among the solutions that fit experimental 13C data . Hence , p13CMFA can be used to select the best flux map in instances where experimental 13C measurements are not enough to fully constrain the 13C MFA solution space . Furthermore , the minimization can be weighted by gene expression allowing seamless integration of 13C with gene expression data ( Fig 1 ) . We have implemented p13CMFA in Iso2Flux , our in-house developed isotopic steady-state 13C MFA software ( https://github . com/cfoguet/iso2flux/releases/tag/0 . 7 . 2 ) . As a proof of concept , we have applied it to the analysis of the metabolic flux distribution in HUVECs ( Human umbilical vein endothelial cells ) through the integration of a small set of 13C enrichment measurements and transcriptomics data . Furthermore , we validated the predictive capacity of p13CMFA using data from a published study of HTC116 cells where fluxes had been estimated with a high degree of confidence[14] . Using only a small subset of the measurements from such study , p13CMFA was able to achieve significantly better flux predictions than both 13C MFA and GIMME .
p13CMFA consists of two consecutive optimizations: first , the optimal solution to the 13C MFA problem is identified ( Eq 1 ) ; secondly , the weighted sum of reaction fluxes is minimized within the optimal solution space of 13C MFA ( Eq 2 ) . The 13C MFA optimization ( Eq 1 ) identifies the flux distribution that minimizes the difference between measured and simulated isotopologue fractions [5 , 7]: Xopt=min∑j ( Ej−Yj ( v ) σj ) 2 ( Eq 1 ) SubjecttoS . v=0 , lb≤v≤ub where , v is a vector of flux values describing a valid steady-state flux distribution; Xopt is the optimal value of the 13C MFA objective; Ej is the experimentally quantified fraction for isotopologue j; Yj ( v ) is the simulated isotopologue fraction for isotopologue j with flux distribution v . Such simulation is performed by solving a complex non-linear system of equations built around isotopologues balances [1] . σj is the experimental standard deviation of the measurements of isotopologue j; S is the stoichiometric matrix; lb and ub are vectors defining the upper and lower bounds for flux values . Flux bounds can be used to integrate experimental flux measurements; Either in large metabolic networks or when small sets of 13C measurements are integrated , the 13C MFA problem can be undetermined and there can be a wide range of possible solutions . Such indetermination emerges from cycles and alternative pathways in the metabolic network , which lead to many possible flux combinations that can result in the measured 13C label patterns . Furthermore , many of the 13C MFA solutions can involve large fluxes through futile cycles , which are usually artifacts of the optimization process as in vivo enzyme activities cannot support such large flux values . Therefore , to select the best solution among the many solutions that fit experimental 13C data , p13CMFA runs a second optimization where the weighted sum of fluxes is minimized ( Eq 2 ) : min∑i|vi|⋅wi ( Eq 2 ) subjecttoS . v=0 , lb≤v≤ub , ∑j ( Ej−Yj ( v ) σj ) 2≤Xopt+T where: wi is the weight given to the minimization of flux through reaction i; T is the maximum value that the 13C MFA objective can deviate from its optimal value ( primary objective tolerance ) when fluxes are minimized; The difference between the optimal 13C MFA objective function value and the objective function value when the total reaction flux is minimized can be assumed to follow a Chi2-distribution with one degree of freedom . Therefore , setting T to 3 . 84 gives a p13CMFA solution within the 95% confidence intervals of 13C MFA[5] . With p13CMFA , the activity through cycles is minimized to the minimum amount needed to account for experimental measurements . Furthermore , gene expression measurements can be integrated to give greater weight to the minimization of fluxes through reactions catalyzed by lowly expressed enzymes . Then , in instances where multiple pathways can result in similar label patterns , those pathways with stronger gene expression evidence are selected . Hence , p13CMFA reduces the solution space towards a unique solution without requiring a simplification of the metabolic network or additional 13C measurements ( Fig 1 ) . As an example of a potential application of p13CMFA , we applied it to analyze the metabolic flux distribution in HUVECs using a publicly available dataset not large enough to make meaningful flux predictions with conventional 13C MFA . In this study , available in the MetaboLights repository[31] ( accession number MTBLS412 ) , HUVECs were incubated in the presence of the tracer [1 , 2-13C2]-glucose , and the relative abundance of 13C isotopologues was quantified in glycogen , ribose , lactate , and glutamate . The rates of production/consumption of glucose , glycogen , lactate , glutamate , and glutamine were also quantified . The data were integrated into a stoichiometric model of central metabolism which includes glycolysis , glycogen metabolism , pentose phosphate pathway ( PPP ) , tricarboxylic acid ( TCA ) cycle , fatty acid synthesis , and energy and redox metabolism ( S1 ZIP ) . To predict the flux distribution using conventional 13C MFA , 95% confidence intervals were computed for each predicted flux value . From this analysis , the space of flux solutions consistent with the measured 13C enrichment was estimated . The resulting space of solution was still mostly undetermined and , in general , 13C MFA was unable to significantly constraint the flux ranges emerging from the stoichiometric and thermodynamic constraints and the measured extracellular fluxes ( Fig 2 , S1 Table ) . For instance , despite integrating measurements of 13C enrichment in ribose , it was not possible to conclude whether the oxidative branch of the pentose phosphate pathway contributed more to de novo ribose synthesis than the non-oxidative branch or vice versa . Nevertheless , p13CMFA can be applied to select the best solution in the 13C MFA solution space . With this aim , transcriptomic data taken from the literature[32] were used to add additional penalties to the flux through lowly expressed enzymes . Indeed , by applying p13CMFA , we can now conclude that , under the condition of the study , glucose is mostly directed towards lactate production except for a small part going to the TCA cycle through pyruvate dehydrogenase ( Fig 2 , Fig 3 ) . Glutamine is mainly metabolized to glutamate or directed to glycolysis through the TCA cycle and phosphoenolpyruvate carboxykinase . In the PPP , the non-oxidative branch contributes to roughly 60% of the net ribose synthesis . Only the glycogen phosphorylase/glycogen synthase futile cycle is predicted to be active , while the remaining futile cycles ( i . e . , the hexokinase/glucose 6-phosphatase , phosphofructokinase/fructose bis-phosphatase , pyruvate carboxylase/phosphoenolpyruvate carboxykinase , and glutaminase/glutamine synthase cycles ) are predicted to be inactive . Concerning redox metabolism , most of the reduced NAD+ ( NADH ) produced in the mitochondria is exported to the cytosol through the malate-aspartate shuttle , where it is used to reduce pyruvate to lactate . To evaluate the contribution of 13C MFA to the p13CMFA solution , GIMME ( i . e . , flux minimization weighted by gene expression without integrating 13C data ) was also performed ( Fig 2 , S1 Table ) . Lacking 13C data , GIMME does not predict any activity in the oxidative branch of the pentose phosphate pathway , nor on the glycogen phosphorylase/glycogen synthase futile cycle . Furthermore , GIMME predicts a significantly larger flux through pyruvate dehydrogenase than p13CMFA . Interestingly , p13CMFA predicts an increased activity of the TCA cycle compared to the GIMME solution . This increased activity is fueled by alternative sources of acetyl-CoA such as fatty acid oxidation or catabolism of ketogenic amino acids . Hence , p13CMFA is able to take advantage of measured 13C enrichments and predict significantly different flux maps than those derived from flux minimization alone . To validate the p13CMFA method , we used data from a metabolic characterization of the colon cancer cell line HCT 116 published by Tarrado-Castellarnau et al . [14] . In this study , 25 direct flux measurements and 24 sets of isotopologue fractions , measured after incubation with either [1 , 2-13C2]-glucose or [U-13C5]-glutamine , had been integrated in the framework of 13C MFA . With such a large set of experimental measurements , 13C MFA had been able to estimate the flux through 62 reactions with a high degree of accuracy . In the same study , transcriptomics data were also collected . From this large data set , we selected a partial data set consisting of 7 experimental flux measurements ( the rates of uptake/secretion of glucose , lactate , glutamine , glutamate and , oxygen and the rate of protein and glycogen synthesis ) and 4 sets of isotopologue fractions ( isotopologue fractions in ribose , lactate , glutamate and glycogen measured after incubation with 1 , 2-13C2]-glucose ) . Those are the sets of isotopologues and fluxes that were analyzed in the HUVECs case study with the addition of the rate of protein synthesis and oxygen consumption which Tarrado-Castellarnau et al . described as key determinants of the metabolic phenotype of HCT 116 cells . The partial data set was used to apply pFBA , GIMME , 13C MFA and p13CMFA in the framework of the metabolic network defined by Tarrado-Castellarnau et al . [14] ( S2 Zip ) . p13CMFA was applied both with and without integrating gene expression data ( p13CMFA+ge and p13CMFA-ge , respectively ) . Two complementary metrics , Pearson’s correlation and Euclidian distance , were used to evaluate the similarity between the predicted flux distributions and the flux maps estimated by Tarrado-Castellarnau et al . using the full dataset[14] ( Fig 4 , S2 Table ) . The results show that p13CMFA-ge yields a significantly more accurate flux prediction than both pFBA ( i . e . , flux minimization without integrating 13C data ) , and 13C MFA . Interestingly , while integrating gene expression significantly enhances the accuracy of p13CMFA ( p13CMFA+ge compared to p13CMFA-ge ) , such effect is less marked than the effect of adding gene expression to standard flux minimization ( GIMME compared to pFBA ) . This is due to the fact that p13CMFA-ge flux predictions have already a remarkable level of accuracy; hence , less information can be gained by adding transcriptomics data . Nevertheless , even if GIMME achieves flux predictions of similar accuracy to p13CMFA-ge , p13CMFA+ge results on flux predictions that are significantly more accurate than those obtained with GIMME . Hence , in instances were only a limited number of 13C measurements are available , p13CMFA is a valid method for obtaining accurate flux estimations , regardless of the availability of gene expression data .
13C MFA is a well-established technique and has proven to be an extremely valuable tool in quantifying metabolic fluxes[9–18] . However , to fully determine fluxes through a large metabolic network , parallel labeling experiments must be performed and 13C propagation must be quantified in many metabolites in the network[19] . Indeed , when applying 13C MFA either with a small set of experimental data or with a large metabolic network , part of the 13C MFA solution space can be too wide to draw meaningful conclusions about the underlying flux distribution . This solution space can be reduced by removing degrees of freedom from the system , for instance , by removing reactions from the network or making reactions irreversible . However , this can introduce an arbitrary bias in the resulting flux distribution . Here we describe p13CMFA , a new approach for 13C data integration which can overcome these limitations of 13C MFA and estimate a realistic solution within an undetermined 13C MFA solution space . This solution will be the flux distribution within the 13C MFA solution space that minimizes the weighted sum of reaction fluxes . Thus , it will be the most enzymatically efficient solution . In that regard , p13CMFA is partially based on a similar principle as pathway activity analysis ( i . e . , the assumption that specific fractions of isotopologues are primarily generated through the simplest combinations of pathways ) . However , unlike pathway activity analysis , p13CMFA is able to integrate all quantified isotopologue fractions and flux measurements ( e . g . rates of metabolite uptake and secretion ) to generate network-wide flux maps consistent with such data . Furthermore , p13CMFA is highly flexible; for instance , here we show that it can be used to seamlessly integrate gene expression data by giving higher weight to the minimization of the fluxes through lowly expressed enzymes . As a proof of concept , we exemplified how p13CMFA can be used to estimate flux distributions integrating only limited sets of 13C measurements in a test case where traditional 13C MFA was unable to provide a narrow solution space . Furthermore , we demonstrated that , when a limited set of measurements are integrated , p13CMFA can yield more accurate flux predictions than both 13C MFA and GIMME . p13C MFA does not aim to be a replacement of 13C MFA; instead , it seeks to supplement it by identifying the more straightforward solution in parts of the network that cannot be uniquely determined . In that regard , it can be used to quantitatively study flux distributions in instances where not enough information can be obtained with conventional 13C MFA . Nor does it aim to replace the direct interpretation of 13C data . The latter is still a suitable technique when the goal of the analysis is to compare the relative activity of well-established pathways across conditions or quantify substrate contributions rather than to generate complete flux maps . 13C data has been widely used to assist in drug discovery . In this regard , tracer analysis coupled with regression and correlation analyses is frequently used to characterize drug response [26–29] . Such approach uses regression and correlation statistics with binary , numeric and visual analysis to integrate drug dosage , time points , as well as all necessary biological variables in order to diagnose disturbed stable isotope labeled matrices[29] . p13CMFA could further expand the role of 13C in drug discovery by allowing the integration of 13C and transcriptomic data in the framework of genome-scale metabolic models . In the framework of such models , drug targets are identified by systematically simulating the effect of reactions or genes knock out to cell function[34] . This is usually attained by applying the ROOM[35] or MOMA[36] algorithms , which take a unique flux solution as input ( wild-type flux distribution ) to predict the most likely effect of a gene KO . Hence , p13CMFA results could be potentially used as ROOM/MOMA inputs allowing to take full advantage of the flux information derived from both 13C and transcriptomics data to predict new drug targets . With atom mappings now available on a genome-scale[37] , the main obstacle to applying p13CMFA at a genome-scale is the high computational complexity of solving the resulting non-linear problem which increases with the size of the network . Hence , the next challenge for p13CMFA will be optimizing its implementation for genome-scale networks .
The flux spectrum[38] ( i . e . , the feasible range of fluxes for a given set of stoichiometric , thermodynamic and flux boundary constraints ) was determined using flux variability analysis [8] . Under this approach , each flux is minimized ( Eq 3 ) and maximized ( Eq 4 ) subject to constraints to find the minimum ( vminiFS ) and maximum ( vmaxiFS ) feasible values for each flux: vminiFS=minvi ( Eq 3 ) subjecttoS . v=0 , lb≤v≤ub vmaxiFS=maxvi ( Eq 4 ) subjecttoS . v=0 , lb≤v≤ub The 13C MFA solution space is estimated by computing the confidence intervals for each flux . Such intervals are obtained by minimizing ( Eq 5 ) and maximizing ( Eq 6 ) each flux subject to constraints[5] . vmini=minvi ( Eq 5 ) subjecttoS . v=0 , lb≤v≤ub , ∑j ( Ej−Yj ( v ) σj ) 2≤Xopt+T vmaxi=maxvi ( Eq 6 ) subjecttoS . v=0 , lb≤v≤ub , ∑j ( Ej−Yj ( v ) σj ) 2≤Xopt+T where , vmini: is the lower bound of the confidence interval for flux i with tolerance T; vmaxi: is the upper bound of the confidence interval for flux i with tolerance T; Provided that the same primary objective tolerance ( T ) is used in computing both the p13CMFA solution and the 13C MFA confidence intervals , the p13CMFA solution will always fall within the boundaries of 13C MFA confidence intervals ( vmini≤vi≤vmaxi ) . To apply GIMME and pFBA , the sum of fluxes is minimized subject only to network stoichiometry and flux boundaries ( Eq 7 ) . In GIMME , flux minimization weights are derived from gene expression measurements , whereas in pFBA all reactions are given the same minimization weight[20 , 22] . Transcriptomic data of HUVECs and HCT 116 cells published by Weigand et al . [32] and Tarrado-Castellarnau[14] et al . , respectively , were obtained from the Gene Expression Omnibus repository[39] . A Robust Multichip Analysis gene-level normalization[40] was performed with the Oligo package for R[41] . Using gene-protein-reaction rules , normalized transcript intensities were mapped to each enzyme-catalyzed reaction or protein-facilitated transport process . The weight given to the minimization of fluxes was assigned according to the following equation: wi=1+max ( Th−gei , 0 ) ( Eq 8 ) where , gei is the gene expression value assigned to reaction i; Th is the gene expression threshold . Fluxes through reactions with gene expression levels below this threshold are given additional minimization weight; Using the same criteria as GIM3E[22] , Th was set at the maximum gene expression value found in the set of genes mapped to the metabolic network ( Eq 9 ) : Th=max ( ge ) ( Eq 9 ) Using this threshold , the information gained from integrating available gene expression measurements is maximized . Other Th values were tested in the validation case study[14] and using the maximum gene expression as the threshold was found to yield the most accurate flux predictions ( S3 Table ) . p13CMFA was implemented in Iso2Flux , our in-house developed 13C MFA software ( https://github . com/cfoguet/iso2flux/releases/tag/0 . 7 . 2 ) . Iso2Flux computes steady-state flux distributions as the product of the null space of the stoichiometric matrix and the vector of free fluxes . Reversible reactions are split into forward and reverse reactions . For each reversible reaction , a turnover variable ( ti ) is introduced defining the flux that is common to the forward ( vif ) and reverse ( vir ) reactions . These variables are used to assign values to the fluxes of the forward and reverse reactions as a function of the steady-state net flux ( vi ) . Iso2flux uses the Elementary Metabolite Unit ( EMU ) framework[1] to build the 13C propagation model . This framework is based on a highly efficient decomposition method that identifies the minimum amount of isotopologue transitions required to simulate the experimentally quantified isotopologues according to the defined carbon propagation rules . The isotopologue transitions are grouped into decoupled systems based on isotopologue size . Balance equations are built around each isotopologue fraction under the assumption of isotopic steady state ( S1 Fig ) . Using the steady-state flux distribution as an input , systems of equations around isotopologues balances are solved sequentially starting with the smallest isotopologue size [1] using the fsolve function of the SciPy library ( https://scipy . org/scipylib/index . html ) . Solving such system predicts the isotopologue distribution associated with a given steady-state flux distribution ( Yj ( v ) ) . The self-adaptive differential evolution ( SADE ) algorithm from PyGMO ( Python Parallel Global Multiobjective Optimizer , https://github . com/esa/pagmo2 ) was used to find the optimal solution of the 13C MFA ( Eq 1 ) and p13CMFA ( Eq 2 ) problems . SADE was parallelized using the generalized island-model paradigm . Under such implementation , SADE is run in parallel in different CPU processes ( islands ) . After a given number of SADE iterations ( generations ) , the best solutions ( individuals ) in each SADE process ( island ) are shared to parallel SADE processes ( migrate to adjacent islands ) . To prevent bias from the starting solutions ( starting populations ) , the islands are seeded through random sampling of all variables . Free fluxes variables are sampled using the optGpSampler implemented into COBRApy[42 , 43] . Turnover variables are sampled using the random . uniform function built into python . The algorithm was run with 7 islands , each with a population of 60 , and with migrations between islands set to occur every 400 generations . For the analyzed 13C MFA and p13CMFA problems , repeated iterations of the algorithm were shown to reliably converge towards the same minimal objective function value . At the beginning of a 13C experiment , all internal metabolites are unlabeled ( m0 ) . Progressively , these products are enriched in 13C , with the subsequent decrease in m0 . Isotopic steady state is quickly reached for small pools of metabolites but not necessarily for larger pools such as those of fatty acids , glycogen or metabolites present in large concentrations in the external medium[44] . For these larger pools , unlabeled isotopologues m0 are oversized and might not quickly decrease to the theoretical value that should be reached at steady-state . However , it is possible to represent the effect of large pools in the framework of steady-state 13C MFA through the addition of a virtual reaction . This reaction replaces labeled isotopologues by unlabeled isotopologues in metabolites with large pools . With p13CMFA , the flux through this virtual reaction can be minimized . Effectively , this allows correcting steady-state 13C simulations for large pools while identifying the solutions that require the minimum amount of correction . The statistical significance of the difference between correlation coefficients was evaluated using the Fisher r-to-z transformation[33] . Following this approach , Pearson’s correlation coefficients ( r ) can be converted to a z-score ( r’ ) : r′=12·Ln ( 1+r1−r ) ( Eq 12 ) The variance of z ( Sz ) will depend only on the sample size ( n ) : Sz=1n−3 ( Eq 13 ) From Eq 12 and Eq 13 , the significance of the difference between two correlation coefficients ( r1 and r2 ) can be evaluated by computing the z score corresponding to such difference ( Eq 14 ) and its associated p-value . Human Umbilical Vein Endothelial Cells ( HUVECs-pooled , Lonza ) were maintained on 1% gelatin-coated flasks at 37°C in a humidified atmosphere of 5% CO2 and 95% air in MCDB131 ( Gibco ) medium , supplemented with the recommended quantity of endothelial growth medium ( EGM ) SingleQuots ( Lonza ) , 10% fetal bovine serum ( FBS ) ( Gibco ) , 2 mM glutamine ( Gibco ) and 0 . 1% Streptomycin ( 100 μg/mL ) /Penicillin ( 100 units/mL ) ( S/P ) ( Gibco ) . 1 × 106 HUVECs were seeded in 1% gelatin-coated cell culture plates for 6h , and then the maintenance medium was replaced with the MCDB131 basal medium , supplemented with 2% FBS , 2 mM glutamine and 0 . 1% S/P and cells were incubated overnight for nutrient deprivation . After nutrient deprivation , the medium was replaced with a restricted medium containing MCDB131 medium supplemented with 2% FBS , 2 mM glutamine and 0 . 1% S/P with 10 mM of 50% [1 , 2-13C2]-glucose ( Sigma-Aldrich ) and cells were incubated for 40h in a humidified atmosphere with 5% CO2 and 1% O2 at 37°C . Both at the beginning ( t = 0h ) and the end ( t = 40h ) of incubation , media and pellets were collected . On the one hand , media and cell pellets were used for analyzing isotopologue abundances for glucose , lactate , glutamate , RNA ribose and glycogen . Raw data are publicly available in the MetaboLights repository at http://www . ebi . ac . uk/metabolights [31] , with accession number MTBLS412 . Isolation , derivatization and analysis details are described in MetaboLights . Glucose , lactate , glutamate , and glutamine concentrations were determined in media samples for estimation of secretion or uptake rates of these metabolites using spectrophotometric methods[45] . Also , the net rate of glycogen re-utilization into glucose was estimated by quantifying glycogen content at initial and final time points using [U-13C-D7]-glucose as recovery standard[46] . All biochemical data were normalized by cell number , and by incubation time ( h ) . The resulting rates–expressed in micromoles of metabolite consumed/produced/transformed per hour per million cells ( μmol·h-1·million-cells-1 ) –were 0 . 463 , 0 . 099 , 0 . 050 and 1 . 169 for glucose uptake , glutamine uptake , glutamate secretion , and lactate secretion , respectively , and a net transformation of glycogen of 0 . 000175 .
|
13C Metabolic Flux Analysis ( 13C MFA ) is a well-established technique that has proven to be a valuable tool in quantifying the metabolic flux profile of central carbon metabolism . When a biological system is incubated with a 13C-labeled substrate , 13C propagates to metabolites throughout the metabolic network in a flux and pathway-dependent manner . 13C MFA integrates measurements of 13C enrichment in metabolites to identify the flux distributions consistent with the measured 13C propagation . However , there is often a range of flux values that can lead to the observed 13C distribution . Indeed , either when the metabolic network is large or a small set of measurements are integrated , the range of valid solutions can be too wide to accurately estimate part of the underlying flux distribution . Here we propose to use flux minimization to select the best flux solution in the13C MFA solution space . Furthermore , this approach can integrate gene expression data to give greater weight to the minimization of fluxes through enzymes with low gene expression evidence in order to ensure that the selected solution is biologically relevant . The concept of using flux minimization to select the best solution is widely used in flux balance analysis , but it had never been applied in the framework of 13C MFA . We have termed this new approach parsimonious 13C MFA ( p13CMFA ) .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"carbohydrate",
"metabolism",
"medicine",
"and",
"health",
"sciences",
"metabolic",
"networks",
"glucose",
"metabolism",
"metabolites",
"network",
"analysis",
"pharmacology",
"molecular",
"biology",
"techniques",
"drug",
"metabolism",
"research",
"and",
"analysis",
"methods",
"cell",
"labeling",
"computer",
"and",
"information",
"sciences",
"gene",
"expression",
"metabolic",
"pathways",
"molecular",
"biology",
"pharmacokinetics",
"metabolic",
"labeling",
"biochemistry",
"glycogens",
"genetics",
"biology",
"and",
"life",
"sciences",
"metabolism",
"glycobiology"
] |
2019
|
p13CMFA: Parsimonious 13C metabolic flux analysis
|
Broadly neutralizing HIV-1 antibodies ( bNAbs ) isolated from infected subjects display protective potential in animal models . Their elicitation by immunization is thus highly desirable . The HIV-1 envelope glycoprotein ( Env ) is the sole viral target of bnAbs , but is also targeted by binding , non-neutralizing antibodies . Env-based immunogens tested so far in various animal species and humans have elicited binding and autologous neutralizing antibodies but not bNAbs ( with a few notable exceptions ) . The underlying reasons for this are not well understood despite intensive efforts to characterize the binding specificities of the elicited antibodies; mostly by employing serologic methodologies and monoclonal antibody isolation and characterization . These approaches provide limited information on the ontogenies and clonal B cell lineages that expand following Env-immunization . Thus , our current understanding on how the expansion of particular B cell lineages by Env may be linked to the development of non-neutralizing antibodies is limited . Here , in addition to serological analysis , we employed high-throughput BCR sequence analysis from the periphery , lymph nodes and bone marrow , as well as B cell- and antibody-isolation and characterization methods , to compare in great detail the B cell and antibody responses elicited in non-human primates by two forms of the clade C HIV Env 426c: one representing the full length extracellular portion of Env while the other lacking the variable domains 1 , 2 and 3 and three conserved N-linked glycosylation sites . The two forms were equally immunogenic , but only the latter elicited neutralizing antibodies by stimulating a more restricted expansion of B cells to a narrower set of IGH/IGK/IGL-V genes that represented a small fraction ( 0 . 003–0 . 02% ) of total B cells . Our study provides new information on how Env antigenic differences drastically affect the expansion of particular B cell lineages and supports immunogen-design efforts aiming at stimulating the expansion of cells expressing particular B cell receptors .
Following HIV-1 infection , serum neutralizing antibody responses against the evolving autologous viral swarm are generated by the vast majority of infected subjects , usually within the first few months of infection [1–6] . In 10–30% of infected subjects , antibodies capable of neutralizing not only the autologous virus but also heterologous viruses are generated , usually following several years of infection [2 , 5 , 7–13] . These neutralizing antibodies are referred to as broadly neutralizing antibodies ( bNAbs ) . Binding but non-neutralizing antibodies ( nNAbs ) are also present in sera from infected subjects . Broadly neutralizing monoclonal antibodies isolated from HIV-1-infected subjects protect animals from experimental infection [14–23] and thus bNAbs are expected to be an important component of the protective immune response elicited by an effective HIV-1 vaccine . The viral envelope glycoprotein ( Env ) is the target for both nNAbs and bNAbs and the epitopes targeted by bNAbs and nNAbs have been identified and in many cases they have been structurally characterized [24–26] . In general , nNAbs target elements of Env that are variable in sequence and are located within the more exposed regions of Env on soluble gp120s or non-stabilized soluble gp140 proteins . In contrast , bNAbs bind conserved elements of Env . Here we examined whether the variable regions of Env stimulate the expansion of B cell lineages that differ from those expanded by more conserved Env regions , and whether such a differential B cell clonal expansion is linked to the elicitation , or not , of neutralizing antibodies . To this end , we performed immunizations studies in rhesus macaques , since they express IGH , IGK , and IGL V alleles with more than 93% homology to human alleles [27 , 28] , using two forms of a clade C Env ( 426c ) whose designs we previously described [29–31]: the full length extracellular form ( WT ) and one where the variable immunodominant regions 1 , 2 , 3 ( V1 , V2 and V3 , respectively ) and three N-linked glycosylation sites ( NLGS ) were artificially eliminated ( ‘NLGS-3 Core’ ) . We performed an in-depth analysis of serum antibodies and of isolated monoclonal antibodies as well as IGH/IGK/IGL deep sequencing analyses of the evolving immune B cell responses in the periphery , lymph nodes , and bone marrow . Although similar serum binding antibody titers were elicited by the two immunogens , the full-length immunogen activated a larger number of B cell clonal lineages than the NLGS-3 Core immunogen . Autologous neutralizing antibodies were elicited only by the NLGS-3 Core immunogen . Binding but non-neutralizing antibodies were derived from B cell clones that became predominant in the periphery , lymph node , and bone marrow during immunization , while neutralizing antibodies were derived from infrequent B cell clonal lineages ( 0 . 003–0 . 02% of total B cells ) . Our study provides a mechanistic explanation as to how the variable regions of Env elicit high titers of non-neutralizing antibodies . As such , our results support efforts to alter the immunogenicity of non-neutralizing epitopes located in these regions . Furthermore , our approach can be used by others to assess how specific Env modifications alter the activation of particular B cell lineages , or how different adjuvant formulations may alter the activation and expansion of particular unmutated B cell receptors by a particular Env .
The immunization schedule and timing of sample collection are summarized in S1 Fig and details are presented in the Materials and Methods section . High titers of autologous binding antibody responses were generated by all animals in both immunized groups , ranging from 7568 to 11924 reciprocal EC50 for the WT immunization group and from 3036 to 7824 for the NLGS-3 Core immunization group ( Fig 1A ) . The titers after the final immunization were not significantly different between the two immunization groups . In both immunization groups , minimal antibody responses against the gp41 subunit were observed , an indication that the elicited binding antibody responses targeted the gp120 subunits of the immunogens used here . A sizable fraction of the antibodies elicited by the NLGS-3 Core immunized animals targeted the CD4bs , something that was not observed with the WT immunized animals ( S2 Fig ) . Thus , the majority of the serum antibody responses elicited by the WT immunogen recognize epitopes outside the CD4bs . Serum antibody neutralizing activities were determined following the first and second DNA/Protein ( DP ) booster immunizations against the autologous 426c WT virus and three NLGS derivatives: NLGS-1: lacking the NLGS in loop D ( N276 ) ; NLGS-2: lacking the two NLGS in V5 ( N460 and N463 ) ; and NLGS-3: lacking all three NLGS . We note that the 426c Core Env ( i . e . , lacking the 3 NLGS and the variable regions 1–3 ) is not functional and cannot be tested as a virus . Therefore , all four autologous viruses used here express the variable regions 1–3 . The 426c WT virus and these three NLGS derivatives exhibit a tier 2 neutralization phenotype when assayed with sera from chronically HIV-1-infected individuals . As additional evidence of a tier 2 phenotype , all these viruses resisted neutralization by a panel of monoclonal antibodies against the gp120 V3 loop ( 2219 , 2557 , 3074 , 3869 , 447-52D and 838-12D ) and CD4bs ( 654-30D , 1008-30D , 1570D , 729-30D and F105 ) that are relatively specific for tier 1 viruses . We do want to emphasize that , although the three NLGS-lacking viral derivatives of the 426c virus display tier 2 overall neutralization phenotypes , they are more susceptible to neutralization by certain VRC01-class MAbs than the WT 426c virus . Irrespective of the immunogen used , neutralization of the WT 426c virus was not recorded ( Fig 1B ) . Only the NLGS-3 Core immunogen elicited serum neutralizing antibody responses against all three autologous NLGS viral variants ( Fig 1B ) . The strongest neutralizing activities ( reciprocal IC50s ) were observed against NLGS-3 and the weakest against NLGS-1 . Although 4/4 animals generated anti-NLGS-3 neutralizing antibody responses , only 2/4 animals ( A13284 and A13286 ) generated neutralizing antibody responses against NLGS-2 ( and only following the last immunization ) . Anti-NLGS-1 neutralizing antibodies were also elicited by 2/4 animals ( A13283 and A13284 ) , only one of which ( A13284 ) elicited anti-NLGS-2 neutralizing antibody responses . Overall , while 4/4 animals generated anti-NLGS-3 neutralizing antibody responses , one animal ( A13284 ) generated neutralizing antibodies against all three viruses . Interestingly , in the case of animal A13283 , vaccine-elicited antibodies neutralized the virus expressing an Env that lacked the NLGS in V5 ( NLGS-1 ) , but not the virus that lacked the NLGS in Loop D ( NLGS-2 ) . Neutralizing activities against the NLGS-1 and NLGS-2 viruses were not detected following the first DNA / Protein immunization . Thus , in animals immunized by the NLGS-3 Core immunogen , only a fraction ( Log differences ) of neutralizing antibodies can bypass the glycan steric blocks presented on Loop D and/or V5 , but not both ( i . e . , the WT Env ) . Overall , the results suggest that: ( a ) the NLGS-3 Core immunogen can elicit autologous NAbs that can ‘by-pass’ the variable regions 1 , 2 and 3 ( which are absent from the immunogen , but present on the virus ) , as long as the NLGS in V5 or Loop D are absent ( individually or in combination ) ; ( b ) that these NAbs have a harder time ‘bypassing’ the NLGS in V5 ( N460 and N463; i . e . , the NLGS-1 virus ) than the NLGS in loop D ( N276; i . e . , the NLGS-2 virus ) ; and ( c ) that the autologous NAbs elicited by the NLGS-3 Core target an epitope whose relative exposure on the virus is regulated by the presence of NLGS in Loop D and V5 , similar to what is known for several anti-CD4bs bNAbs [31–35] . The sera did not display neutralizing activity against several heterologous viruses ( Clade A: Q168a2 , Q461e2 , Clade B: QHO692 , SF162 , Bal . 26 , and Clade C: 706c , 823c ) . Elimination of the NLGS ( in Loop D and V5 ) from most of these viruses ( Q168a2 , Q461e2 , Bal . 26 , 706c and 823 ) did not lead to neutralization either . Thus , the neutralizing antibodies elicited by the NLGS-3 Core in non-human primates ( NHPs ) using this immunization regimen target epitopes that are either absent or are present but are less accessible on heterologous viruses . To determine whether the different abilities of these two immunogens to elicit neutralizing antibody responses were linked to a differential stimulation of B cell lineages by the two immunogens , we performed next generation Illumina MiSeq deep sequencing analysis of the variable domains of the heavy ( IGHV ) and light ( IGKV and IGLV ) chains from B cells isolated from the periphery , pre- and post-immunization . A number of IgHV genes from circulating memory B cells became commonly enriched among the animals from the group immunized with WT 426c Env . In the IGH locus , genes belonging to the IGHV3 family underwent the most expansion ( 13 total genes ) , and to a much lesser extent we observed expansion in the IGHV1 ( 2 genes ) , IGHV4 ( 3 genes ) and IGHV7 ( 2 genes ) ( Fig 2A , S3 Fig Top panel ) . Stimulation of these genes was observed after both DNA immunization and after protein plus DNA immunization ( Fig 2B ) . This , despite the fact that IGHV1 represents <5% and IGHV3 ~20% of circulating IGHV in IgM+ B cells in NHP , while IGV4 is the most frequently expressed IGVH ( ~70% ) in IgM+ B cells [28] . However , we previously reported [28] that the IgGHV3 family becomes more prevalent in the IgG+ B cell compartment compared to the IgM+ compartment , indicating a preferential stimulation of IGHV3 into class-switched memory B cells , in agreement with what we observed here . The particular expansion of the above-mentioned VH genes was also reported by others in studies were a different full length soluble Env was used [36 , 37] . Thus , it is probable that the stimulation of IGHV1 and IGHV3 of the circulating memory B cells is due to immunization with Env . In the light chain loci , we observed enrichment in both the IgK and IgL loci after immunization with WT 426c ( Fig 2A ) . In the IgK locus , IgKV1 was the most enriched , followed by IgKV2 , IgKV3 , and IgKV4 . In the IgL locus , the IgLV2 family was the most enriched after immunization , followed by IgL1 , IgLV3 , IgLV5 , and IgLV8 families . As with the IgH locus , stimulation of the light chain families was observed after both DNA and DNA plus protein immunization ( Fig 2B ) . IgKV1 and IgLV2 are the predominantly expressed gene families from their respective loci [28] . We did not observe any significantly enriched IGHV , IGKV , or IGLV gene families after immunization with NLGS-3 Core ( Fig 2 ) , indicating that there was not a widespread stimulation of the same V genes within the group . We determined that this finding was not due to the NGS sequence data sets themselves , as quality and Hill’s diversity analysis of all sequence sets reported here revealed all data sets to be roughly equivalent in structure and quality , no matter the chain that was amplified nor the origin of the libraries ( S4–S7 Figs ) [38] . These findings were confirmed by principal component analyses , which clusters large , multi-dimensional data sets by the most significant sources of variation . In the WT animals , the NGS data sets clustered by time point , indicating that the statistically significant changes in gene abundance were due to vaccination time point . In contrast , the NLGS-3 NGS data sets cluster by animal and not time point , confirming that vaccination did not drive significant changes in common gene usage among the animals in this group ( S8 Fig ) . This stark dichotomy implies that , while the NLGS-3 is immunogenic and elicits IgG titers similar to that of WT 426c , it does not broadly stimulate a diversity of V genes during immunization . Potentially , this is a direct , measurable consequence of the elimination of the highly immunogenic variable loops . To better characterize the B cells that produce neutralizing antibodies and those that produce binding but not neutralizing antibodies , we isolated Env-specific IgG B cells from individual animals following immunization based on their CD4bs specificity ( based on the D368R and E370A mutations , DREA ) . Thus , two populations of B cells were isolated from animals immunized with either immunogen: CD4bs-specific cells ( Env+/CD4bs-KO- B ) cells and non-CD4bs-specific cells ( Env+/CD4bs-KO+ B cells ) . The corresponding recombinant Env used to immunize the animals was used for B cell-isolation . B cells were cultured in bulk in multiple wells , each well containing ~1000 B cells , due to the high number of sorted B cells . The cell supernatants were evaluated for anti-WT 426c and anti-NLGS-3 virus neutralizing activities ( Fig 3 ) . Supernatants from wells containing B cells ( irrespective of their CD4bs specificities ) isolated from the WT-immunized animals did not display neutralizing activities . In contrast , supernatants from 4 of 6 wells containing non-CD4bs specific B cells isolated from the NLGS-3 Core-immunized animals neutralized the autologous NLGS-3 virus , but not the WT virus . Thus , the neutralization results obtained from B cell supernatants and those obtained from sera ( Fig 1B ) were in agreement . The NLGS-3 neutralizing activity was derived from non-CD4bs-specific B cells . Since these B cells were isolated from animals immunized with the NLGS-3 Core immunogen , by definition they do not bind elements of V1 , V2 , or V3 regions . Thus , although the majority of serum binding antibodies target the CD4bs ( S2B Fig ) , the neutralizing activity in the sera is due to antibodies whose binding is independent of the DREA mutation that is widely used to identify anti-CD4bs antibodies in sera [5 , 39–41] . To better define the characteristics of the neutralizing antibodies elicited by NLGS-3 Core and to compare them to those of non-neutralizing antibodies elicited by the same immunogen , we isolated individual CD4bs-specific and non-CD4bs-specific and peripheral IgG+ B cells from the four animals immunized with the NLGS-3 Core ( S9 Fig ) . The CD4bs-specific B cells represented the minority of Env-specific B cells ( between 4 . 5% and 8 . 6% of total Env+ B cells . Within the non-CD4bs-specific B cell population , the neutralizing B cells represented a small fraction ( between 0 . 4% and 2 . 4% ) . Thus , only a very small fraction ( 0 . 003 to 0 . 02% ) of total periphery IgG+ B cells display neutralization potential . IGH and IGK/IGL genes from wells containing individual B cells that displayed neutralizing activities against the NLGS-3 virus and from wells displaying 426c NLGS-3 Core-binding , but not neutralizing activity were amplified and sequenced . Seventy-nine IGH ( S10 Fig ) and 32 IGK/IGL ( S11 Fig ) genes were fully sequenced from peripheral B cells secreting binding , but not-neutralizing antibodies . The majority of IGH ( ~90% ) were derived from IGHV3 and IGHV4 alleles . The majority of the IGHV3 sequences ( ~52% ) were derived from the IGHV3-Korf19 allele ( human homologue is IGHV3-33*01 [28] ) . A majority of the IGK light chains were derived from the IGKV1 family ( ~28% ) ( S11 Fig ) , including IGKV1-I21 ( human homologue IGKV1D-16*01 ) which represented 10 . 9% of IGK present in the peripheral B cells . The second most frequently expressed LC was derived from IGLV5 . Within the IGLV5 , the predominant allele was IGLV5-S28 . Overall , we estimate that over 40% of the serum response was represented by these binding , but non-neutralizing antibodies , due to their allelic dominance in the peripheral B cell repertoire .
Different concepts to elicit HIV-1 bNAbs through immunization are under investigation . One concept is based on the ‘germline-targeting’ approach during which a ‘germline-targeting’ immunogen is used to initiate the activation of naïve B cells expressing specific germline ( unmutated ) BCRs and subsequently , booster immunizations with specifically-designed immunogens to guide the maturation ( through somatic hypermutation ) of these BCRs towards their broadly neutralizing forms [42] . This approach was recently shown to be effective in eliciting PGT121 bNAbs in a knock-in mouse model where the PGT121 germline BCR was expressed by every B cell [43] . On a polyclonal BCR background however , Env immunogens ( including ‘germline-targeting’ immunogens ) will activate not only the desired B cells , but many B cells that recognize irrelevant , but immunogenic , epitopes on the immunogen [44] . These off-target B cells will expand even further during the booster immunizations , potentially limiting the expansion of the desired B cells ( although this has not yet been experimentally demonstrated ) . We ( and others ) believe that an in-depth understanding of BCR clonal lineages expanding during immunization with Env-based immunogens will be important to identify the booster immunogens for the optimal development of bNAbs [45] . The NLGS-3 Core Env immunogen used here , was engineered by introducing specific mutations on the tier 2 clade C 426c viral Env . These include deletions of the variable domains 1 , 2 and 3 and the targeted elimination of three NLGS: one in Loop D ( N276 ) and two in V5 ( N460 and N463 ) [29 , 31 , 46] . These modifications were introduced so that this Env engages B cells expressing germline VRC01-class BCRs , which are derived from the human VH1-2*02 allele paired with LCs expressing infrequent 5 amino acid long CDRL3 domains . We now know that rhesus macaques ( and other animal species such as mice , rats and rabbits ) do no express an exact orthologue to the human VH1-2*02 allele [28 , 47 , 48] . The present study was not therefore conducted to inform on the ability of the NLGS-3 Core immunogen to activate and expand B cells expressing germline VRC01-like BCRs in vivo , but to generate new mechanistic information on why the variable V1 , V2 and V3 domains of Env dominate the B cell responses upon Env immunization . In this regard we note that although the observed relative changes in IGH and IGK/IGL were assessed from ‘total’ B cells and not exclusively from Env-specific B cells , we expect that these changes are due to differences in the antigenic/immunogenic properties of the two immunogens we evaluated here , as the immunogens were prepared ( expressed and purified ) identically , the same adjuvant was used and the immunization protocols were identical . We also note that although the NGS analyses of IGH and IGK/IGL were performed on the same samples , they were derived from bulk , but not individual B cells and thus presently we cannot assess whether the expansion of a particular IGH lineage was linked with the expansion of a particular IGK/IGL lineage . At first glance , it is not surprising that the WT immunogen stimulated a larger number of IGH and IGK/IGL lineages , as it expresses more epitopes than the NLGS-3 Core; especially the variable V1 , V2 and V3 domains which are known to be immunogenic , both in the context of HIV-1 infection and immunization with Env [39 , 40 , 49–55] . One would expect that the immunogenicity of epitopes present on the core part of Env to increase in the absence of the immunodominant variable domains . This was the case , as the NLGS-Core elicited similar serum antibody titer responses as the WT immunogen . We note that both immunogens examined here are not stabilized trimers . It is anticipated that the immunogenicity of variable domains will be reduced on such constructs [56–59] . The fact however , that neutralizing antibodies against the autologous NLGS viruses were elicited by the immunogen lacking the variable regions 1 , 2 and 3 , indicate that neutralization epitopes are located outside these variable regions of Env; within the core components of Env . These epitopes are also present in the WT immunogen ( as the two immunogens share a common amino acid sequence ) , but since that immunogen did not elicit neutralizing antibodies we assume that these epitopes are occluded and thus poorly immunogenic . The lack of neutralization of the WT 426c virus by the neutralizing antibodies elicited by the NLGS-3 Core immunogen supports this assumption . In part the occlusion of the epitopes is due to carbohydrates present on NLGS in Loop D and V3 , as viruses lacking these NLGS are susceptible to the neutralizing antibodies elicited by the NLGS-3 Core immunogen . One ( 13284 ) of four animals immunized with NLGS-3 Core elicited serum neutralizing antibody responses against both the NLGS-2 virus ( lacking the two NLGS in V5; positions N460 and N463 ) and the NLGS-1 virus ( lacking the NLGS in Loop D; position N276 ) . The neutralization titers against the NLGS-2 virus were 30 fold higher than those against the NLGS-1 virus . This , combined with the fact that none of the isolated neutralizing antibodies neutralized the NLGS-1 virus , but 5/8 neutralized the NLGS-2 virus , suggests that B cells producing antibodies capable of bypassing the restrictions imposed by the V5 NLGS were less frequently expanded during immunization than the B cells producing antibodies capable of bypassing the Loop D NLGS . As the conserved NLGS in Loop D ( N276 ) is a major block in the engagement of germline VRC01-class BCRs by Env [31–33 , 35 , 47 , 60 , 61] , our results suggest that the 426c NGLS-3 Core immunogen presents CD4bs epitopes more favorably as compared to the WT Env even in animals without VRC01-like naïve B cells . An alternative possibility is that the NLGS-3 Core mutations themselves are the cause of the development of the neutralizing antibody responses discussed here . As we did not immunize animals with a 426c Env that only lacked the 3 NLGS , we do not know the relative impact these NLGS or the variable regions had on the observed B cell lineage expansions observed with the 426c NLGS-3 Core immunogen . Regardless , our results suggest that in animals with a polyclonal naïve BCR repertoire capable of producing VRC01-class B cells immunization with the 426c NGLS-3 core immunogen may lead to activation of VRC01-like naïve B cells . Our underline hypothesis for the observations made here , is that the epitopes targeted by the neutralizing antibodies elicited by the NLGS-3 Core immunogen are occluded on the 426c WT Env and thus are not immunogenic . It is however possible that they equally immunogenic on the 426c WT Env , but because the immunogenicity of the variable domains is so high , it overwhelms the response to the neutralizing epitopes . We previously reported that non-neutralizing anti-CD4BS antibodies can prevent the uptake of Env by B cells expressing precursors of broadly neutralizing antibodies ( [29] ) . The outcome of the competition between ‘on target’ and ‘off target’ B cells responses to Env depends on the relative frequencies of ‘on-target’ and of ‘off-target’ B cells ( that are always going to be present at some level ) and on the relative affinities of the ‘on-target’ and ‘off-target’ BCRs to their respective epitopes on the same immunogen ( [62 , 63] ) . Thus , immunogen-design approaches that aim at reducing the immunogenicity of non-neutralizing epitopes on Env immunogen and at increasing the affinity of BCRs for neutralizing epitopes are fully warranted .
All Env constructs are based on the HIV-1 clade C 426c Env ( GenBank: KC769518 . 1 ) . Mutations that disrupt N-linked glycosylation sites ( NLGS ) were introduced , individually and in combination , in Loop D ( N276 ) and V5 ( N460 and N463 ) to generate the following single , double and triple mutants: N276D ( NLGS-1 ) , N460D+N463D ( NLGS-2 ) , and N276D+N460D+N463D ( NLGS-3 ) [31] . Deletions of the variable regions 1 , 2 , and 3 were also introduced on the NLGS-3 background ( this construct is referred to as ‘NLGS-3 Core’ [46] . D368R/E370A mutations that knock-out the binding of many anti-CD4-binding site antibodies ( CD4-binding site KO , CD4bs-KO ) were also introduced on some of the above-mentioned constructs . CD4bs-KO reagents were employed during the B cell-sorting experiments ( see below ) . Soluble gp140 or gp120 forms of these Envs were expressed from the pTT3 vector [29 , 31] . Soluble recombinant gp120 envelopes produced by transient transfection of 293E/F suspension cells and purified using a size-exclusion chromatography AKTA purifier ( GE , Fairfield , CT ) as described previously [64 , 65] . Avi-tagged versions of WT , NLGS-3 Core , or NLGS-3 Core with CD4bs-KO mutations Envs were biotinylated overnight using the BirA enzyme in vitro biotinylation kit ( Avidity , Aurora , CO ) with an excess of biotin . Excess biotin was removed via Amicon Ultra-4 centrifugal membrane filtration ( EMD-Millipore , Billerica , MA , USA ) . Streptavidin-allophycocyanin ( SA-APC ) or streptavidin-allophycocyanin-Cy7 ( SA-APC-Cy7 ) was conjugated to biotinylated proteins at an optimized ratio . Two groups ( four animals each ) were immunized with either 426c WT ( Animals IDs: A13279 , A13280 , A13281 , and A13282 ) or 426c NLGS-3 Core ( Animals IDs: A13283 , A13284 , A13285 , and A13286 ) ( gp140 forms ) ( S1 Fig ) . The latter construct expressed the I423M / N425K / G431E mutations that reduces binding to human and macaque CD4 [66] . At weeks 0 and 4 , the animals were immunized with DNA vectors expressing the gp140 Env forms . 2mg DNA in 1mL endotoxin-free water was administered intradermally in 2 sites in the back ( 0 . 2mg each ) and intramuscularly in 2 sites in the quads ( 0 . 8 mg each ) . Protein immunizations were administered with 20% Adjuplex at weeks 12 and 20 . 0 . 1mg protein in 0 . 5mL 20% adjuvant mixture was administered intramuscularly in the deltoids . Blood was collected at weeks -4 , -2 , 1 , 2 , 5 , 6 , 12 , 13 , 14 , 20 , 21 , and 22 . Lymph nodes ( axillary and/or inguinal ) were collected at weeks -2 , 13 , 21 , and 39 , and bone marrow collected at week 22 . PBMCs , plasma , and bone marrow were purified from freshly-collected blood using density gradient centrifugation with Ficoll-Paque and SepMate columns ( StemCell Technologies , Vancouver , BC , Canada ) according to adapted manufacturer’s instructions . Isolated PBMCs were resuspended ( 20 x 106 cells/mL ) in freezing media ( 90% heat-inactivated FBS , 10% DMSO ) , placed in Mr . Frosty containers ( ThermoFisher Scientific , Waltham , MA ) , and stored at -80°C overnight before transfer to liquid nitrogen , where they were stored until further use . Plasma was aliquoted and stored at -80°C . Lymph nodes were sliced into grindable parts and cell strained using a 40μm strainer followed by a rinse with RPMI-1640 media , and then a rinse with 10mL PBS ( Thermo Fisher Scientific , Waltham , MA ) . Isolated lymph node cells were resuspended ( 10 x 106 cells/mL ) in freezing media ( 90% heat-inactivated FBS , 10% DMSO ) , placed in Mr . Frosty containers , and stored in -80°C overnight before transfer to liquid nitrogen , where they were stored until further use . All NHP studies were conducted at the Washington National Primate Research Center at the University of Washington ( Seattle , WA , USA ) . The study was reviewed and approved by the UW Institutional Animal Care and Use Committee , Office of Animal Welfare , University of Washington under Protocol Number: 3408–04 and Protocol Title: Optimizing HIV Immunogen-BCR Interactions for Vaccine Development . Housing and care procedures were within guidelines of the National Institutes of Health ( NIH ) ( National Research Council , Guide for the Care and Use of Laboratory Animals , 8th edition ) and in compliance with federal regulations relating to animal welfare . All efforts were made to minimize suffering . Details of animal welfare and steps taken to ameliorate suffering were in accordance with the recommendations of the Weatherall report , "The use of non-human primates in research" . Rhesus macaques ( Macaca Mulatta ) of Indian origin , approximately 3 years old , were habituated to the housing conditions ( > 4 weeks ) before the initiation of the study . All procedures were conducted under anesthesia ( 10mg/kg ketamine HCL ) . Animals were individually housed in suspended stainless steel wire-bottomed cages and provided with a commercial primate diet . Fresh fruit was provided once daily and water was freely available at all times . A variety of environmental enrichment strategies were employed . The animals were not terminated at the conclusion of study , and were released back into the colony . 10–20 million PBMCs , lymph nodes ( LNs ) , or bone marrow ( BM ) were thawed and resuspended in 12mL complete RPMI ( 2% Penn-strep , 10% heat-inactivated FBS ) , centrifuged at 1 , 400rpm for 5 min , then rinsed with FACS Buffer ( 2% heat-inactivated FBS in sterile PBS ) . The cells were resuspended in 1 – 2mL RBC lysis buffer ( Sigma , St . Louis , MO ) for 10min at RT to lyse red blood cells according to manufacturer’s instructions . 8 – 10mL of 1x PBS was used to rinse cells . The cells were resuspended in 50–100μL PBS , and then stained with 0 . 5–1μL of Live/Dead Fixable Aqua Dead Cell Stain according to manufacturer’s protocol ( Invitrogen/Life Technologies , Grand Island , NY ) . Cells were then stained with NLGS-3 Core CD4bs-KO–APC-C7 ( 426c . NLGS-3 . D368R . E370A . MKE . DV1/2/3-APC-Cy7 ) for 10min on ice in the dark , then with either WT- APC or NLGS-3 Core rEnv—APC ( 426c . WT-APC or 426c . NLGS-3 . MKE . DV1/2/3-APC ) for 10min . Cells were then stained with a master mix of CD3-FITC , clone SP34 ( BD Biosciences , San Jose , CA ) , CD14-FITC , clone MφP9 ( BD Biosciences , San Jose , CA ) , CD19-PE , clone J3-119 ( Beckman Coulter , Brea , CA ) , and IgG PECF594 , clone G18-145 ( BD Biosciences , San Jose , CA ) . For compensation set-up , PBMCs collected prior to immunization were used . IgG-APC , clone G18-145 was utilized for the compensation control in the APC channel due to the minimal binding of the rEnv-APC to naïve rhesus macaque PBMCs ( BD Biosciences , San Jose , CA ) . All samples were resuspended in 1 mL media ( ~1M cells /mL ) and filtered through 70μm Flowmi strainers ( Scienceware Bel-Art , Wayne , NJ ) . LD Aqua- / CD3- / CD14- / CD19+ / IgG+ / WT+ or NLGS-3 Core rEnv+ / NLGS-3 Core CD4bs-KO rEnv+/- cells were bulk sorted into 100μL complete IMDM media using a FACS Aria II cell sorter ( BD Biosciences , San Jose , CA ) . Sorted rEnv-specific cells were plated on 3T3-msCD40L feeder cells ( provided by Dr . J . R . Mascola , NIH/VRC , 3T3-msCD40L are NIH 3T3 mouse embryonic fibroblast cells engineered to express the CD40 ligand ) at a final dilution of 1 . 4 B cells/well and cultured as previously described [67] . After 12 days , supernatants were collected for ELISA and neutralization ( see below ) testing and the cells were lysed with 30μL of RLT supplemented with β-ME/glycogen and frozen at -80°C . Prior to single cell sorting , negative and positive populations of WT rEnv+ / NLGS-3 Core rEnv+ and NLGS-3 Core CD4bs-KO rEnv+/- , cells were sorted into 100μL of complete IMDM media supplemented with 2% Penn-strep and 10% heat-inactivated FBS then plated at 1000 cells/well on 12-well plates with 3T3-msCD40L feeder cells with conditions described above . After 12 days , supernatants were removed and were tested for neutralizing activity ( see below ) and the cells were lysed with RLT supplemented with β-ME/glycogen and stored at -80C . RNA recovery , cDNA synthesis , and PCR amplification were carried out as previously described [37 , 68 , 69] with a few minor modifications . 45μL of RLT lysis buffer supplemented with β-ME was added to previously lysed cells for a total volume of 75μL RLT , and RNA was column-purified with RNeasy Micro Kit ( Qiagen , Venlo , Netherlands ) . 14μL of RNA was used directly in a 20μL cDNA synthesis reaction using the High-Capacity cDNA Reverse Transcription kit according to manufacturer’s instructions ( Applied Biosystems / Life Technologies , Grand Island , NY ) . IGH , IGK , and IGL gene transcripts were then amplified independently from cDNA using first and second round primers followed by nested PCR previously published ( S2–S4 Tables ) [37 , 69 , 70] . First and second round PCR were done using Phusion High Fidelity DNA polymerase according to manufacturer’s protocol . The first round PCR included 2 mins at 94C followed by 50 cycles of 94°C 10s , 55°C 30s , 72°C 30s with a final extension at 72°C 5mins . 3μl of primary PCR product was used in the nested second round PCR , which included 2 min at 94°C followed by 50 cycles of 90°C 30 s , 72°C 30 s , 72°C 5 min , and cooling at 4°C 15min . PCR products were evaluated on 1 . 2% flash gels ( Lonza , Rockland , ME ) and band sizes at ~450–500 bp were purified via Qiaquick purification columns ( Qiagen , Venlo , Netherlands ) or via Agencourt AM Pure XP beads ( Beckman Coulter , Brea , CA ) . A third and final PCR using Accuprime pfx ( Life Technologies , Carlsbad , CA ) was performed to add MiSeq adaptors that were used to prime direct amplicon sequencing . The PCR program was initiated with 5 min at 95°C followed by 10 cycles of 95°C 15 s , 55°C 30 s , 68° 30 s , 68°C 5 min , and cooling at 4°C 15min . Third round PCR product was sent for Sanger sequencing ( Genewiz , Plainfield , NJ or SeattleBiomed , Seattle , WA ) . If sequencing reads were unclear , second round nested PCR products were TOPO cloned following manufacturer’s instructions ( Life technologies , Carlsbad , CA ) and then sent for Sanger sequencing . Paired heavy and light chain sequences were matched against both the human genes via IGMT/V-quest and against the rhesus macaque genes via a customized IgBlast database search using an IGH/IGK/IGL database created and described previously [28 , 37 , 69 , 71–73] . Further analysis was conducted with Geneious Alignment software ( cite PMID: 22543367 ) . Antibody sequences can be found in Genbank database with accession numbers MF346735 , MF346736 , MF346737 , MF346738 , MF346739 , MF346740 , MF346741 , MF346742 , MF346743 , MF346744 , MF346745 , MF346746 , MF346747 , MF346748 , MF346749 , MF346750 , MF346751 , MF346752 , MF346753 , MF346754 , MF346755 , MF346756 , MF346757 , MF346758 . RNA was recovered from 1 , 000–100 , 000 sorted B cells ( CD19+ / IgG+ / CD27+/- ) from PBMCs using the flow staining protocol described above . RNA was recovered using the RNeasy Micro Kit , as described above . All 14μL of recovered RNA were ran through a speed vacuum for 3 min then used directly in the 12μL cDNA synthesis reaction using the Superscript III First-Strand Synthesis SuperMix kit according to manufacturer’s instructions ( Invitrogen / Life Technologies , Grand Island , NY ) . The RT program was initiated with a pre-warmed PCR for 5 min at 65°C followed by ice for 1 min and addition of enzyme mix , then by 1 cycle of 50°C 50 min , 25°C 10 min , 50°C 50 min , and termination at 85°C 5 min followed by chilling on ice . IGH , IGK , and IGL gene transcripts were amplified independently from cDNA using adaptor PCR described above . The adaptor PCR step was performed on 5μL diluted cDNA using Accuprime pfx according to manufacturer’s protocol ( Life Technologies , Carlsbad , CA ) . Adaptor PCR product was purified via Agencourt AM Pure XP beads as described above for 35–50 cycles ( Beckman Coulter , Brea , CA ) . The index PCR step was performed on 5μL cleaned adaptor round PCR product using the Nextera XT DNA Library Preparation kit ( Illumina , San Diego , CA ) with the Kapa HiFi DNA polymerase PCR program initiated at 3 min at 95C followed by 15–25 cycles of 98°C 20 s , 55°C 15 s , 72° 15 s , 72°C 1 min , and cooling at 4 C 15min ( Kapa Biosystems , Wilmington , MA ) . Index PCR product band size of 600 – 650bp was confirmed on a gel and purified via Agencourt AM Pure XP beads as described above ( Beckman Coulter , Brea , CA ) . Each library was diluted to 10nM , quantitated with Qubit and Bioanalyzer , and ran on Illumina HiSeq 2500 ( Illumina , San Diego , CA ) at 2 x 300 with the v3 25M kit at the Genomics Core at the Fred Hutchison Cancer Research Center ( FHCRC ) . Sequencing of multiple libraries ( limit of up to 20 per chip ) were performed during every sequencing run and a single library was never sequenced alone , thus the error PCR rate was the same for all libraries per chip . Additionally , during each sequencing run , an internal control was included to ensure the proper performance of the sequencer . Illumina data was processed , as previously described ( PMID: 27525066 ) . Briefly , raw data obtained from the forward and reverse MiSeq reads were merged to reconstruct the amplicon with FLASH ( ver . 1 . 2 . 11 ) ( PMID: 21903629 ) . The resulting amplicon sets were filtered to select only sequences containing the amplification primers ( a procedure during which the primer sequences themselves were removed ) using cutadapt ( ver . 1 . 14 ) ; amplicons containing low-confidence base calls ( N’s ) were then removed from the set , and deduplicated using FASTX-toolkit ( ver . 0 . 0 . 14 ) ( MARTIN , Marcel . Cutadapt removes adapter sequences from high-throughput sequencing reads . EMBnet . journal , [S . l . ] , v . 17 , n . 1 , p . pp . 10–12 , may . 2011 . ISSN 2226-6089 ) . Samples were annotated using a local IgBLAST ( ver . 1 . 6 . 1 ) [74] installation equipped with a custom database of previously published rhesus macaque gene segments ( described here [28] . The resulting annotated datasets contained assignments for the most likely matches to the database V/D/J segments and identified the CDR3 sequences for each processed amplicon . Amplicons representing productively-rearranged immunoglobulin sequences were then clustered based on three parameters: 1 ) V-family assignment , 2 ) J-family assignment , 3 ) CDR3 amino acid sequence . Amplicons that shared the CDR3 amino acid sequence but disagreed on the V- or J-family assignment were labeled as "chimera" and filtered out from further analysis . Only clusters containing 5 or more members were considered in further analysis . Cluster V-gene segment assignment was made based on the most abundant assignment found within each cluster; these assignments were then compiled into a gene counts table for each dataset for subsequent analysis with Bioconductor R package edgeR ( ver . 3 . 16 . 5 ) [75] . Data sets representing WT and NLGS3-core immunizations were analyzed separately , and log fold-change ( logFC ) in enrichment for each gene segment as well as the associated false discovery rate ( FDR ) were used to identify changes associated with the immunization regimens . Structures of clustered sequence populations were analyzed using the R-package alakazam ( PMID: 26069265 ) by calculating the values for the general Hill diversity index ( encompassing representations of Shannon’s entropy , evenness , etc . ) . ELISA assays were carried out using the following antigens: 426c . WT gp140 , 426c . NLGS-1 gp140 , 426c . NLGS-2 gp140 , 426c . NLGS-3 gp140 , 426c . NLGS-3 . MKE Core gp140 , 426c . NLGS-3 . MKE CD4bs-KO gp140 , 426c . WT . MKE gp120 , 426c . NLGS-3 . MKE gp120 , and HXB2 gp41 . 50ng of rEnv was adsorbed onto each well of 96-well or 384-well MaxiSorp ELISA plates ( Sigma , St . Louis , MO ) overnight in 0 . 1M NaHCO3 , pH 9 . 4–9 . 6 at RT . Plates were blocked using dilution buffer , a solution of Millipore H2O , 1X PBS , 10% non-fat dry milk , and 0 . 03% Tween-20 ( Sigma , St . Louis , MO ) for 1 h at 37°C . All supernatant from sorted single B cells growing on feeder cells was diluted 1:5 and 1:10 in complete RPMI and incubated for 1 h at 37°C . Serum was diluted 1:10 in dilution buffer and serially titrated 3-fold . Antibodies were diluted to 100μg/mL in dilution buffer and titrated 3-fold . A 3-fold titration of germline and mature VRCO1 was performed as a positive control . Bound antibodies were detected at 37°C for 1h with goat-anti-human-IgG ( H+L ) HRP conjugate ( Invitrogen , Grand Island , NY ) , diluted 1:6 , 000 with dilution buffer . Plates were developed with 30μl ( 384-well ) or 50μl ( 96-well ) SureBlue Reserve TMB Microwell Peroxidase Substrate ( KPL , Gaithersburg , MD ) , stopped with an equal amount of 1N H2SO4 . Absorption at 450nm was read on a Spectramax spectrophotometer ( Molecular Devices , Sunnyvale , CA ) . ELISA positive wells were above background of 0 . 2nm at a 1:5 dilution . Assays were performed in triplicate . Paired IGH and IGL/IGK sequences from isolated single B cells were fabricated into gBlocks ( IDT , Coralville , IA ) containing flanking regions with corresponding enzyme sites for vector-ligation . gBlocks were digested with the appropriate restriction enzymes EcoRI , NheI ( γ ) , Bsi-WI ( κ ) , XhoI ( λ ) ( NEB , Ipswich , MA ) and purified with the Qiagen PCR Purification Kit ( Qiagen , Venlo , Netherlands ) . Ligation was performed for 1–2 hours in a total volume of 10–20μL with 4U T4 DNA Ligase ( Invitrogen/Life Technologies , Grand Island , NY ) , 15ng digested and purified DNA , and 70ng linearized pt1-732 gL γ , pt1-695 κ , pt1-341 λ vectors . Chemically competent DH5α T1 E . coli cells were transformed with 2 . 5μL of ligation product . Single colonies were expanded for 16h at 37°C in 4mL LB Broth containing 1 μg/mL ampicillin . Plasmid DNA was purified via Qiagen Qiaprep Mini-Prep kit ( Qiagen , Venlo , Netherlands ) and eluted with 30μL EB Buffer . Plasmid DNA was quantified using a NanoDrop spectrometer ( Thermo Fisher Scientific , Waltham , MA ) and sequenced by the Sanger sequencing method ( Genewiz , South Plainfield , NJ ) using a pTT3 5’ forward primer . Plasmids containing confirmed sequences were expanded for 12h at 37°C in 100mL LB Broth containing 1μg/mL ampicillin . Plasmid DNA was purified with a HiSpeed Plasmid Maxi Kit ( Qiagen , Venlo , Netherlands ) , eluted with 50μL TE Buffer , and filtered with a 0 . 2μM filter before co-transfection . IgGs were produced by transient co-transfection of two plasmids: one expressing the IGH and the other the IGL/IGK . Briefly , 12 . 5μg of IGH plasmid DNA and 12 . 5μg of IGL/IGK plasmid DNA were incubated at RT for 15 m with 293F transfection reagent ( EMD Millipore , Temecula , CA ) in 1x PBS prior to addition to 293F cells ( HEK-293F suspension cells from the American Type Culture Collection ( ATCC ) ) , these are a variation of human embryonic kidney cells derived from an unknown original patient ) ) at 1 million cells/mL in 50mL of FreeStyle 293 Expression Media ( Life Technologies , Grand Island , NY ) . After 4–6 days incubation , the cell supernatants were centrifuged at 6 , 000rpm for 10min and the clarified supernatant was filtered through a 0 . 2μM filter ( Millipore , Billerica , MA ) before loading onto a pre-rinsed Protein-A/G agarose resin column ( Thermo Fisher Scientific , Waltham , MA ) . After washing the agarose beads with 10x column volumes of 1x PBS , IgG was eluted from the column with 0 . 1 M citric acid ( pH 3 ) in 1mL fractions into tubes containing 100μL 1M Tris-HCl ( pH 9 ) . Fractions with high IgG content were pooled and buffer exchanged into 1x PBS using Amicon Ultra-4 centrifugal units with 30kDa membrane cutoffs ( Millipore , Billerica , MA ) . IgG concentrations were determined using a NanoDrop spectrometer ( Thermo Fisher Scientific , Waltham , MA ) and antibody size confirmed via SDS-PAGE/Western Blot expression . BLI was performed on purified biotinylated IgG using an Octet Red instrument ( ForteBio , Inc . , Menlo Park , CA ) . Antibodies were biotinylated in water using the EZ-Link ( NHS-PEG4-Biotin ) Kit according to manufacturer’s instructions ( Thermo Fisher Scientific , Waltham , MA ) . Biotinylated antibodies were buffer exchanged with 1x PBS and purified via Amicon Ultra-4 centrifugal units with 30kDa membrane cutoffs ( Millipore , Billerica , MA ) . IgG concentrations were determined using a NanoDrop spectrometer ( Thermo Fisher Scientific , Waltham , MA ) . For these assays , gp140 trimeric Env forms were resuspended at 80nM in 1x Kinetics Buffer . Streptavidin ( SA ) biosensors ( ForteBio , Inc . , Menlo Park , CA ) were activated by immersion into 1x Kinetics Buffer ( 1x PBS , 0 . 1% BSA , 0 . 02% Tween-20 , 0 . 005% NaN3 ) for 10m . Biotinylated IgGs ( at 10μg/mL in 1x KB ) were immobilized on SA biosensors for 300s , and then biosensors were re-immersed in 1x Kinetics Buffer for 60s to establish a ‘baseline’ . Biosensors were then immersed into wells containing NLGS-3 Core rEnv gp140 or NLGS-3 Core rEnv gp140 previously incubated with saturating concentrations ( 160nM in 1x Kinetics Buffer ) of mature VRC01 , mature b12 , or CD4-IgG [ ( CD4-IgG obtained through the National Institutes of Health ( NIH ) AIDS Research and Reference Reagent Program , Division of AIDS , National Institute of Allergy and Infectious Diseases , NIH ( cat . no . 11780; contributors: Progenics Pharmaceuticals , Inc ) ] . After an association phase of 300s , SA biosensors were re-immersed into wells containing only 1x Kinetics Buffer for dissociation for 600s . Binding shift ( nm ) was determined by alignment to baseline , interstep correction to dissociation , and final processing with Savitsky-Golay Filtering . Heat-inactivated serum from immunized animals , supernatants from Env-specific sorted B cells ( see above ) , or monoclonal antibodies ( MAbs ) , were tested for neutralizing activity using the TZM-bl ( also known as JC53BL-13 cells , a Henrietta Lacks ( HeLa ) cell clone engineered to be CXCR4-positive was obtained from the Center for AIDS Reagents ( CFAR ) ) cell line-based neutralization assay , as previously described [31 , 76] . Briefly , MAbs ( starting concentration 50μg/mL or 200μg/mL ) , B cell supernatant ( starting at 1:2 dilution ) , or sera ( starting at 1:10 ) were serially diluted 3-fold for at a final volume of 30μL . 30μL of pseudovirus , previously determined to result in ~2 x 105 luciferase units per well , was added to each well for 90 min at 37°C . 50μL of the pre-incubated mixture was added to TZM-bl cells , previously incubated with polybrene for 30min at 37°C . 72 hours later , the media were removed and Steady-Glo Luciferase reagent ( Promega , Madison , WI ) was added . All assays were performed in duplicate or triplicate as indicated . Significant difference were assessed by ANOVA via Prism v6 software . Bioinformatics analysis significant differences were calculated in various R packages as indicated above . Respective p-values and FDR values are indicated in figure legends .
|
Broadly neutralizing HIV-1 antibodies ( bNAbs ) display protective potentials against experimental animal infection and thus are believed to be a key component of an effective HIV vaccine . bNAbs are derived from B cells that express B cell receptors formed by specific VH/VL alleles . We report that the variable domains of recombinant HIV-1 Env immunogens activate a large number of B cell clones that give rise to many non-neutralizing antibodies , and that removing the variable domains from the immunogen reduces the number of activated B cell lineages and leads to the development of autologous neutralizing antibodies , a step towards bNAb-production . Our findings shed new light into how HIV-1 evades detection from B cells that can produce bNAbs and also provides information that is relevant for the design of optimal immunization strategies .
|
[
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"blood",
"cells",
"medicine",
"and",
"health",
"sciences",
"immune",
"cells",
"immune",
"physiology",
"immunology",
"microbiology",
"vertebrates",
"viral",
"structure",
"animals",
"mammals",
"primates",
"molecular",
"biology",
"techniques",
"antibodies",
"antibody",
"response",
"old",
"world",
"monkeys",
"research",
"and",
"analysis",
"methods",
"sequence",
"analysis",
"immune",
"system",
"proteins",
"white",
"blood",
"cells",
"monkeys",
"artificial",
"gene",
"amplification",
"and",
"extension",
"animal",
"cells",
"proteins",
"viral",
"core",
"bioinformatics",
"antigens",
"biological",
"databases",
"molecular",
"biology",
"immune",
"response",
"macaque",
"antibody-producing",
"cells",
"biochemistry",
"eukaryota",
"sequence",
"databases",
"cell",
"biology",
"b",
"cells",
"polymerase",
"chain",
"reaction",
"virology",
"physiology",
"database",
"and",
"informatics",
"methods",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"amniotes",
"organisms"
] |
2018
|
B cell clonal lineage alterations upon recombinant HIV-1 envelope immunization of rhesus macaques
|
We propose a working hypothesis supported by numerical simulations that brain networks evolve based on the principle of the maximization of their internal information flow capacity . We find that synchronous behavior and capacity of information flow of the evolved networks reproduce well the same behaviors observed in the brain dynamical networks of Caenorhabditis elegans and humans , networks of Hindmarsh-Rose neurons with graphs given by these brain networks . We make a strong case to verify our hypothesis by showing that the neural networks with the closest graph distance to the brain networks of Caenorhabditis elegans and humans are the Hindmarsh-Rose neural networks evolved with coupling strengths that maximize information flow capacity . Surprisingly , we find that global neural synchronization levels decrease during brain evolution , reflecting on an underlying global no Hebbian-like evolution process , which is driven by no Hebbian-like learning behaviors for some of the clusters during evolution , and Hebbian-like learning rules for clusters where neurons increase their synchronization .
A plethora of phenomena in nature can be effectively described by networks . Neuroscientists have used tools for the analysis of complex networks that help realize even more deeply the functionality and structure of the brain . It was found that many aspects of brain network structures are typical of a wide range of non-neural or non-biological complex networks [1 , 2] . One of the main findings in neuroscience is the modular organization of the brain , which in turn implies an inherent parallel nature of brain computations [1] . Modular processors have to be sufficiently isolated and dynamically differentiated to achieve independent computations , but also globally connected to be integrated in coherent functions [1] . It has been revealed that the cortical network is a hierarchical and clustered network with a complex connectivity [3] . A possible network description for this modular organization is that brain networks may be small-world structured [4] with properties similar to many other complex networks [5] . This viewpoint has been driven by the systematic finding of small-world topology in a wide range of human brain networks derived from structural [4] , functional [6] , and diffusion tensor MRI [7] studies . Small-world topology has also been identified at the cellular-network scale in functional cortical neural circuits in mammals [8] and also in the nervous system of the nematode Caenorhabditis elegans ( C . elegans ) [9] . Moreover , this topology seems to be relevant for the brain function because it is affected by diseases [10] , normal ageing , and by pharmacological blockade of dopamine neurotransmission [11] . Synchronization is ubiquitous in nature . Insightful findings regarding synchronization in complex networks were reviewed recently in Ref . [12] . Recently , synchronization in complex modular or clustered networks has been investigated [13 , 14] . It appears as the interplay between the intrinsic dynamics associated to the nodes of the network and its graph topology and connecting functions . In this work , synchronization will be considered as a mean to quantify functional behaviors of the brain dynamical networks ( BDNs ) studied . By BDN we mean a network that represents the connectome equipped with neural dynamics for its nodes to account for their time evolution . Mathematical and computational approaches have a long tradition traced back to the early mathematical theories of perception [15] and of current integration by a neural cell membrane [16] . Hebb’s idea on assembly formation [17] inspired simulations on the largest computers available at that time ( 1956 ) [18] to understand the relation between neural connectivity strength and response , i . e . behavior . It is a learning rule which proposes an explanation for the adaptation of neurons during the learning process . The simultaneous activation of pairs of neurons leads to pronounced increases in their synaptic strength . There is also the possibility of many other kinds of learning [19] . In this work we find evidence of Hebbian-like and no Hebbian-like learning rules characterized by the relationship between the addition of synapses and the increase or decay in the synchronization behavior of neurons in the evolved BDNs . Several brain models have been studied so far that do not take into account particular behaviors of isolated neurons , but higher level functions of the external stimulus and the behavioral response attributed to ensembles of neurons of cortical areas . There are two classes: Those based on collective functional dynamics of local groups of neurons , such as the Wilson-Cowan model for the cortical and thalamic nervous tissue [20] , and those based on the conditional probabilities ( as well as information and mutual information ( MI ) ) of stimuli and responses , prominent examples of which are the Bayesian brain hypothesis [21] and the infomax theory [22] . Recently , it was shown that most of the probabilistic brain models can be unified under a single free-energy principle [23] , the one that interprets the brain as a system that tries to minimize surprises of the sensations from the world . If the brain learns by maximizing the MI between stimuli and response [22] , or by updating the internal model of probabilities by using Bayesian techniques [21] , or by minimizing the free-energy [23] , the surprise of the response provides little insight about the dynamical mechanisms appearing in a brain network when it evolves based on such principles . The drawback however of the probabilistic brain models is that they describe little about the underlying dynamical structure of what is really happening in the neural level and the representation of the stimuli [24] . For deterministic systems with correlations , an appropriate quantity for measuring the transfer of information is the Mutual Information Rate ( MIR ) , the MI per time unit . In Ref . [25] , the authors have developed alternative methods to overcome problems that stem from the definition of probabilities from time series and derived an upper bound for the MIR , Ic , between two nodes of a complex dynamical network from time averages that do not rely on probabilities , but instead on the two largest Lyapunov exponents l1 , l2 of the subspace of the network formed by the two nodes ( see Eq ( 10 ) in Materials and Methods ) . In our study , Ic stands for the upper bound for the information transferred per time unit between any two nodes of the BDN , what represents the information flow capacity of the BDN . We discuss more on these in Materials and Methods , Section Upper Bound for MIR . Inspired by the infomax theory and by theoretical studies that proposes that the maximization of information transmission between subsystems can be used as a principle for understanding the development and evolution of complex brain networks ( see Ref . [2] and references therein , and Ref . [26] ) and , aiming at elucidating the microscopic dynamic mechanisms associated to the evolution of brain networks , we propose a working hypothesis and , provide evidence that a brain dynamical network may evolve based on the maximization of the information flow capacity it can internally handle at each step of its evolution process , i . e . a new inter-neuron connection is established if it leads to a subsequent increase of the information flow capacity of the new brain circuitry . Our hypothesis is based on the internal neural network dynamics and on a plausible model for brain structure per se without the need to resort to probabilistic models based on the external influence in the brain and its response . We have been able to show that our evolved BDNs present similar synchronization and information flow capacity behaviors with those found for the simulated dynamical networks for the brain structure of the C . elegans and humans . Moreover , we show that BDNs evolved with coupling strengths that maximize the information flow capacity are the ones with the smallest spectral graph distance from the BDNs of the C . elegans and humans , and that , during the growing process , their information flow capacity increase is related to moderately low amounts of global neural synchronization . This work provides ample evidence that brain networks may grow by maximizing the capacity of information flow they can internally handle , driven by global no Hebbian-like evolution processes , according to which the addition of interconnections between clusters during the evolution process leads to a decrease in the global synchronization level of the BDN . This behavior is accompanied by a similar no Hebbian-like learning process for neurons in some of the clusters and by Hebbian-like processes for neurons in the remaining clusters , leading to an increase in the synchronization level between these neurons during brain network evolution . Effectively , the no Hebbian-like mechanism is akin to the unlearning anti-Hebbian mechanism of Crick and Mitchison [27] that refers to the elimination of unnecessary connections to prevent overload and to render the network more efficient . In our evolution we do not delete links . However , both mechanisms lead to a decrease of synchronization and to more efficient networks , being in our case the evolved networks able to maximize their information flow capacity . Global synchronization takes into consideration the synchronous behavior of all neurons in the BDN whereas local synchronization of neurons in a cluster of the BDNs . Finally , our work shows that optimising information flow capacity leads to evolved networks that are heterogeneous , in accordance with the line of research in Ref . [2] , where the authors report on a mathematical model for the evolution of heterogeneous modules in the brain based on the maximization of bidirectional information flow transmission .
We present in panels ( A ) , ( B ) of Fig 1 the global synchronization measure ρ and upper bound of information flow Ic for the BDN of the C . elegans in the parameter space of chemical coupling gn in [0 , 2] and electrical coupling gl in [0 , 2] . We refer the reader to Materials and Methods , Subsection C . elegans Data . A direct comparison between the two panels reveals a number of conclusions for the different parameter space regions . At first , for relatively high chemical and electrical couplings almost full global synchronization can be achieved ( yellow and red regions in Fig 1 ( A ) ) . For the same region , panel ( B ) shows an almost absence of capability of information transmission as the upper bound for MIR , Ic ≈ 0 ( dark blue region ) . High levels of global synchronization accompanied by low values of Ic indicate that not only neural activities are similar but also they have very low entropy since both Lyapunov exponents λ1 , λ2 are practically zero leading to their difference Ic ≈ 0 ( see Materials and Methods , Subsection Upper Bound for MIR ) . Second , for chemical couplings smaller than 0 . 3 ( i . e . gn ∈ [0 , 0 . 3] ) and electrical couplings gl ∈ [0 , 2] , we observe a multitude of different functional behaviors: There are regions of high synchronization ( red region in panel ( A ) ) and low Ic ( blue region in panel ( B ) ) and others with exactly the opposite behavior ( i . e . low global synchronization accompanied by high information flow capacity ) . These different functional behaviors will become even more evident in Fig 2 , where we plot the regions between the left vertical axes and the white dotted lines of Fig 1 in a finer resolution . Our findings suggest that for the C . elegans BDN , for most of the chemical and electrical couplings , global brain synchronization is roughly speaking inversely related to the information flow capacity , Ic . As we shall see in Materials and Methods , Section Similarities between C . elegans and human BDNs for a more detailed analysis of the C . elegans and human BDNs , high synchronization as depicted by ρ implies small information flow capacity . The relation between ρ and Ic can be better understood in the basis that typically ρ ∝ 1/λ2 , where λ2 is the second largest Lyapunov exponent of the BDN , for the so-called non-excitatory networks [28 , 29] . The non-excitatory character is expected to be prominent when the electrical coupling has a dominant contribution to the behavior of the network with respect to the chemical , a situation that promotes global neural synchronization . We have performed a similar study for the global synchronization ρ and upper bound of information flow Ic for the human brain networks in Materials and Methods , Subsection Human Subjects Data , shown in Fig 1 ( C ) and 1 ( D ) . Particularly , we first prepared parameter spaces for all six human subject BDNs with the same coupling ranges with those of the first two panels of the C . elegans and then computed their average , presented in Fig 1 ( C ) and 1 ( D ) . The averaged ρ and Ic quantities from the six subjects are indicated by ⟨ρ⟩6 and ⟨Ic⟩6 , respectively . A direct comparison between panels ( C ) and ( D ) of Fig 1 for the humans reveals a number of parameter space regions associated to different functional behaviors , similar to those observed for the C . elegans . For relatively high chemical and electrical couplings almost full global synchronization is achieved ( yellow and red regions in Fig 1 ( C ) ) since ⟨ρ⟩6 ≈ 1 . For the same region , Fig 1 ( D ) shows an almost absence of information flow capacity as ⟨Ic⟩6 ≈ 0 , the reason being the same as for the C . elegans BDN . For chemical couplings smaller than 0 . 2 ( i . e . gn ∈ [0 , 0 . 2] ) and electrical couplings gl ∈ [0 , 2] we observe regions of high synchronization ( red regions in Fig 1 ( C ) ) and low ⟨Ic⟩6 ( blue region in Fig 1 ( D ) ) , as well as others with exactly the opposite property , having low synchronization and large ⟨Ic⟩6 . In order to look deeper into the details of the functional behaviors of these BDNs and , to understand the structural and functional similarities between them , we study in Materials and Methods , Section Similarities between C . elegans and human BDNs , zoom-in plots of the previous parameter spaces . There has been enormous research devoted on the C . elegans worm which has revealed its ability to learn about mechano , chemo and thermosensory inputs and stimuli [30 , 31] . It was also shown that its neural system has the ability to distinguish between tastes , odours or any indication related to the presence or absence of food . It also shows different kinds of learning behavior , such as associative ( classical conditioning and differential classical conditioning ) , and non-associative forms of learning , such as habituation and dishabituation [32] . These properties are reminiscent of the human brain ability to adapt to different stimuli and environments . In Fig 2 , we present a finer resolution version of the parameter spaces of Fig 1 for the C . elegans and humans . They allow us to reveal the extraordinary similarity on the functional level , the global synchronization and Ic patterns between the C . elegans and human BDNs . In these plots , wherever we observe high synchronization as evidenced by orange and red regions in panel ( A ) for the C . elegans and ( C ) for the human brain network , the upper bound for MIR , Ic , that stands for the internal information flow capacity of the BDN , is small ( blue region ) and vice-versa . Since the C . elegans brain connectivity network is about four times smaller in size than the human BDNs [33] studied here , we should expect that the global synchronization and upper bound for MIR patterns could occur for different ranges of chemical and electrical couplings as compared to those of Fig 1 ( see for example Ref . [34] ) . Therefore , a rescaling of the coupling strengths was employed to allow for both BDNs to have the possibility to produce equivalent dynamical behaviors . This rescaling is described in Materials and Methods , Subsection Rescaling of Chemical and Electrical Couplings for Parameter Spaces of Networks with Different Eigenvalue Spectra . It is worth noting that there is an optimal coupling range for both BDNs that allows for large information flow capacity in the brain networks ( orange and yellow regions in panels ( B ) , ( D ) of Fig 2 ) , a coupling range that promotes moderately low global synchronization ! In Materials and Methods , Subsection A Model for Brain Network Evolution Based on the Maximization of Information Flow Capacity , we propose an artificial brain network evolution model that presents important structural and functional properties of the BDNs of the C . elegans and humans . It is based on the combined effect of chemical and electrical synapses and , on a topology reminiscent of interconnected brain communities found in these BDNs . We use chemical synapses for the communication of neurons of different clusters ( inter-cluster connections ) and electrical for the communication of neurons within each cluster ( intra-cluster connections ) . Here , we compare the functional properties of this brain network evolution model with those of the C . elegans and humans . By functional we mean the properties of the dynamics of the BDN such as local , global synchronization and information flow capacity . Particularly , we report on the similarities we found for the functional properties of the BDNs of the C . elegans and humans , and for those of the proposed model for brain network evolution . To support further the validity of the presented results for the model for brain network evolution and to show its independence on the particular initial small-world cluster configuration , we computed its parameter spaces averaging over the functional measurements obtained from five different initial small-world cluster configurations ( for a discussion about the creation of small-world networks or clusters see Materials and Methods , Subsection Analysis of Networks and Communities ) . In all cases , we used the same evolution process as described in Materials and Methods , Subsection A Model for Brain Network Evolution Based on the Maximization of Information Flow Capacity . The results of this study are shown in Fig 3 , where ⟨ρ⟩5 stands for the average of the global synchronization of the evolved BDNs for the five realizations and ⟨mMIR⟩5 for the average maximal value of the MIR for the same realizations . We first provide evidence in panel ( A ) and ( C ) to ( H ) that the finally evolved BDN of small-world clusters captures similar local and global synchronization properties to those of the C . elegans BDN . The local synchronization ρci of the i-th community ( Fig 3 ( C ) –3 ( H ) ) also reproduces similarly the global synchronization patterns of Fig 2 ( A ) and 2 ( C ) for the C . elegans and human BDNs . Comparing these panels , we conclude that the averaged synchronization measure ⟨ρ⟩5 of the brain network evolution model ( Fig 3 ( A ) ) attains almost similar values in the same coupling regions to those of the C . elegans in Fig 2 ( A ) . The different ranges on the horizontal axes of the chemical coupling strength gn can be explained by the fact that the finally evolved network consists of 60 neurons whereas the C . elegans of 277 neurons ( for more details see Materials and Methods , Subsection Rescaling of Chemical and Electrical Couplings for Parameter Spaces of Networks with Different Eigenvalue Spectra ) . It is remarkable that the finally evolved BDN exhibits almost identical synchronization and information flow capacity features as the BDN of the C . elegans and humans for almost all coupling ranges considered . Particularly , focusing on panels ( A ) , ( B ) ( for the C . elegans ) and ( C ) , ( D ) ( for the humans ) of Fig 2 and , on ( A ) , ( B ) of Fig 3 ( for the brain network evolution model ) , we observe an almost identical pattern of global synchronization and information flow capacity as depicted by Ic and its averages . Again , here we have used the rescaling in Materials and Methods , Subsection Rescaling of Chemical and Electrical Couplings for Parameter Spaces of Networks with Different Eigenvalue Spectra , to create networks that can potentially reproduce similar dynamical behaviors for the finally evolved BDNs , in agreement with those identified for the C . elegans and human BDNs earlier . The upper bound for MIR , Ic , depends on various factors , such as network topology and connectivity patterns , coupling strengths and types ( chemical , electrical ) , synchronicity , etc . The values of Ic , the information flow capacity , can be comparable ( or different ) for networks with different topologies . This is due to the fact that Ic is a function of the chemical and electrical coupling strengths . The surprising fact is that as we evolve a network with an initial small-world network configuration , by maximizing information flow capacity , the final network exhibits not only similar topological ( structural network characteristics ) but also similar functional or behavioral ( synchronization and upper bound for MIR ) properties as those found for the brain dynamical networks of the C . elegans and human subjects ( see Materials and Methods , Subsection Spectral Similarity of C . elegans and Human Brain Networks with those of the Model for Brain Network Evolution ) . For completeness , in S2 Fig , we present a similar analysis to the one of Fig 3 based on Erdős-Rényi random networks ( panels ( A ) , ( B ) ) , scale-free ( Barabási-Albert ) ( panels ( C ) , ( D ) ) and star topologies perturbed by 20% for the clusters of the model for brain evolution of 60 neurons and 6 small-world clusters ( panels ( E ) , ( F ) ) and found out that the evolution model fails to capture similar functional ( local and global synchronization and , information flow capacity patterns in the parameter spaces ) as the same model equipped with small-world topologies for its clusters ( see Fig 3 ) . Here , we are interested in studying and proposing a model for brain network evolution that is able to reproduce not only similar functional properties such as information flow capacity and , local and global synchronization properties , but also importantly to reproduce similar structural properties for the finally evolved full brain network and of its clusters . Based on our results so far , we show next that the initial cluster configuration able to fulfil both requirements is the small-world cluster topology . The discussion about the results of Fig 4 in Materials and Methods suggest that the evolution process of a basic small-world clustered network is capable of generating an evolved one with similar structural properties with those for the C . elegans and human BDNs ( see Materials and Methods , Subsection Structural Properties of the Model for Brain Network Evolution ) . The structural similarity to the human brain network is even more remarkable if the couplings of the network to be evolved are within the range that promotes high levels of information flow capacity and low neural synchronization , a prominent example of which is case 𝓑 ( for a definition and discussion about cases 𝓐 , 𝓑 , see Materials and Methods , Subsection Brain Network Evolution Promotes Global no Hebbian-like and , Local Hebbian-like and no Hebbian-like Evolution Learning ) . Here we study how close the normalized Laplacian spectral plots of the networks considered in Fig 4 ( Q ) are from those of evolved BDNs . Particularly , we examined the spectral similarity by comparing spectral plots and computing their average Euclidean distance following Ref . [35] . We refer the reader to Materials and Methods , Subsection Normalized Laplacian Spectra , for the details of our computations . We compared the normalized Laplacian spectral plot of the C . elegans and of the six human brain networks with the spectral plots of all finally evolved networks of the five realizations used to compute the averaged parameter spaces of the model for brain network evolution of Fig 3 ( A ) and 3 ( B ) . We provide additional support by demonstrating in the last four panels of Fig 5 results from a similar comparison between the C . elegans and humans with a double-sized version of 120 neurons of the model for brain network evolution of the Materials and Methods , Subsection A Model for Brain Network Evolution Based on the Maximization of Information Flow Capacity . The normalized Laplacian spectra for the C . elegans and for the averaged over the six human BDNs are shown in Fig 4 ( Q ) . As it can be seen , both spectra share some common features: Both show a left-skewed distribution in which the largest eigenvalue is closer to one in agreement with results in Ref . [35] . Also , the distributions show peaks around one and the eigenvalues are scattered at the beginning of the spectra , suggesting similarities in their community structure , reminiscent of their small-worldness . Although the spectral plots of both cases of the model for brain network evolution shown in Fig 4 ( R ) do not exhibit all properties of the spectral plots of the C . elegans and humans of Fig 4 ( Q ) , maybe due to the considerably smaller size of the former networks , they do exhibit interesting similarities: They are both left-skewed distributions with a peak around 1 . 3 , being closer to 1 than to 2 . We also observe low relative frequency eigenvalues at the beginning of both spectra . Both spectral properties suggest similarities in their community structure [35] as well ( i . e . their small-worldness ) . This is in accordance with the spectral plots of the averaged human and C . elegans brain networks of Fig 4 ( Q ) . We argue that this similarity comes from the small-worldness of the communities . The close relation between the normalized Laplacian spectra of Fig 4 suggests the existence of common underlying structural properties of the neural networks of the C . elegans , the humans and the model for brain network evolution . We measured the similarity between the spectral plots of the C . elegans and the averaged human brain network with the averaged model for brain network evolution and , plot in panels ( A ) , ( B ) of Fig 5 their spectral distance D for different chemical and electrical coupling ranges . In this framework , the closer D is to zero , the closer structurally the compared networks are . In particular , panel ( A ) is the parameter space for the spectral distance between the C . elegans and the averaged model for brain network evolution and , panel ( B ) is a similar plot for the spectral distance between the averaged human brain network and the same model for brain network evolution . We pinpoint case 𝓐 by a ▲ and case 𝓑 by ● . Such results allow one to draw interesting relations between structure and function of the proposed model for brain network evolution with the structural properties of the brain networks of C . elegans and humans . Panels ( A ) and ( B ) of Fig 5 already reveals that the smallest mean spectral distance happens for the pair of chemical and electrical couplings that gives rise to BDNs that present small amount of synchronization and high information flow capacity , in other words to cases such as 𝓑 . In contrast , one of the largest mean spectral distances was found for case 𝓐 that promotes high amount of neural synchronization and small information flow capacity in the brain network !
The complexity of the circuitry of the nervous system of the human brain is still a big challenge to be resolved as it contains about 86 billion neurons and thousands times more synapses . A synapse is a junction between two neurons and it is a mean through which neurons communicate with each other . There are electrical and chemical synapses: An electrical synapse is a physical connection between two neurons which allows electrons to pass through neurons by a very small gap between nerve cells . Electrical synapses are bidirectional and of a local character , happening between neurons whose cells are close . They are believed to contribute to the regulation of synchronization in the brain network . In contrast , chemical synapses are special junctions through which the axon of the pre-synaptic neuron comes close to the post-synaptic cell membrane of another neuron or non-neural cell . In Ref . [37] , the authors report on the self-consistent gap junctions and chemical synapses in the connectome of the C . elegans . In our work , we use both kinds of synapses . Following Ref . [34] , we endow the nodes ( i . e . neurons for the c . elegans and neural ensembles for the humans ) of the networks with Hindmarsh-Rose brain dynamics [38]: p ˙ = q - a p 3 + b p 2 - n + I ext , q ˙ = c - d p 2 - q , n ˙ = r [ s ( p - p 0 ) - n ] , ( 1 ) where p is the membrane potential , q is associated with the fast current , Na+ or K+ , and n with the slow current , for example Ca2+ . The rest of the parameters are defined as a = 1 , b = 3 , c = 1 , d = 5 , s = 4 , p0 = −1 . 6 and Iext = 3 . 25 for which the system exhibits a multi-scale chaotic behavior characterized as spike bursting . r modulates the slow dynamics of the system and was set to 0 . 005 so that each neuron or neural ensemble be chaotic . For these parameters , the HR model enables the spiking-bursting behavior of the membrane potential observed in experiments made with a single neuron in vitro . It is also a relatively simple model that provides a good qualitative description of the many different patterns empirically observed in neural activity . We couple the HR system to create an undirected BDN of Nn neurons connected simultaneously by electrical ( linear diffusive coupling ) and chemical ( nonlinear coupling ) synapses: p ˙ i = q i - a p i 3 + b p i 2 - n i + I ext - g n ( p i - V syn ) ∑ j = 1 N n B i j S ( p j ) - g l ∑ j = 1 N n G i j H ( p j ) , q ˙ i = c - d p i 2 - q i , n ˙ i = r [ s ( p i - p 0 ) - n i ] , ϕ ˙ i = q ˙ i p i - p ˙ i q i p i 2 + q i 2 , i = 1 , … , N n . ( 2 ) In our study , ϕ . i is the instantaneous angular frequency of the i-th neuron and ϕi is the phase defined by the fast variables ( pi , qi ) of the i-th neuron . We consider H ( pi ) = pi and: S ( p j ) = 1 1 + e - λ ( p j - θ syn ) , ( 3 ) with θsyn = −0 . 25 , λ = 10 , and Vsyn = 2 to create excitatory BDNs . Eq ( 3 ) is a sigmoidal function that acts as a continuous mechanism for the activation and deactivation of the chemical synapses and , also allows for analytical calculations of the synchronous modes and synchronization manifolds of the coupled system of Eq ( 2 ) [34] . In Eq ( 2 ) , gn is the coupling strength associated to the chemical synapses and gl to the electrical synapses . For the chosen parameters , we have ∣pi∣ < 2 and that ( pi − Vsyn ) is always negative for excitatory networks . If two neurons are connected under an excitatory synapse then , when the presynaptic neuron spikes , it induces the postsynaptic neuron to spike . We adopt only excitatory chemical synapses in our analysis . We use as initial conditions for each neuron i: pi=−1 . 30784489+ηir , qi=−7 . 32183132+ηir , ni=3 . 35299859+ηir and ϕi = 0 , where η i r is a uniformly distributed random number in [0 , 0 . 5] for all i = 1 , … , Nn , following Ref . [34] . These initial conditions place the trajectory quickly in the attractor of the dynamics and thus , there is less need to consider longer transients . Gij accounts for the way neurons are electrically ( diffusively ) coupled and it is a Laplacian matrix ( i . e . Gij = Kij − Aij , where A is the binary adjacency matrix of the electrical connections and K is the degree identity matrix based on A ) , and so ∑ j = 1 N n G i j = 0 . By a binary adjacency matrix , we mean an adjacency matrix with entries either 0 ( no connection ) or 1 ( connection ) . Bij is a binary adjacency matrix and describes how the neurons are chemically connected and therefore its diagonal elements are equal to 0 , giving thus ∑ j = 1 N n B i j = k i , where ki is the degree of the i-th neuron , i . e . it represents the number of chemical links that neuron i receives from all other j neurons in the network . A positive ( i . e . 1 ) off-diagonal value in both matrices A , B in row i and column j means that neuron i perturbs neuron j with an intensity given by glGij ( electrical diffusive coupling ) or gnBij ( chemical excitatory coupling ) , respectively . Therefore , the binary adjacency matrices C of the BDNs considered in this work are given by: C = A + B . We numerically integrated Eq ( 2 ) in Fortran 90 using the Euler integration method ( order one ) with time step δt = 0 . 01 . We decided to do so to reduce the numerical complexity and CPU time of the required simulations to feasible levels as a preliminary comparison of trajectories computed for the same parameters ( i . e . δt , initial conditions , etc . ) with integration methods of order 2 , 3 and 4 , revealed similar results . We evolve the dynamics of the brain networks and calculate their two largest Lyapunov exponents λ1 , λ2 for the estimation of the upper bound for MIR , Ic . We use the well-known method of Refs . [39 , 40] to compute the Lyapunov exponents needed for the estimation of the upper bound Ic for MIR ( see also Materials and Methods , Subsection Upper Bound for MIR ) . The numerical integration of the HR system of Eq ( 2 ) for the C . elegans and human BDNs was performed for the final integration time tf = 5000 and the computation of the different quantities such as the order parameter ρ of Materials and Methods , Subsection Synchronization Measures in BDNs and the Lyapunov exponents , starts after the transient time tt = 300 to make sure that orbits converged to the attractor of the dynamics . The same parameters for the model of brain network evolution of Materials and Methods , Subsection A Model for Brain Network Evolution Based on the Maximization of Information Flow Capacity were set to tf = 2500 and tt = 300 to reduce the numerical complexity and CPU time to feasible levels , retaining similar results . We have been careful to check that using exactly the same values as for the C . elegans and humans , the conclusions were practically the same . In this work we propose an artificial evolution model for brain network connectivity that captures important structural and functional properties of the BDNs of the C . elegans and humans . Our idea is reminiscent of modular processors that are sufficiently isolated and dynamically differentiated to achieve independent computations , but also globally connected to be integrated in coherent functions [1] . The model is based on the consideration of the combined effect of chemical and electrical synapses between neurons based on a topology reminiscent of interconnected brain network communities found in the BDNs of C . elegans and humans . In this study , we consider chemical synapses solely for the communication of neurons of different clusters of the network ( inter-cluster connections ) and electrical synapses for the communication of neurons within each cluster ( intra-cluster connections ) . This idea comes from the biological local and non-local nature of these connections . Particularly , we consider a starting network topology for brain network evolution , where Nc clusters of electrically coupled neurons are connected in a closed ring as shown in S1 Fig of the Supporting Information . We endow each cluster with a small-world topology [9] as this is what we found to be more plausible to happen on the brain networks of the C . elegans and humans ( see Subsection . Analysis of Networks and Communities in Materials and Methods ) . We also use in our evolution model , for simplicity but this is not mandatory , clusters of the same number of neurons . We denote the total amount of neurons in the network by Nn . Each small-world cluster is connected by an inter-cluster connection with its two nearest neighbour clusters by only chemical excitatory connections . In S1 Fig of the Supporting Information , one can see such an example of a small-world network topology that comprises Nc = 6 clusters and Nn = 60 neurons , where the red links denote the chemical inter-cluster connections and black the electrical intra-cluster connections . For this model network , we subsequently compute the two largest Lyapunov exponents λ1 , λ2 of the BDNs following Refs . [39 , 40] to estimate the upper bound Ic = λ1 − λ2 for the MIR of the network , i . e . the maximum amount of information per time unit that can be exchanged between the neurons of the basic ( not yet evolved ) network , aka its information flow capacity ( for the details see Subsection . Upper Bound for MIR in Materials and Methods ) . We then evolve the starting network , such as the one in S1 Fig , by adding new chemical excitatory inter-cluster connections to simulate the creation of new chemical synapses between neurons of different clusters . The electrical connections , topology and the values of the chemical and electrical coupling strengths are not modified during the evolution process . We adopt the following evolutionary rule to imitate brain plasticity [19]: If the newly added inter-cluster chemical connection leads to an increase of Ic prior to the addition , the new synapse is retained . If , instead , it is found not to increase Ic then it is deleted from the network and the random search for another one starts , being this procedure iterative . We choose the nodes of the different small-world clusters so that to simulate the addition of new inter-cluster connections in a random fashion ( i . e . the candidate links are randomly chosen from a uniform distribution ) . The iterative procedure is repeated until the maximum number of possible pairs of neurons from different clusters is exhausted . We denote by mMIR the value of Ic of the finally evolved BDN which is always bigger or equal than the Ic of the starting BDN . For different values of the coupling strengths , mMIR can be achieved for different numbers of added interconnections . In all cases studied , we also compute the global synchronization measure ρ of the finally evolved BDN ( for the details , see Materials and Methods , Subsection Synchronization Measures in BDNs ) to allow for direct comparisons with mMIR and for the identification of relations between synchronization and information flow capacity in the BDNs . Following Ref . [34] , a rough estimation on the range of chemical and electrical couplings based on those used for the parameter space of another network that is capable of reproducing similar synchronous behaviors [28] and similar amounts of Kolmogorov-Sinai entropy [29] , can be computed as following: Suppose we have produced a parameter space such as those of Fig 2 showing behaviors of the BDN of the C . elegans as a function of gn and gl . Let us denote the maximum electrical coupling as g l C , the maximum chemical as g n C , the smallest positive eigenvalue of the Laplacian matrix of the electrical connections as ω m C and the average degree of the chemical connections as d ‾ C . Suppose now we want to compute a similar parameter space for another BDN . Let us denote by g l max , g n max , ω m max and d ‾ max the corresponding values of the new parameter space . Then , for the new maximum couplings we have: g n max = ( d ¯ C d ¯ max ) g n C ( chemical coupling ) , ( 4 ) g l max = ( ω m C ω m max ) g l C ( electrical coupling ) . ( 5 ) To arrive at Eqs ( 4 ) and ( 5 ) using the results of Ref . [34] , we have assumed that a network with an average degree d ‾ for its chemical connections behaves similarly to a network with the same degree for its chemical connections . Eqs ( 4 ) and ( 5 ) provide a rough estimation on the maximum coupling strengths that can be used for the new parameter spaces . S1 Table presents ωm , d ‾ , g n max and g l max for the different BDNs considered in our work . Based on these rough predictions , we then identified as best matching ranges , those depicted in the figures of the paper . Based on the maximum values of the parameter space ranges of Fig 2 ( A ) for the C . elegans ( g n C = 0 . 3 and g l C = 2 ) , we get for the average humans g n max = 0 . 21 and g l max = 1 . 71 , in good agreement with the maximum values of the ranges in Fig 2 ( C ) and 2 ( D ) . For the model for brain network evolution we get g n max = 1 . 14 to 2 . 28 and g l max = 0 . 97 to 1 . 17 depending on the particular BDN , in accordance with the range of couplings used in Fig 3 and , 2 in S2 Fig . Similarly , for the large version of our model for brain network evolution ( Nn = 120 , Nc = 6 ) we estimated g n max = 2 . 16 and g l max = 2 . 9 , consistent with the maximum coupling strengths used in the last four panels of Fig 5 . Consequently , our methodology allowed us to identify regions of synchronization ρ and upper bound for MIR , Ic , for the different BDNs of this work with similar functional and structural properties . We initially identified the communities of the networks using the walktrap method [41] of the igraph software with six steps . The algorithm detects communities through a series of short random walks , with the idea that the vertices encountered on any given random walk are more likely to be within a community . The algorithm initially treats all nodes as communities of their own , then merges them into larger communities , and these into still larger , and so on . Essentially , it tries to find densely connected subgraphs ( i . e . communities ) in a graph via random walks . The idea is that short random walks tend to stay in the same community . Following this procedure we have been able to identify 6 communities in the C . elegans BDN , 10 in human subject A1 , 5 in A2 , 9 in B , 6 in C , 10 in D and finally , 7 in E . After this step , we computed various statistical quantities such as the global clustering coefficient , the average of local clustering coefficients , the mean shortest path , the degree pdf of the network and the small-worldness measure . The latter property is characterized by a relatively short minimum path length on average between all pairs of nodes in the network , together with a high clustering coefficient . Even though small-worldness captures important aspects of complex networks at the local and global scale of the structure , it does not provide information about the intermediate scale . Properties of the intermediate scale can be more completely described by the community structure or modularity of the network [42] . The modules of a complex network , also-called communities , are subsets of nodes that are densely connected to other nodes in the same module but sparsely connected to nodes belonging to other communities . Since nodes within the same module are densely intra-connected , the number of triangles in a modular network is larger than in a random graph of the same size and degree distribution , while the existence of a few links between nodes in different modules plays the role of topological shortcuts in the small-world topology . Systems characterized by this property tend to be small-world networks , with high clustering coefficient and short path length with respect to random networks . To infer the small-worldness of a network or community , we first compute the mean local clustering coefficient C of the network or community and the mean of the local clustering coefficients of one hundred randomly created networks ⟨Cr⟩100 of the same degree pdf with the studied network and also , the mean shortest path of the studied network L and the average of the mean shortest paths of the same one hundred random networks ⟨Lr⟩100 . Watts and Strogatz [9] measured that many real-world networks have an average shortest path length comparable to those of a random network ( L ∼ Lr ) , and at the same time a clustering coefficient significantly higher than expected by random chance ( C ≫ Cr ) . Then , they proposed a novel graph model , currently named the Watts-Strogatz model , with a small average shortest path length L , and a large clustering coefficient C . We adopt this as a working definition of a small-world network or community . Therefore , small-world networks are in between the limit cases of regular graphs with large L and C and random networks with small L and C . To quantify small-worldness we use the ratios [43]: μ = L ⟨ L r ⟩ 100 , γ = C ⟨ C r ⟩ 100 , ( 6 ) in such a way that , for a small-world network or community , we compute: σ = γ μ > 1 , ( 7 ) being the small-worldness measure . The higher is σ from unity for a given network or community , the better it displays the small-world property . For completeness , we note that for the human subject D , the walk-trap community analysis with step equal to six detected eleven communities for which the last one comprised only one neuron . Hence , in all computations , we disregarded this community as a trivial case . We present the results of the above analysis in S2 Table . We have performed all structural analyses of this paper using the igraph software . The evolution of the basic network of S1 Fig , under the principle of the maximization of the information flow capacity , is able to capture behaviors of real brain connectivity networks such as those for the communities of the C . elegans BDN , being their small-world structure a prominent reason . We found out that in all networks studied , the small-world measure σ gets values much higher than unity ( see S2 Table ) , clearly indicating that they all display the small-world property , though in different degrees . Here , based on the extraordinary functional similarities found so far , we focus on two characteristic behaviors of the model for brain network evolution defined in Materials and Methods , Subsection A Model for Brain Network Evolution Based on the Maximization of Information Flow Capacity , to reveal important modular behaviors and underlying learning evolution processes: We associate the first one to the case of high global synchronization and low information flow capacity ( which we call case 𝓐 ) and the other one to the opposite situation of low synchronization and high information flow capacity ( case 𝓑 ) . For illustration purposes , we select from the proposed model for the brain network evolution of 60 neurons and 6 clusters of Materials and Methods , Subsection A Model for Brain Network Evolution Based on the Maximization of Information Flow Capacity , two finally evolved BDNs with the following coupling strengths: For case 𝓐 the pair gn = 0 . 2 , gl = 1 . 8 from the fifth realization ( indicated by ▲ in Figs 3 and 5 ) and for case 𝓑 the pair gn = 0 . 9 , gl = 1 . 5 from the fourth realization ( indicated by ● in Figs 3 and 5 ) . We have checked that the conclusions are valid independently of the realization and other similar pairs of coupling strengths . Also , our results reported here are independent on the initial small-world cluster configuration and initial conditions . In panels ( A ) , ( B ) of Fig 6 we demonstrate the relation between global synchronization ρ and mMIR for these cases . Case 𝓑 of moderately low global synchronization ( dashed black curve in panel ( A ) ) and high information flow capacity ( dashed black curve in panel ( B ) ) is characterized by a larger number of added interconnections ( i . e . 30 ) present in the finally evolved network with respect to case 𝓐 of only 12 ( corresponding to the solid black curves in panels ( A ) , ( B ) ) ! In both cases , ρ of Eq ( 8 ) for global synchronous behavior , has the tendency to decrease with different slopes ( denoted by θ in Fig 6 ( A ) ) during brain network evolution as is evident by the fitting to the data in dashed blue lines . We attribute this behavior to an evolutionary brain network behavior , where global neural synchronization levels decrease during brain network evolution , reminiscent of a possible underlying global no Hebbian-like evolution process that promotes a decrease in global synchronization levels as new connections are added in the network . We regard this as an important global property of the model for brain network evolution that needs to be further clarified , as we need to account for what happens on the local , cluster level as well . We thus use the following procedure to infer about the underlying learning rules for the clusters: During each step of the evolution process of the model of 60 neurons and 6 small-world clusters ( which is a particular BDN ) , we compute ⟨ρij⟩cl based on Eq ( 9 ) in Materials and Methods , to account for the average pair-wise synchronization of cluster cl , l = 1 , … , 6 . At the end of the time evolution , we have six such values for the clusters , lying in the interval [0 , 1] . We then record these values if the newly added chemical interconnection leads to an increase of the information flow capacity as depicted by the corresponding Ic and disregard them if not . At the end of the process , we result with a relationship between ⟨ρij⟩cl and the added interconnections that maximizes Ic , for all clusters ( see panels ( C ) , ( D ) of Fig 6 ) . Doing so , we can account for the underlying cluster learning processes . In particular , the terminology “Hebbian-like” is employed to represent a learning rule that not only involves quantities for synchronous events defined between pairs of neurons ( see Eq ( 9 ) ) , but also a direct relation between synapse strength and synchronization increase . The “no Hebbian-like” terminology refers to a learning rule that involves not only a global measure of synchronous behavior ( see Eq ( 9 ) ) , but also refers to a phenomenon where the addition of synapses is accompanied by a decrease in the synchronization levels . We present these results in Fig 6 ( C ) and 6 ( D ) . Following our considerations , we already know that globally , both finally evolved BDNs are following a no Hebbian-like learning rule . However , panels ( C ) , ( D ) reveal a substantial difference for the synchronization behavior for pairs of neurons in the clusters in the two cases that allows us to assign different kinds of learning rules to them . For case 𝓑 in panel ( D ) , the slopes θcl of the fitted lines ( in blue ) to the data ⟨ρij⟩cl , l = 1 , … , 6 ( in black ) show that the third and fifth cluster have the tendency to increase their internal synchronization level as the BDN evolves . In other words , neurons belonging to these two clusters follow a Hebbian-like learning process that comes in contrast to the no Hebbian-like learning behavior of neurons belonging to the other clusters as they show the opposite trend during brain network evolution , exhibiting negative slopes for the synchronization . These findings are in agreement with the conclusions drawn from panel ( F ) in which neurons belonging to the third and fifth cluster are seen to be more synchronized with respect to the others ( in yellow and red ) , a situation that promotes the modular organization in the brain with different internal levels of synchronization . On the other hand , case 𝓐 demonstrates a completely different situation in which the level of global synchronization decreases and cluster synchronization is very strong between neurons in all clusters during the whole brain network evolution , a behavior that results in a network whose neurons are unable to exchange but very small amounts of information . This is a case where the brain network can not transmit information by increasing its synaptic efficacy and thus , it is unable to learn new information ! Panels ( E ) , ( F ) of Fig 6 show the pair-wise synchronization level of the neurons denoted as ρij ( see Eq ( 9 ) in Materials and Methods , Subsection Synchronization Measures in BDNs ) of the finally evolved BDNs for cases 𝓐 and 𝓑 , resulting in networks of 12 and 30 chemical inter-links , respectively . Panel ( F ) of case 𝓑 for moderately low global synchronization and high information flow capacity reveals a clusterized synchronization behavior . The third and fifth small-world clusters show much higher levels of internal synchrony ( yellow and red points ) , i . e . synchronization of pairs of neurons in the same small-world community , with respect to the blue or dark blue points of low neural pair synchronization in other small-world clusters . The same also happens for a small number of pairs of neurons belonging to different clusters ( off-diagonal points ) . One can notice that given the cluster that has the same internal level of synchronization , there can always be found two subnetworks , each belonging to one cluster , that have an equal amount of synchronization . This means that case 𝓑 corresponds to a BDN that has clustered multilayer synchronization , where different clusters become functionally connected with different common behaviors . A situation reminiscent of findings in neuroscience that demonstrate the modular organization of the brain in which modular processors are sufficiently isolated and dynamically differentiated to achieve independent computations , but also globally connected to be integrated in coherent functions [1] . Our results for case 𝓑 ( see Fig 6 ( F ) ) come in contrast to those of 𝓐 of high global synchronization and low information flow capacity shown in Fig 6 ( E ) which demonstrates that all clusters and pairs of neurons attain almost the same state of almost complete synchronization , revealing a highly synchronized brain dynamical network that is not able however to exchange but only very small amounts of information between its different parts ! The no Hebbian-like mechanism is effectively similar to the unlearning anti-Hebbian mechanism of Crick and Mitchison [27] that proposes the elimination of unnecessary connections to prevent overload and to render the network more efficient . Both mechanisms lead however to more efficient networks , being in our case the evolved networks able to maximize their information flow capacity . Typically , low synchronization implies λ1 ≈ λ2 leading to Ic ≈ 0 . In critical points however , Ic tends to be large as long as most of the oscillation modes are stable resulting in a situation where λ1 > 0 and , λ2 ≈ 0 and positive . Thus , maximum Ic at low synchronization corresponds to maximum Ic near the critical point of the dynamics where λ1 > 0 and , λ2 ≈ 0 and positive . Self critical phenomena happen when the network has marginal Lyapunov exponents , in other words at the critical point that results in typically large Ic . In this context , our results are in agreement with the work in Ref . [44] , supporting our findings for the existence of Hebbian and no Hebbian-like learning mechanisms in the level of the communities ( local synchronization ) at self criticality , which is responsible for the maximization of the information flow capacity of the evolved BDNs ( low global synchronization ) , such as in case 𝓑 ( for a similar result see Ref . [45] ) . We do not study our BDNs under different stimuli . Our main hypothesis is that the final topology and behavior of an evolved BDN that maximizes mutual information rate between its neurons is similar to real brain networks , such as those from the C . elegans and human subjects . Since our results for the maximization of the information flow capacity of the evolved BDNs happen when self critical phenomena emerge , they are in agreement with the results reported in Refs . [46 , 47] . We study here the structural properties of the evolved BDNs and compare with those of the C . elegans and human BDNs . As we have already demonstrated , they share common functional properties . We present the results of this study in Fig 4 . Particularly , panel ( A ) shows the degree probability distribution function ( pdf ( k ‾ i ) ) , panel ( B ) the clustering coefficient CC ( k ‾ i ) as a function of the normalized degree k ‾ i , panel ( C ) the average degree k n n ( k ‾ i ) of the neighbors of nodes with degree k ‾ i and panel ( D ) the network with its distinct communities depicted by different color-shaded neighborhoods , for case 𝓐 . In this context , k ‾ i is the normalized with respect to the maximum , degree . Panels ( E ) to ( H ) show similar plots for case 𝓑 . Panels ( I ) to ( L ) are similar plots for the C . elegans brain network and , panels ( M ) to ( P ) for human subject A1 . The plots in the second column show the correlation between different normalized degrees of the network whereas those of the third the tendency of the nodes of a certain normalized degree k ‾ to link with other nodes of a given degree . From the plots of the second column we observe that high degree nodes have the tendency to link with low degree nodes following an exponential dependence ( with different exponents ) implying disassortative mixing by degree . Our results from the second column of Fig 4 suggest that evolving BDNs based on the maximization of the upper bound for MIR ( cases 𝓐 and 𝓑 ) gives rise to disassortative mixing by degree meaning that high degree nodes are preferentially connected to other low degree nodes and low to high degree nodes . It is worth noting that the structural properties of all human subjects are similar . Fig 4 ( I ) –4 ( O ) show that human subject A1 has similar structural properties to the C . elegans structure . Cases 𝓐 and 𝓑 seem also to present strong structural similarities , for the quantities considered in Fig 4 , despite the profoundly different functional behaviors . We found out that case 𝓑 of moderately low global synchronization and high information flow capacity is characterized by a big number of added interconnections ( i . e . 30 ) present in the final network whereas case 𝓐 by only 12 ! We have checked that this relationship between a large ( small ) number of added chemical inter-links with low ( high ) synchronization and large ( low ) information flow capacity happens for all similar cases of the parameter spaces . It is also noteworthy that in case 𝓑 ( see Fig 4 ( H ) ) , the number of clusters of the finally evolved network is reduced by one ( since we start with six and end up with five ) and is in contrast with what is happening in case 𝓐 ! Depending on the couplings which give rise to different functional behaviors , the brain network evolution model is capable of merging different communities , i . e . of restructuring the initial network configuration . Case 𝓑 is a more globally connected network . The previous observation can be quantified in terms of the modularity of the network , i . e . by the strength of the division of the network into modules ( groups , clusters or communities ) . Networks with high modularity have dense connections between the nodes within modules and sparse connections between nodes of different modules , being more clustered . We have used the igraph software for these computations . By applying this idea here , we find that the modularity of the final BDN of case 𝓑 is 0 . 596 , very close to the average modularity of the six human subjects 0 . 588 ± 0 . 023 . In contrast , the modularity of case 𝓐 is 0 . 702 . For the sake of completeness , we also report the modularity of the C . elegans topology which is 0 . 375 , the smaller of all cases we considered . The last result shows a network with sparser connections between the nodes within the modules and denser between nodes of different modules ! For the C . elegans , the natural evolution process led to a smaller clusterization of its brain network . The results of Fig 4 suggest that the evolution process of a basic small-world clustered network is capable of generating evolved ones with similar structural properties to those for the C . elegans and human BDNs . The structural similarity to the human brain is even more remarkable if the couplings of the network to be evolved are within the range that promotes high levels of information flow capacity such as in case 𝓑 . Synchronous activity has been observed in neural systems and reported to be associated not only with pathological brain states [48] but also with various cognitive functions [49] . It has been found that burst synchronization of neural systems may be strongly influenced by many factors , such as coupling strengths and types [50] , noise [51] , and the existence of clusters in neural networks [14] . In this paper we use the order parameter ρ to account for the synchronization level of the neural activity of the studied BDNs and of their communities [52] . It is originated from the theory of measures of dynamical coherence of a population of Nn oscillators of the Kuramoto type [53] and , can be computed by a complex number z ( t ) defined as: z ( t ) = ρ ( t ) e i Φ ( t ) = ∑ j = 1 N n e i ϕ j ( t ) . ( 8 ) By taking the modulus ρ ( t ) of z ( t ) , one can measure the phase coherence of the population of the Nn neurons of the BDN , and by Φ ( t ) to measure the average phase of the population of oscillators . In this context , ϕi is the phase variable of the i-th neuron of the HR system Eq ( 2 ) given by its fourth equation . Actually , one averages over time ρ ( t ) to obtain the order parameter ρ = ⟨ρ ( t ) ⟩t , the tendency of ρ in time . A value of ρ = 1 corresponds to complete synchronization of the oscillators , whereas ρ = 0 to complete desynchronization . We use Eq ( 8 ) , adapted accordingly , wherever in the paper we need to compute the synchronization level of BDNs or clusters . In particular , in the case of BDNs , Nn is the number of neurons of the BDN and j runs through all Nn neurons of that network whereas in the case of clusters , Nn represents the number of neurons of the particular cluster and j refers to the particular neurons which are members of this cluster . We also compute and plot in Fig 6 ( C ) and 6 ( D ) , for cases 𝓐 and 𝓑 respectively of the model for brain network evolution of 60 neurons and 6 small-world clusters , the pair-wise neural synchronization by looking at the synchronization patterns between all pairs of neurons i , j of the network as: ρ i j = lim Δ t → ∞ | C i j Δ t ∫ τ τ + Δ t e i [ ϕ i ( t ) - ϕ j ( t ) ] d t | , ( 9 ) where Cij is the adjacency matrix of the brain network and ϕi is the phase variable of the i-th neuron of system Eq ( 2 ) . ρij are bounded in the interval [0 , 1] , being ρij = 1 when neurons i , j are fully synchronized and 0 when they are dynamically uncorrelated . To correctly compute ρij , we take the averaging time large enough in order to obtain good measurements of the coherence degree of each pair . We similarly compute and plot in Fig 6 ( C ) and ( D ) for the same cases , the averaged quantity ⟨ρij⟩cl , l = 1 , … , 6 over the 6 clusters , where i , j run through the neurons of each cluster . After Shannon’s pioneering work [54] on information , it became clear that it is a very useful and important concept as it can measure the amount of uncertainty an observer has about a random event and thus provides a measure of how unpredictable it is . Another related concept to the Shannon entropy that can characterize random complex systems is the MI [54] which is a measure of how much uncertainty one has about a state variable after observing another state variable . In Ref . [25] , the authors have derived an upper bound for the MIR between two nodes or two groups of nodes of a complex dynamical network that depends on the two largest Lyapunov exponents l1 , l2 of the subspace of the network formed by these nodes . In particular , they have shown that: MIR ≤ I c = l 1 - l 2 , l 1 ≥ l 2 , ( 10 ) where l1 , l2 are the two finite time and size Lyapunov exponents calculated in the bi-dimensional observation space of the two considered nodes [25 , 55] , which typically should approach the two largest Lyapunov exponents λ1 , λ2 of the dynamical network if it is connected and the time considered to calculate l1 , l2 is sufficiently small . In our study , the upper bound Ic for the MIR between any two nodes of the BDNs is effectively estimated by Ic = λ1 − λ2 ( i . e . l1 = λ1 and l2 = λ2 ) and will stand for the upper bound for the information transferred per time unit between any two nodes of the BDN ( i . e . between the neurons ) , what represents the information flow capacity of the BDN . The phase spaces of the dynamical systems associated to the neural networks we study here are excessively highly multi-dimensional and thus , estimating an upper bound for the MIR using λ1 and λ2 calculated by the methods of Refs . [39 , 40] instead of the MIR itself between all pairs of nodes , reduces enormously the computational complexity of the numerical calculations of this work . Besides , parameter changes that causes positive or negative changes in the MIR are reflected in the upper bound with the same proportion [25] . It is well-known that similarities between the structure of networks can be used for their classification [56] . The architecture of brain networks that describe the organization of maps of connections between neurons and brain elements at a systems level can be achieved by examining the eigenvalue spectrum of the normalized Laplacian of the connectome [35 , 57] . In our study , the connectome is given by the adjacency matrix of the brain network and thus we compute the eigenvalues of the normalized Laplacian based on this matrix . The eigenvalues νi , i = 1 , … , Nn of the normalized Laplacian matrix L are in [0 , 2] which helps to compare networks of different sizes , and is defined as: L i j = { 1 if i = j , - 1 k i if i , j are connected , 0 otherwise , ( 11 ) where i , j represent any two nodes of the brain network , Lij the link between nodes i , j and ki the degree of node i . Therefore , the Laplacian spectrum of the network is given by the set of all eigenvalues of L , namely by its eigenspectrum .
In this paper we propose a working hypothesis , and provide evidence , that neural networks that evolve based on the principle of the maximization of their internal information flow capabilities produce networks whose functional behavior and topology are similar to those features observed in dynamical neural networks whose topology is provided by the C . elegans and humans . Our hypothesis goes along the lines of the infomax theory that proposes that the brain evolves by maximizing the mutual information between external stimuli and its response . When maximizing the internal information flow capacity , we are creating a network capable of processing information about external stimuli for which its information content is smaller than the information flow capacity of the evolved network . Notably , the brain evolves by the action of input signals . Our working hypothesis simplifies enormously the complexity of the involved calculations and , allow us to understand function and behavior in the brain , without the need to externally perturb non-autonomous neural networks . We have been able to show that our evolved brain networks present similar synchronization and information flow capacity behaviors with the ones found for the simulated dynamical networks for the structure of the C . elegans and human brain . Moreover , we have shown that BDNs evolved with coupling strengths that maximize the information flow capacity are the ones that have the smallest spectral graph distance from the BDNs of the C . elegans and humans , and that , during the growing process , their MIR increase is related to moderately low amounts of global neural synchronization . Actually , the global neural synchronization levels decrease during brain network evolution , revealing an underlying global no Hebbian-like evolution process ( where synapse strength leads to global decay of synchronization ) driven by a mix of local no Hebbian-like learning rules for neurons in some clusters and by Hebbian-like learning rules in neurons belonging to other clusters where synapse strength leads to cluster synchronization . In this context , the no Hebbian-like mechanism is effectively similar to the unlearning anti-Hebbian mechanism as both lead to more efficient networks , in the sense that in our case the evolved networks are able to maximize their information flow capacity . We note that if other models than Hindmarsh-Rose will be used , such as the Morris-Lecar , Izhikevich , or spiking map-based neural models , then the parameter regions that maximize information flow capacity and minimize synchronization ( or vice versa ) will be different , however we expect that this would not change the main results and conclusions of this work in the sense that brain network evolution based on the maximization of information flow capacity will lead to similar topologies , behaviors and relations for the evolved networks . For the human subjects , the graphs represent functionally connected brain regions . The Wilson-Cowan model could be appropriate to model the human brain , whereas it will not be suitable to model the C . elegans brain . As we have done here , using neurons to represent nodes in the human connectome do not reproduce the real dynamics of the brain but gives us a mean to compare results with the C . elegans and the evolved networks . In relation to our work , the maximization of the information flow capacity for low synchronization corresponds to the critical point of the dynamics in which self critical phenomena occur when the second or larger Lyapunov exponents of the BDNs are marginally positive . Finally , our results support further the hypothesis made in Ref . [2] that maximization of the information flow capacity can serve as a principle for the development of heterogeneous structures in brain dynamical networks , such as the neocortex of mammalian brains .
|
The study of the function of the brain is of primordial importance in neuroscience . Several brain models have been studied so far that take into account higher level functions of the external stimulus and the behavioral response attributed to ensembles of neurons of cortical areas . If the brain learns by maximizing the Mutual Information between stimuli and response , or by updating the internal model of probabilities by using Bayesian techniques , or by minimizing the free-energy , it provides little insight about the dynamical mechanisms appearing in brain networks when evolved based on such principles . In our work we propose a working hypothesis supported by numerical simulations that brain dynamical networks evolve based on the principle of the maximization of their internal information flow capacity , i . e . the upper bound for the information transferred per time unit between any two nodes . We make a strong case to verify our hypothesis by showing that the neural networks with the closest spectral graph distance to the brain networks of Caenorhabditis elegans and humans are the Hindmarsh-Rose neural networks evolved with coupling strengths that maximize information flow capacity . We also find that synchronous behavior and capacity of information flow of the evolved neural networks reproduce well the same behaviors observed in the brain dynamical networks of Caenorhabditis elegans and humans . Finally , we find that global neural synchronization levels decrease during brain network evolution .
|
[
"Abstract",
"Introduction",
"Results",
"Materials",
"and",
"Methods",
"Discussion"
] |
[] |
2015
|
Do Brain Networks Evolve by Maximizing Their Information Flow Capacity?
|
Human mobility is a key component of large-scale spatial-transmission models of infectious diseases . Correctly modeling and quantifying human mobility is critical for improving epidemic control , but may be hindered by data incompleteness or unavailability . Here we explore the opportunity of using proxies for individual mobility to describe commuting flows and predict the diffusion of an influenza-like-illness epidemic . We consider three European countries and the corresponding commuting networks at different resolution scales , obtained from ( i ) official census surveys , ( ii ) proxy mobility data extracted from mobile phone call records , and ( iii ) the radiation model calibrated with census data . Metapopulation models defined on these countries and integrating the different mobility layers are compared in terms of epidemic observables . We show that commuting networks from mobile phone data capture the empirical commuting patterns well , accounting for more than 87% of the total fluxes . The distributions of commuting fluxes per link from mobile phones and census sources are similar and highly correlated , however a systematic overestimation of commuting traffic in the mobile phone data is observed . This leads to epidemics that spread faster than on census commuting networks , once the mobile phone commuting network is considered in the epidemic model , however preserving to a high degree the order of infection of newly affected locations . Proxies' calibration affects the arrival times' agreement across different models , and the observed topological and traffic discrepancies among mobility sources alter the resulting epidemic invasion patterns . Results also suggest that proxies perform differently in approximating commuting patterns for disease spread at different resolution scales , with the radiation model showing higher accuracy than mobile phone data when the seed is central in the network , the opposite being observed for peripheral locations . Proxies should therefore be chosen in light of the desired accuracy for the epidemic situation under study .
One of the biggest challenges that modelers have to face when aiming to understand and reproduce the spatial spread of an infectious disease epidemic is to accurately capture population movements between different locations or regions . In developed countries this task is generally facilitated by the existence of data or statistics at the national or regional level tracking individuals' movements and travels , by purpose , mode , and other indicators if available ( see e . g . transport statistics in Europe [1] , commuting , migration data or other types of mobility at country level [2]–[6] ) . Access to highly detailed and updated data may however still be hindered by national privacy regulations , commercial limitations , or publication delays . The situation becomes increasingly complicated in less-developed regions of the world , where routine data collection may not be envisioned at similar levels of details [7] , but which , most importantly , may be characterized by a high risk of emergence and importation of infectious disease epidemics or may suffer of endemic diseases . Depending on the infectious disease under study , different mobility processes may play a relevant role in the spatial propagation of the epidemic while others appear to be negligible , as determined by the typical timescales and mode of transmission of the disease , and the geographic scale of interest . For rapid directly transmitted infections , daily movements of individuals represent the main mean of spatial transmission . At the worldwide scale , air travel appears to be the most relevant factor for dissemination , as observed during the SARS epidemic [8] , [9] and the 2009 H1N1 pandemic [10] , [11] . On smaller regional scales , instead , daily commuting is significantly linked to the spread of seasonal influenza [12] , [13] , affecting the epidemic behavior at the periphery of the airline transportation infrastructure [14] . To overcome issues in accessing commuting data when simulating spatial influenza spread , epidemic models have traditionally relied on mobility models to synthetically build patterns of movements at the desired scale [14]–[16] . The gravity model [17] and the recently proposed radiation model [18] have been shown to fit well the commuting patterns observed in reality on different spatial scales [12] , [14]–[16] , [18]–[20] . Next to mobility modeling approaches , alternative tools for understanding daily human movements have more recently flourished thanks to the availability of individual data obtained from different sources , namely mobile phone call records carrying temporal and spatial information on the position of the cell phone user at the level of tower signal cells [21]–[23] . Such direction of research has gained great popularity , leading to the discovery of universal characteristics of individual mobility patterns , and the possibility to study mobility in space and at timescales that were unreachable before [21]–[26] . Such increasing volumes of finely resolved human mobility data , thanks to the near ubiquity of mobile phones , also offered an opportunity to contrast the huge deficit of quantitative data on individual mobility from underdeveloped regions . They were indeed used to shed light on malaria diffusion and identify hotspot areas [24] , [26] , [27] , to monitor human displacements in case of natural disasters [25] , [28] and to study disease containment strategies in Ivory Coast [29] . Despite the variety of modeling approaches and data sources , the impact of using different proxies for human commuting in epidemic models for rapidly disseminated infections is still poorly understood . Each approach or source of data clearly has its own intrinsic strengths and weaknesses , related to accuracy and availability of the dataset . More specifically , mobility models require some assumptions or input data for calibration and fit to the real commuting behavior . The gravity model requires full knowledge of mobility data for its parameter fitting and can be extended to other regions where data is not available in case of empirical evidence pointing to “universal” commuting behavior at a given resolution scale , i . e . well described by the same set of parameter values [14] , or by making assumptions on generalizability . The radiation model requires population distribution values and the total commuter flows out of a given region , a quantity that may not be easily accessible at the desired level of resolution or with sufficient coverage . While mobile phone data can provide mobility information at a high granularity level , they are also characterized by a number of issues that may hinder their use . Phone data are inevitably affected by biases related to the population sampling: coverage is usually not homogenous across space and it depends on the market share of the operator providing the data . Phone ownership and usage may differ across social groups , gender or age classes depending on the country under study [30] , [31] , and access to users' metadata to evaluate the representativeness of the sample is limited by privacy concerns [32] . Given the recent availability of these data , the impact of such biases on mobility estimates is still poorly understood . Recent studies have assessed the effects of using gravity models in mathematical epidemic models [12] , [33] , however similar works on the use of data-saving options like the radiation models or of alternative strategies like mobile phone activity data for epidemic applications are still missing . The aim of this paper is therefore to assess the adequacy of two specific proxies – mobile phone data and the radiation model – to reproduce commuter movement data for the modeling of the spatial spread of influenza-like-illness ( ILI ) epidemics in a set of European countries . We first compare the commuting networks extracted from the official census surveys of three European countries ( Portugal , Spain and France ) to the corresponding proxy networks extracted from three high-resolution datasets tracking the daily movements of millions of mobile phone users in each country . More specifically , we examine through a detailed statistical analysis the ability of mobile phone data to match the empirical commuting patterns reported by census surveys at different geographic scales . We then examine whether the observed discrepancies between the datasets affect the results of epidemic simulations . To this aim , we compare the outcomes of stochastic SIR epidemics simulated on a metapopulation model for recurrent mobility that is based either on the mobile phone commuting networks or the radiation model commuting networks , with respect to the epidemics simulated by integrating the census data . We evaluate how the simulated epidemic behavior depends on the underlying mobility source and on the spatial resolution scale considered , by investigating the time to first infection in each location and the invasion epidemic paths from the seed .
The study relied on billing datasets that were previously recorded by a mobile provider as required by law and billing purposes , and not for the purposes of this project . To safeguard personal privacy , individual phone numbers were anonymized by the operator before leaving storage facilities , in agreement to national regulations on data treatment and privacy issues , and they were identified with a security ID ( hash code ) . The research was reviewed and approved by the MIT's Institutional Review Board ( IRB ) . As part of the IRB review , authors , who handled the data , and the PI participated in ethics training sessions at the outset of the study . We use a metapopulation modeling approach [40] , [41] to perform numerical simulations of epidemic scenarios . We assume the national population of every country to be spatially structured in subpopulations defined by the administrative subdivisions described in the previous subsection . We focus on rapid directly transmitted infections , such as influenza-like-illnesses , for which daily regular movements of individuals for commuting purposes were found to correlate well with the observed regional spread [12] , [13] . We consider a simple SIR compartmental model [41] , where individuals can be either susceptible ( S ) , infectious ( I ) or recovered ( R ) from the infection , assuming a life-long immunity for recovered individuals . The dynamics is discrete and stochastic and individuals are assumed to be homogeneously mixed within each subpopulation . No additional substructure of the population is considered ( e . g . schools or workplaces ) , as our aim is to introduce a rather simple epidemic model to test the adequacy of different commuting sources for the simulation of ILI dissemination within a country . We therefore neglect unnecessary details that may hinder the interpretation of results . Subpopulations are coupled by directed weighted links representing the commuting fluxes between two locations , thus defining the metapopulation structure of the model [40] , [41] . No other type of movement is considered . Human mobility is described in terms of recurrent daily movements between place of residence and workplace so that the infection dynamics can be separated into two components , each of them occurring at each location [42] . The number of newly infected individuals during the working time in location is randomly extracted from a binomial distribution considering trials ( susceptible individuals living and working in location , , and susceptible individuals living in and working in , ) and a probability equal to the force of infection being the transmissibility of the disease , and the total population and the total number of infectious individuals living in location and working in , respectively . Similarly , the infection events taking place at the resident location during the remaining part of the day are randomly extracted from a binomial distribution considering susceptible individuals and probability equal to the force of infection . We model an influenza-like-illness transmission characterized by an exponentially distributed infectious period with average days [43] , [44] , and explore three epidemic scenarios by varying the transmissibility β and corresponding to the following values of the basic reproductive number ( average number of secondary cases per primary case in a fully susceptible population [41] ) : , , , representing a mild , moderate , and severe epidemic , respectively . Simulations are fully stochastic , individuals are considered as integer units and each process is modeled through binomial and multinomial extractions ( more details on the simulation algorithm are reported in Section 4 in Text S1 ) . Each day of the simulation is modeled with commuting movements informed by the three sources considered for a typical working day; therefore no weekends or holidays are envisioned in the model . Simulations are initialized with 10 individuals localized in a given seed . As seeds we consider the country's capital ( Lisbon , Madrid and Paris ) , a peripheral location with a small population ( Barrancos , Lleida and Barcelonnette ) , and a medium size location , characterized by an average population and an average number of connections through commuting links ( Braga , Jaen and Rennes ) . Although the countries under study are geographically contiguous , they are considered as independent entities since the investigated datasets do not include refined data about cross-border commuters . A sensitivity analysis on the role of cross-border commuting in the spread of ILI is reported in Section 3 in Text S1 . Once a set of initial conditions is defined ( mobility network , , and seeding location ) , we simulate 1 , 000 stochastic realizations for each epidemic scenario , for a total duration of 8 months . Such timeframe is chosen as a reference estimate of the expected time comprising the interval from the initial seeding of a pandemic event to the international alert ( approximately two months in the case of the 2009 H1N1 pandemic [45] ) and the average time period needed to develop a vaccine against the circulating virus ( approximately six months ) [46] . During this timeframe the value of the basic reproductive number is kept constant , and no change in behavior that could be self-initiated in response to the epidemic [47] , [48] , or imposed by public health interventions is considered , for the sake of clarity in the comparison of the results .
The census commuting networks for Portugal include ( i ) 1 , 643 , 938 commuters traveling between the 278 municipalities through 25 , 634 weighted directed connections , and ( ii ) 469 , 089 commuters traveling between the 18 districts on a fully connected network . In Spain we consider the provinces' geographical scale only , as constrained by the information available in the census survey . The commuting network is formed by 47 nodes and 722 weighted directed edges , representing the daily travel flows of 537 , 331 commuters . The commuting networks for France are defined at the district scale ( 8 , 019 , 636 commuters moving along 38 , 077 weighted directed edges connecting 329 nodes ) , and at the department level ( 4 , 957 , 193 commuters for 7 , 994 weighted directed links among 96 nodes ) . For all countries , at all scales considered , all administrative units are included in the datasets ( i . e . they have at least one incoming or outgoing commuting flux to another administrative unit in the country ) . A summary of the basic statistics of the networks extracted from census data is reported in Table 1 . Commuting patterns from mobile phone records are extracted from a sample of 1 , 058 , 197 anonymous users in Portugal , 1 , 034 , 430 in Spain , and 5 , 695 , 974 in France . Records referred to 2 , 068 towers in Portugal , 9 , 788 towers in Spain , and 18 , 461 in France . Once mapped onto the administrative units , we find 452 , 113 , 460 , 211 and 1 , 676 , 103 total commuters in the mobile data samples in Portugal , Spain , and France , respectively , corresponding to the lowest administrative hierarchy . Population tracked by the operators' samples is distributed nationwide and approximately equal to 9% of the census population in Portugal and France , and 2% of the census population in Spain . By taking into account these scaling factors , cell phone population correlates well with the census population at the highest geographical resolution considered , with a Pearson correlation coefficient between the two quantities equal to for Spanish provinces , Portuguese municipalities and French districts . Population coverage is rather uniform in France with more than half of the districts in the interval of the national coverage value ( grey colored units in Figure 1 ) , while larger discrepancies are observed in the geographic distribution of the tracked population in Spain and Portugal . In Spain we observe a significant undersampling of the population in Galicia and Basque regions . In Portugal , we observe larger regional fluctuations around the national coverage value: most of the municipalities report an undersampled population , whereas the region close to the capital , Lisbon , shows an oversampling as large as 3 times the national coverage . Commuting networks obtained from census data and mobile phone activity data share the same number of nodes at all hierarchies considered in all countries , given that all administrative units were covered by both datasets , however variations are observed in the number of commuting links ( Table 1 ) . The set of links common in both datasets in the Portugal case at the municipality level account for about 60% of the total links of each network and include more than 96% of the total travel flux of both networks . Aggregating the datasets at the level of Portuguese districts , both networks become very close to fully connected , almost achieving a perfect overlap ( more than 99% of links falling in the intersection ) . Similar figures are obtained for French districts , though the common 95% of traffic is distributed over 82% of the census links and only 52% of the mobile phone links . Spain displays a different situation , with the census commuting network topology being completely included into the mobile phone one . Census commuting links represent only 37% of connections of the mobile phone dataset , however accounting for 87% of its total traffic . We compare the probability density distributions of the travel fluxes in both networks ( Figure 2 ) , after considering the basic normalization scaling to the population of each administrative unit ( see Methods ) . All distributions display a broad tail and very similar shapes in each country , and differences are observed in particular for small traffic values . In Portugal and France , the very weak commuting flows are not captured by the mobile phone dataset , clearly as an outcome of the smaller users' sample size in the mobile phones case with respect to census . Such discrepancy disappears when we move to larger spatial scales , as in the case of Spain . Restricting our analysis on the topological intersection , a side-by-side weight comparison on each link shows a high correlation between the two datasets ( Spearman's rank correlation coefficient >0 . 7 for the largest administrative units , Table 2 ) , however commuting fluxes in the mobile phone network are found to be larger than the census ones across almost the entire interval of values ( panels d-f of Figure 2 ) . Deviations appear larger for smaller fluxes ( commuters ) in Portugal and France , with a good agreement for the largest values , whereas they are uniform in the case of Spain . Similar results are obtained when we analyze the total number of commuters leaving a given administrative unit , as well as the total number of incoming commuters in a given unit . A strong correlation between the two datasets is found for both quantities , generally independent of the level of aggregation considered ( Spearman's coefficient >0 . 88 for Portugal and France ) , whereas small values of the Lin's coefficient indicate the presence of strong differences in the absolute values for the two datasets ( <0 . 53 across all countries and for all administrative levels , for both quantities , Table 2 ) . Spain has a rather low Spearman's coefficient for the incoming fluxes of commuters with respect to the other countries ( 0 . 54 vs . values >0 . 88 ) , showing a poor capacity of the mobile phone data to properly account for the attraction of commuters of a given location . The correlations found along the various indicators do not ensure the statistical equivalence of the two datasets ( a Wilcoxon-test for matched pairs would reject the null hypothesis of zero median differences between paired values of the same quantities ) . We further analyze whether the observed discrepancies between the weights in the mobile phone networks and the census networks show any dependency on the variables that characterize the underlying spatial and social structure , namely the Euclidean distance between the connected nodes ( calculated from the coordinates of the administrative unit's centroid ) , the population of the origin node and the population of the destination node ( Figure 3 ) . The overestimation of the magnitude of commuting fluxes in the mobile phone dataset does not show a significant dependence on the population sizes . Fluxes are instead found to be more similar when they connect units at shorter distances with respect to longer distances across the countries . Such variation disappears if we consider the topological distance defined by a neighbor joining approach ( see Section 3 . 4 in Text S1 ) . Spatial aggregation into larger administrative units does not alter this overall picture but weakens the effect observed on distance ( see details in Section 2 . 1 in Text S1 ) . If we refine the normalization of the mobile phone networks by taking into account the total number of commuters in each administrative unit , the agreement with the census dataset improves in the side-by-side weight comparison on every link ( see Section 3 in Text S1 ) . This approach allows us to explicitly discount the systematic overestimation found with the basic normalization , resulting in higher Lin concordance coefficients ( Table S1 in Text S1 ) ; discrepancies between mobile phone and census data are however still observed for very small and very large commuting flows . We examine whether the observed non-negligible discrepancies in the commuting fluxes of the two datasets are also significant from an epidemic modeling perspective , altering substantially the outcome of disease spreading scenarios . We compare scenarios obtained from stochastic metapopulation models equally defined and initialized , except for the mobility data they integrate ( see Methods ) . In addition to the census commuting network and the mobile phone commuting network , we also consider the synthetic commuting network generated with the radiation model . Epidemics starting from different seeds in the three countries , and characterized by different values of the basic reproductive number , yield large variations of the Jaccard index value measuring the similarity in the epidemic invasion paths produced by the use of mobile phone data and of the radiation model with respect to the census benchmark ( in , see Figure 4 ) . Epidemic invasion trees obtained from proxies for mobility are more similar to the ones obtained from the model integrating census data when the seed is located in the capital city of the country . In addition , increases with larger values of . If the seed is instead located in a peripheral node , values of the Jaccard similarity index fall always below 0 . 4 in the three countries , and decrease with larger values of the transmissibility . Mobile phone data performs similarly to the radiation model once the corresponding epidemic models are seeded in a central location , except for the case of Lisbon , and performs better or similar when they are seeded in a peripheral location . If the epidemic starts from a mid-size populated region , the relative performance of the radiation model against mobile phone data in the epidemic outcomes depends on , with improvements observed as increases . To test for the role of overestimation of flows , we also performed the same analysis by considering the refined normalization of the mobile phone commuting data that keeps the same total number of commuters per administrative region as in the census dataset and explicitly discounts overestimation biases . The refined normalization allows the mobile phone data to better reproduce the invasion paths obtained from census commuting flows for central and medium-type locations for all , and to perform slightly worse in case the seed is located in a peripheral location ( Figure 5 for the case of France ) . When focusing on the time of arrival in a given location , we find a systematic difference between models based on proxy networks and the benchmark model integrating census data . Mobile phone data , overestimating the census commuting fluxes if a basic normalization is considered , leads to a positive difference corresponding to a faster spreading ( Figure 4 ) . On the other hand , epidemics on the radiation model tend to unfold slower than simulations on the census network , with later arrival times as indicated by negative values of ( except in the case of France where the median of is approximately equal to zero in all cases ) . For small values of , the arrival times of simulations running on a proxy network may be substantially different from the ones obtained with census data , with of the order of months . While the transmission potential of the disease drives the magnitude of the impact of the discrepancies , the role of the seed location appears to be less relevant here than what previously observed in the study of the invasion paths . A slightly decreasing trend in the positive median values of is observed in the mobile phone vs . census results , going from peripheral to medium to central location , the effect being more pronounced in Spain and in France . By discounting a posteriori the average anticipation of the model built on mobile phone data , which is trivially due to the overestimation of the census commuting fluxes , we find a very good correlation between arrival times for the models built on the census network and on the mobile phone network , with most of the points lying close to the identity line ( Lin concordance correlation coefficient ranging from 0 . 77 to 0 . 88 , panels c , f and i of Figure 4 ) . If we consider the refined normalization , anticipation effects produced with the mobile phone data are preserved but reduced in magnitude ( Figure 5 ) . Epidemic peak times are also affected by the different distributions of commuting flows in the two networks ( see Section 2 . 2 in Text S1 ) . As soon as the disease reaches most of the nodes , the epidemic model integrating the mobile phone network displays a more homogeneous behavior , with epidemic peaks that follow very shortly after each other in all the subpopulations , while peak times in the census networks span a wider time frame . On coarser spatial scales ( Portuguese districts , French departments ) , we obtain a higher similarity between simulated results with proxies vs . census ( see Section 2 . 1 in Text S1 ) , closer to the results observed for Spanish provinces . The performance of the epidemic model built on the radiation is noticeably poorer than the mobile phone network if we consider the coarse-grained scale , for all seeds but the capital . The differences between arrival times are generally reduced by the coarse-graining , but remain significant when the reproduction number is small ( ranging between 0 and 120 days ) .
Next to traditional census sources or transportation statistics , several novel approaches to quantifying human movements have become recently available that increase our understanding of mobility patterns [21]–[28] , [49]–[52] . Adequately capturing human movements is particularly important for improving our ability to simulate the spatiotemporal spread of an emerging disease and enabling advancements in our predictive capacity [53] , [54] . Previous work has focused on testing mobility models' performance in reproducing the movements of individuals [18] , [19] , and its impact on epidemic simulation modeling results when fully supported by data [19] . The full knowledge of mobility data from national statistics is however largely limited to few regions of the world [14] , whereas in many others it may not be routinely collected nor accessible . If mobility models often require aggregated input data from national statistics on movement habits [18] or the full mobility census database [19] for the fitting procedure , mobile phone data may be thought as an ideal alternative candidate for a proxy of human movements in absence of ( complete and/or high-resolution ) mobility data from official sources [24] , [26] , [27] . To systematically test this hypothesis exploiting the full resolution of both the proxy data and the official census data for commuting , we have compared these two datasets in three European countries and performed a rigorous assessment of the adequacy of proxy commuting patterns – extracted from mobile phone data or synthetically modeled – to reproduce the spatiotemporal spread of an emerging ILI infection . Mobility data from mobile phones is able to capture well the fluxes of the commuting patterns of the countries under study , reproducing the large fluctuations in the travel flows observed in the census networks . In all countries the intersection between the two networks includes the vast majority of the commuting flows and the correlation measured on links' traffic and nodes' total fluxes of incoming or outgoing commuters is high ( though not statistically equivalent ) . This suggests that mobile phone data can be used as a surrogate tracking the commuting patterns of a given country , identifying the relative importance of its mobility connections in terms of flows' magnitude , with a resolution that is equivalent to the one adopted by official census surveys or higher . This is a particularly relevant result for data-poor situations , where census data may not be available and official statistics may not be enough to correctly inform a mobility model . Discrepancies are however found , especially in the overestimation of commuting flows per link and in the larger variations observed for weaker flows and longer distances , that appear to be responsible for the differences observed in the simulated epidemics . Epidemics run on mobile phone commuting networks reproduce well the invasion pattern simulated on the census commuting when the seed is located in a central location and is large . The capital city is indeed strongly connected to the rest of the country; therefore it behaves as a potential seeder of the direct transmission to the majority of the other cities , leading to very similar star-shaped infection trees from the seed . These rather similar sets of infected locations at the first generation of the invasion path provide a twofold contribution to the increase of : on the one side , they correspond to a large fraction of the total number of infected subpopulations , so they contribute a large relative weight in the computation of ; on the other , common infected locations are likely to maintain the similarity of the invasion paths at the second generation too , repeating the process in an avalanche fashion . Such behavior becomes increasingly stronger as grows larger . The opposite situation is instead found when seeds are located in peripheral nodes , reporting low values of the Jaccard index . The analysis of the commuting networks has indeed shown that larger discrepancies exist for small weights . Once considered in the framework of an epidemic propagation , such discrepancies are expected to lead to strong differences in the invasion already at the first generation of infected locations . If these locations directly infected by the seed strongly differ , their contribution to the decrease of the similarity of the invasion paths will become increasingly stronger for further generations: different nodes are infected and likely different neighbors of those nodes will be affected by the disease , so that deviations cumulate at each successive step of the invasion ( Figure 6 ) . Diseases with a higher transmission potential would enhance this behavior , as with a large value of the peripheral seed can more quickly infect a large fraction of the system in the mobile phone network , than in the census dataset . Such effect is also present in the radiation model that is not able to describe the epidemic behavior better than the mobile phone data when the seeding location is characterized by a small population or degree . Not being able to capture well the mobility coupling between peripheral regions and the rest of the country , the radiation model misses most of the seeding events on long distances even when is large ( Figure 6 ) . Using a synthetic proxy is therefore not always preferable to data alternatives , and mobile phones appear to be more reliable in matching the spatial epidemic spread starting from peripheral locations . A clear bias , which is observed consistently across all countries and for all resolution scales considered , is the faster rate of spread of the simulation based on the mobile phone commuting network with respect to the census one . This is clearly induced by the larger commuting flows obtained following the extraction of commuting patterns from mobile phone data using a basic normalization . The effect is stronger for as it is enhanced by the intrinsic large fluctuations characterizing epidemics close to the threshold . In such scenarios , even relatively small differences between networks' topologies can strongly alter the invasion path of the disease , consistently with the results of previous work on the effect of network sampling on simulated outbreaks [53] . Increasing the value of the reproduction number leads to narrower ranges , because the larger disease transmissibility accelerates the spreading , synchronizing the epidemic behavior at distant locations and , in general , reducing the system's heterogeneity . Time of arrival of the infection in a given location is better matched by the epidemic model built on the radiation model , though with large fluctuations for small values of . However , it is important to keep in mind that the total number of commuters per administrative unit is an input of the radiation model and no overestimation effects , as the ones resulting from the use of mobile phone data in the basic normalization approach , are possible in the model . If we inform the extraction of commuting patterns from mobile phone data with the same input data of the radiation model , i . e . through the refined normalization , predictions on the time of arrival consistently improve with respect to the basic normalization approach . Fixing the total number of commuters equally in the two datasets is however not enough to obtain an equivalent picture in terms of arrival times , as a considerable anticipation for small values of the transmissibility is still observed . These results need to be taken into account when considering epidemic simulations integrating mobility proxies , as a high accuracy in predicting arrival times can be used for assessing the epidemic situation at the source of the infection , estimating important epidemiological parameters during the early phase of the outbreak in a backtracking fashion [11] , [45] , [54] . Nodes ranking according to time to first infection also improve in the epidemic simulations based on the refined normalization with respect to the baseline one . The similarity in the invasion paths equals ( or even improves ) the levels reached once the radiation model is considered . Similar results are therefore obtained from two different sources however employing the same type and amount of input data ( for calibration/normalization ) . Jaccard index values display anyway the presence of important differences in the way the epidemic propagates on proxies with respect to census , being only when the outbreak is seeded in Paris . Effects of flows overestimation are visible in the analysis of the epidemic peaks too , but less prominent . The larger number of commuters that travel in the mobile phone networks tends to synchronize the epidemic peak between different subpopulations , leading to shorter overall timespan for all subpopulations to peak in the mobile phone case with respect to census . Differences between the datasets mostly range in a time interval of 2–3 weeks , a time resolution that still allows a meaningful comparison of epidemic results with the average reporting period of standard surveillance systems . In the case of France and Portugal we have also studied multiple hierarchical levels of the administrative units , by aggregating both datasets . Overall , our analysis indicates that the epidemic behavior on aggregated proxy network better matches the results obtained on census data , with respect to higher resolution level . This is however obtained at the cost of studying the epidemic on a lower geographic resolution , which would then provide less information on the predicted time course of the epidemic and may compromise our ability to use models to extract valuable public health information for epidemic control [54] . On the other hand , the radiation model displays an opposite behavior when aggregating on space . This suggests that at each scale of resolution there exists an optimal proxy for the description of the spatial spread of an infectious disease epidemic , similar to what observed in a comparison of mobility models [55] . The overall picture we presented clearly shows that proxies integrated into epidemic models can provide fairly good estimation of the ranking of subpopulations in terms of time to first infection . A good agreement in the simulated arrival times is intrinsically related to proxies' calibration and normalization aspects , and observed biases can be reduced by using additional information , such as the knowledge of the total number of commuters in each location . On the other hand , the most probable path of infection from one subpopulation to another appears to be affected by more substantial discrepancies between the different sources of data or synthetic flows that cannot be overcome through a simple normalization . To further improve predictions on the path of invasion , we would need to comprehensively understand the causes behind the differences observed in the data analysis . These are inevitably related to the methods used to account for the population sample considered in the mobile phone data and to define the commuting mobility per user . First , in extracting the commuting behavior of each user from mobile phone data we necessarily have to make assumptions on the identification of home and work locations ( in absence of metadata on the user ) . If we identify these two locations as the two most visited ones [22] , by definition , we are assuming that place of residence and place of work are two distinct locations , yielding that every mobile phone user is a commuter at the resolution level of the cell phone towers . Once aggregated at a larger scale ( i . e . the various administrative units under consideration ) , we obtain a population made of individuals living and working in the same unit ( non-commuters ) and of individuals commuting between two different units . While aggregation leads to a certain fraction of non-commuters , the resulting commuting behavior – expressed by the ratio of commuters vs . non-commuters – is anyway more pronounced likely because of the intrinsic assumption made on the original identification of home/work locations from the data . Different choices can be made that can improve the correct identification of home/work locations , leveraging on the availability of additional data . If timing of the call activity is provided , one possible refined definition would be to identify as home location the tower cell with the largest activity during nighttime , and the work location as the one with the largest activity constrained to daytime ( with variations of the definition of these intervals ) [32] , [33] , [38] , [56] . We tested this approach in the Portuguese dataset and found that the identification of the two locations was not substantially altered by the time-constrained definition chosen , and did not affect our results ( see Text S1 ) . Increasingly sophisticated approaches can also be envisioned , based on clustering methods applied to calling behavior [57] , [58] . In addition to the need for access to the metadata associated to the activity data , results from time-constrained or clustering methods may anyway be affected by biases induced by users' call plans ( influencing the pattern of calls to given timeframes during the 24 h or depending on the day of the week , e . g . weekday vs . weekend ) , job types ( altering the expected timing pattern of call activity from work ) , and more generally the definition of normal business hours that may have a strong cultural component . Second , our basic normalization may be too simplistic , thus inducing strong overestimation because the population sampled through the mobile phone data is not representative of the general population , being characterized by specific different features affecting the resulting mobility behavior . Biases may be induced by mobile phones ownership , with fluctuations strongly dependent on socio-economic status [31] , [58] , and by market share of the specific operator providing the data , In Spain , for example , the strong undersampling of the population in Galicia and Basque region , characterized by a strong political and cultural identity , may be due to the presence of local operators that account for a larger market share than what observed at national level . In Portugal , we observe larger fluctuations per region around the average national coverage than in other countries . Predictions for the invasion path obtained with mobile phones for epidemics starting in the capital of the country are not in good agreement with those obtained with census flows ( , Figure 4 ) . In this case , the central role of the capital , responsible for leading to higher similarity as discussed before , is reduced by the presence of larger ( and overestimated ) flows connecting less central regions in the mobile phone dataset . This leads to the creation of leaves stemming from peripheral nodes and infecting the closest neighbors , thus strongly reducing the role of the seed in infecting the large majority of nodes at the first generation of invasion . This phenomenon is effectively similar to the one encountered when the epidemic is seeded in a peripheral location . Small-scale studies targeting specific populations ( such as e . g . a city or a college town ) with additional metadata accompanying the activity records may possibly shed more light in the identification of such biases . In poorer countries these effects are expected to be of a larger magnitude , given that mobile phone users still represent a privileged minority of the population [31] . Recent work has however showed that mobility estimates in Africa are very robust to biases in phone ownership [59] . The introduction of a refined normalization to account for the non-representative nature of the mobile phone sample fixes the total number of commuters equally in the two datasets and leads to an improvement of the comparison of the commuting fluxes on a link-by-link basis . Discrepancies on traffic flows along links are however still observed that are responsible for differences in the resulting epidemic observables , even though the overall systematic overestimation obtained with the basic normalization has been discounted . Increasingly sophisticated approaches can be developed that use iterative proportional fitting , fixing two marginal values that need to be assumed , i . e . the total numbers of incoming commuters and of outgoing commuters per location ( or additional data , such as points of interest in the case of intra-city commuting ) [60] . Knowledge of these quantities may however not be largely accessible across different regions of the world . Third , there may be inconsistencies in the definition of commuting for both datasets , or differences in the year of collection of each dataset . We have no information on users' age in the mobile phone dataset , therefore movements for work or study are both tracked in users' trajectories . Commuting for study reasons is included in the Portuguese and French census data , whereas Spain reports about workflows only . The impact of not considering students' commuting in the Spanish case is however estimated to be rather low . Spanish data is indeed collected at a high administrative level ( provinces ) , where students' commuting flows may be very weak given that they are usually more localized than those of workers . Data from France shows that 95% of students ( aged<15 ) travel on distances less than 10 km [4] . To estimate the impact of missing students in the Spanish dataset on the province scale , we examined the fraction of commuter movements of students in the French census commuting network aggregated at the level of regions , i . e . similar to the size of Spanish provinces . In France , students represent about 10% of the total commuting flows across regions . If we assume a similar statistics for Spain too , such ratio is not sufficient to explain the discrepancy observed between the normalized mobile phone commuting flows and the census commuting flows in Spain ( Table 1 ) . In addition , the lack of a portion of individuals in the dataset ( e . g . students ) would have no impact when using the refined normalization because , in that case , the total number of commuters is set to be equal in both data sets by definition . Discrepancies in the year of data collection for the two sources range from two years for Spain ( 2005 is the year of collection of census data , 2007 the year of collection of phone data ) to five years for Portugal ( 2001 , 2006 ) . In the case of France , the two datasets belong to the same year ( 2007 ) . To assess the possible changes in commuting flows with time , we analyzed French yearly data between 2006 and 2009 , given their availability ( see Table S1 in Text S1 ) . In three years , the total number of commuters in the country increased by about 3% , and every year , the total number of commuters increased by 1% or less . Assuming that similar trends apply to the other countries , we conclude that the total census commuting flows of Spain and Portugal may have increased by 2% and 5% , respectively , a difference being much smaller than the average discrepancy observed between census data and mobile phone data . Finally , the epidemic model considered adopts some approximations that we would like to discuss in the following . Even if countries under consideration belong to a contiguous area in continental Europe , numerical simulations for the epidemic spread were performed for each country in isolation . This choice is driven by the lack of mobile phone data for cross-border movements ( given their national nature ) , and by the negligible fraction of commuting across countries with respect to national commuting ( about 780 , 000 people in the EU , including EEA/EFTA , were cross-border commuters in the year 2006/2007 [61] over a total of more than 100 million national commuters ) . For the sake of completeness , we also checked our results against the inclusion of cross-border commuting in the census network of France , where international movements are predominant in a subset of districts ( for instance , in those bordering Switzerland ) . Results reported in Text S1 show that including cross-border movements in the census commuting networks does not significantly alter the simulated epidemic patterns , keeping our conclusions unchanged . The modeling approach we proposed was fairly simple and did not consider additional substructure of the population , interventions , change of behavior or weekend vs . weekday movements . Our aim not being to reproduce historical epidemics , we chose to include only the basic ingredients that were the object of the analysis in order to achieve a clearer understanding and interpretation of results . Simulations were performed assuming a continuous series of working days , given the purpose of the study and the knowledge that the inclusion of weekend movements has little or no effect in the resulting epidemic profile [42] . To apply this framework to real case studies , more refined compartmental models , movements and interactions between individuals may need to be considered . Our study was performed on three European countries , and we expect that our conclusions are applicable to other developed countries in the world characterized by similar cultural , social , and economic profiles . Our approach for the extraction of commuting patterns from mobile phone data was based on minimal assumptions in order to facilitate its generalizability in other settings where data knowledge may be limited or completely absent . Further work is necessary to extend this work to the analysis of the adequacy of mobile phone data as proxy for human mobility in underdeveloped countries where cultural and socio-economic factors may affect differently the biases here exposed . We also note that diseases other than ILI may be of higher interest for these regions , and in that case the relevant mobility mode and epidemic model would need to be updated in the approach we presented . For instance , the transmission of the disease under study may be strongly affected by seasonal forces , such as the variations in human density and contact rates due to agricultural cycles that drive the spatial spread of measles in Niger [62] , or other factors , such as poor sanitation standards that were linked to the persistence of the poliovirus in India [63] . Also , long-term migration may play a more significant role than commuting movements in the spread of the polio virus . On the other hand , the concern for the emergence of new infectious diseases with pandemic potential , as in the recent cases of the H7N9 flu in China [64] and the MERS-CoV virus in the Middle East [65] , [66] , is significant for developing countries as well , given they may have access to fewer resources for preparedness and control . In this context , our work can provide useful insights for the development of epidemic models for the spatial spread of such rapidly disseminated directly transmitted emerging diseases .
|
The spatial dissemination of a directly transmitted infectious disease in a population is driven by population movements from one region to another allowing mixing and importation . Public health policy and planning may thus be more accurate if reliable descriptions of population movements can be considered in the epidemic evaluations . Next to census data , generally available in developed countries , alternative solutions can be found to describe population movements where official data is missing . These include mobility models , such as the radiation model , and the analysis of mobile phone activity records providing individual geo-temporal information . Here we explore to what extent mobility proxies , such as mobile phone data or mobility models , can effectively be used in epidemic models for influenza-like-illnesses and how they compare to official census data . By focusing on three European countries , we find that phone data matches the commuting patterns reported by census well but tends to overestimate the number of commuters , leading to a faster diffusion of simulated epidemics . The order of infection of newly infected locations is however well preserved , whereas the pattern of epidemic invasion is captured with higher accuracy by the radiation model for centrally seeded epidemics and by phone proxy for peripherally seeded epidemics .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"Methods",
"Results",
"Discussion"
] |
[
"biology",
"and",
"life",
"sciences",
"infectious",
"disease",
"modeling",
"population",
"modeling",
"computational",
"biology"
] |
2014
|
On the Use of Human Mobility Proxies for Modeling Epidemics
|
Microbial symbionts can modulate host interactions with biotic and abiotic factors . Such interactions may affect the evolutionary trajectories of both host and symbiont . Wolbachia protects Drosophila melanogaster against several viral infections and the strength of the protection varies between variants of this endosymbiont . Since Wolbachia is maternally transmitted , its fitness depends on the fitness of its host . Therefore , Wolbachia populations may be under selection when Drosophila is subjected to viral infection . Here we show that in D . melanogaster populations selected for increased survival upon infection with Drosophila C virus there is a strong selection coefficient for specific Wolbachia variants , leading to their fixation . Flies carrying these selected Wolbachia variants have higher survival and fertility upon viral infection when compared to flies with the other variants . These findings demonstrate how the interaction of a host with pathogens shapes the genetic composition of symbiont populations . Furthermore , host adaptation can result from the evolution of its symbionts , with host and symbiont functioning as a single evolutionary unit .
Animals and plants live in close association with numerous symbiotic bacteria that often cause strong phenotypic changes in their hosts [1] . For example , defensive symbionts can increase host resistance to pathogens and parasitoids [2–8] . In insects , several defensive symbionts are maternally transmitted [3–7] , such that the fitness of the symbiotic bacteria is dependent on that of their female hosts . Therefore , one can expect that selection on host phenotypes , including resistance to ( other ) parasites , impacts the evolution of the bacterial symbiont population . Host parasite burden can impact symbiont populations . For example , experimental evolution of the pea aphid Acyrthosiphon pisum or of Drosophila hydei in the presence of parasitoid wasps , caused an increase in the frequency of individuals carrying the protective symbionts Hamiltonella defensa and Spiroplasma , respectively [9 , 10] . Also , the recent spread of a Spiroplasma symbiont in natural populations of D . neotestacea in North America has been associated with the arrival of a parasitic nematode to this continent [7] . In agreement with this , the frequency of Spiroplasma in a D . neotestacea population increases in the presence of the parasitic nematode during experimental evolution [11] . These studies show changes in the prevalence of endosymbiont infection in host populations , but do not address selection at the level of the genetic diversity of the symbiont itself . However , some evidence suggests that this could be the case: 1 ) some defensive symbiont populations display genetic and phenotypic variability [12–18] and 2 ) variants or strains of endosymbionts change in frequency in natural populations or during experimental evolution [19–21] . Nonetheless , a clear link between the selective pressure exerted on hosts and the genetic changes observed in the symbionts has been missing . In this study , we establish a relation between host adaptation to parasites and changes in the genetic composition of endosymbiont populations . Wolbachia is a maternally-transmitted bacterial endosymbiont widespread in arthropods [22] . In some natural hosts it induces strong protection against infection with several RNA viruses [3 , 4 , 23 , 24] . Importantly , genetic variation in the Wolbachia strain of Drosophila melanogaster ( wMel ) , can be linked to the strength of antiviral protection [14 , 15] . Using experimental evolution , we have previously shown that D . melanogaster populations adapt to Drosophila C virus ( DCV ) challenge [25] . Resistance to this pathogen increases over twenty generations and we identified the genetic bases of this adaptation at the host level [25] . However , all individuals of the outbred founder population carried Wolbachia [26] . Therefore , we used this unique setup to ask if the genetic composition of the Wolbachia wMel populations also changed during host adaptation to DCV challenge and whether this change could impact on Drosophila fitness .
We performed experimental evolution on four replicate populations of D . melanogaster under selection with systemic DCV infection ( Virus-Selected ) and four replicates with mock infection ( Control ) [26] . DCV infection was performed at every generation using the same virus strain , at the same dose . As previously described [25] , we performed genome-wide sequencing of DNA from pools of each population ( Pool-Seq ) [27 , 28] . Using Pool-Seq on the Ancestral populations and on the Control and Virus-Selected populations after 20 generations [25] , we determined the genetic diversity of Wolbachia in these populations . We found statistically significant changes in the frequency of 125 single nucleotide polymorphisms ( SNPs ) between the Ancestral and the Virus-Selected populations ( Fig 1A , S1 and S2 Figs , S1 Dataset ) . Of these , 111 were also significantly different between Control and Virus-Selected populations , but not between Control and Ancestral populations , showing that these changes in the genetic composition of the Wolbachia populations are mostly specific to the response to viral infection . Phylogenetic analysis , based on whole genome sequencing of Wolbachia and mitochondria , indicate that in the recent past wMel has been strictly vertically transmitted [13 , 29 , 30] . Moreover , there is no evidence of fly lines simultaneously carrying wMel variants from distant haplotypes or recombination between these [29] . Therefore we inferred Wolbachia haplotypes in the Ancestral , Control and Virus-Selected populations from the Pool-Seq data ( S1 Text ) . Overall , we identified diagnostic SNPs ( i . e . SNPs present in all variants , and only in variants , of a specific clade ) for three of the major clades of wMel ( S1 Dataset ) [13 , 14 , 21] . The Ancestral Wolbachia populations consisted of approximately 88% clade V variants and 12% of variants of clades I and III . In the Virus-Selected populations all these diagnostic SNPs became fixed with the nucleotide that matches clade V ( S1 Dataset ) . In fact , in all the 123 SNPs that became fixed between Ancestral and Virus-Selected populations the fixed nucleotides match clade V . Moreover , between Ancestral and Control populations , the 8 SNPs that significantly changed have only been detected before in clades I and III variants , whereas Clade V specific SNPs did not significantly change in frequency between these populations ( S1 Text and S1 Dataset ) . Therefore , we conclude that selection of D . melanogaster with a viral challenge changed the frequencies of wMel variants in the host populations and led to fixation of clade V wMel variants . To confirm that the fixation of clade V variants was specific to the Virus-Selected populations , we analyzed individual flies from the Ancestral , Control and Virus-Selected populations as well as from a parallel selection regime in which Drosophila was challenged with systemic bacterial infection [26] ( Fig 1B ) . We determined the wMel variant carried by 96 individual flies from each replicate population through restriction analysis of a PCR fragment containing a clade V diagnostic SNP . This analysis distinguishes flies carrying wMel variants of clades I/III or clade V . The frequencies of flies carrying clade V wMel variants in Ancestral , Control and Virus-Selected populations are in agreement with the Pool-Seq data ( S1 Text ) and clade V variants are only fixed in the Virus-Selected populations . We observed significant differences in frequencies between the Virus-Selected populations and the other tested populations but not between any other regimes ( generalized linear mixed model ( GLMM ) , Selection Regime effect , χ23 = 31 . 648 , p < 0 . 001 , Tukey HSD , |z| > 3 . 437 , p < 0 . 005 for all comparisons with the Virus-Selected populations , |z| < 2 . 067 , p > 0 . 23 for all other comparisons ) . These data argue against drift being responsible for the fixation of clade V variants since Bacteria-Selected populations had fewer surviving individuals for a larger number of generations than Virus-Selected populations ( S1 Text and S2 Dataset ) and wMel variants of clade V did not reach fixation in any of the Bacteria-Selected populations . Moreover , a time-course analysis of wMel clade V frequencies in the Virus-Selected regime , based on individual genotyping , also shows that changes in frequencies were parallel in all four replicates ( Fig 1C ) . Finally , based on the frequencies of clade I/III and clade V variants in generations 0 , 5 , 10 and 20 we estimated a strong selection coefficient against clade I/III variants of 0 . 263 ( 0 . 177–0 . 349 ) ( estimated log-linear slope using GLMM , Generation effect , χ21 = 42 . 466 , p < 0 . 001 ) [31] . Therefore , fixation of clade V variants in all the Virus-Selected populations is unlikely due to drift , injury or a generic immune challenge , but the consequence of the specific adaptation to viral challenge . To analyze the phenotype of clade V wMel variants against clade III variants we established eleven different isofemale lines carrying wMel from clade V and eleven different isofemale lines carrying wMel from clade III . These lines were established from the Control populations . To directly compare the differences between wMel from the different clades we set up reciprocal crosses between eleven independent pairs of clade III and clade V isofemale lines . Since wMel is only maternally transmitted , the female progeny of each of these paired crosses differed in the wMel variant , but had the same host genotype . During the virus-selection protocol , reproduction of surviving adults took place five to seven days after DCV infection [25] . At five , six and seven days after DCV infection , flies carrying wMel clade III variants had lower survival than flies with clade V variants ( Fig 2A , GLMM , wMel clade effect , χ21 > 16 . 44 , p < 0 . 001 in all daily comparisons , see also analysis of S3A Fig , below ) . Analysis of the survival data until 20 days post-infection confirms an overall lower susceptibility upon viral infection of flies with wMel variants of clade V compared with flies carrying clade III variants ( Fig 2B , mixed effect Cox model , wMel clade effect , χ21 = 25 . 817 , p < 0 . 001 , see also analysis of S3B Fig , below ) . We also analyzed the reproductive output of flies with the different wMel variants ( from the same reciprocal crosses ) , five to seven days post-infection with DCV ( Fig 2C ) . Flies with clade III variants had fewer progeny than flies carrying clade V variants ( linear mixed model ( LMM ) , wMel clade effect , χ21 = 39 . 217 , p < 0 . 001 ) . This difference between variants is contingent on viral infection , since their reproductive output is not significantly different in the absence of infection ( LMM , χ21 = 2 . 321 , p = 0 . 128 , S4 Fig ) . The differences between flies carrying wMel clade III variants and flies carrying clade V based on reproductive output and survival at five to seven days post-infection could explain the relative fitness of 0 . 723 ( 0 . 651–0 . 823 ) calculated from the above estimated selection coefficient ( w = 1-s ) . Mitochondria are co-inherited with Wolbachia . Therefore , the phenotypic differences we observed between flies carrying different wMel variants could , hypothetically , be due to phenotypic differences of their associated mitochondria variants . If this were the case , selection could have acted on the mitochondria and indirectly affected frequencies of Wolbachia variants . To test for the contribution of mitochondria to the phenotypic differences observed , we repeated these assays with the same isofemale lines and matching isofemale lines from which wMel was removed by tetracycline treatment . We found a significant interaction between wMel/mitochondria clade ( cytotype ) and Wolbachia presence , both in survival 5 , 6 or 7 days after infection and in overall survival ( S3A and S3B Fig–clade by presence of Wolbachia , GLMM and Mixed effect Cox model , p < 0 . 001 ) . Importantly , both models showed a significant difference in survival between flies carrying wMel clade V or wMel clade III , but not between flies of the same cytotypes without Wolbachia ( pairwise comparisons between clades in Mixed effect Cox Model , |z| = 5 . 739 , p < 0 . 001 and |z| = 0 . 868 , p = 0 . 385 , for flies with and without Wolbachia , respectively and in pairwise comparisons between clades using GLMM at 5 , 6 or 7 days post-infection |z| > 3 . 794 , p < 0 . 001 and |z| < 1 . 678 , p > 0 . 093 , for flies with and without Wolbachia , respectively ) . Analysis of differential reproductive output had similar results . There was a significant interaction between cytotype and Wolbachia presence ( S3C Fig , LMM , clade by presence of Wolbachia , χ21 = 4 . 2 , p = 0 . 040 ) . Pairwise comparisons of reproductive output between cytotypes with Wolbachia showed a significant difference ( t = 4 . 27 , p < 0 . 001 ) but not between cytotypes in the absence of Wolbachia ( t = 1 . 2 , p = 0 . 087 ) . Overall , these data indicate that there is no significant difference in survival or reproductive output , upon viral infection , between flies only carrying different mitochondria . Therefore , the phenotypic differences we observe are due to differences between wMel variants and not between mitochondria variants . Finally , we tested if lower fitness upon viral infection of flies carrying wMel clade III variants was associated with higher DCV load as different wMel variants have been shown to confer differential resistance to DCV infection [14] ( S5A Fig ) . Flies carrying these variants had 5 . 4 fold higher levels of DCV compared with flies carrying clade V variants ( log-LMM , wMel variant effect , χ21 = 11 . 479 , p < 0 . 001 ) . The lower resistance to viruses of flies carrying Clade III variants , compared to Clade V , may explain their lower survival and fertility upon infection . Flies with Clade III variants also had lower Wolbachia levels when compared with flies carrying clade V variants ( S5B Fig , LMM , χ21 = 16 . 292 , p < 0 . 001 ) . This may explain lower antiviral resistance of these variants , in line with previous findings [14 , 15 , 32 , 33] .
Our data show that ( a ) the frequencies of Wolbachia variants specifically change when Drosophila populations evolve in the presence of viruses , ( b ) this exposure to DCV leads to fixation of clade V wMel variants , and ( c ) genetically identical individuals are more protected against DCV infection and display lower viral loads when they harbor these Clade V variants , relative to when they harbor other variants still present in the Control ( and Ancestral ) population . Moreover , the selection coefficient inferred from the evolutionary dynamics of clade V in DCV-exposed populations could be explained by the fitness advantage of clade V over clade III wMel variants in flies subjected to DCV infection . These results demonstrate that host infection by parasites can be a selective force leading to genetic changes in the endosymbiont population such that the most protective variants become fixed . In turn , this evolution can contribute to host adaptation to pathogens . We have previously identified two regions in the D . melanogaster genome that mediate adaptation of this population to DCV infection [25] . Here we show that this adaptation also leads to change in wMel genetic diversity . There may be interactions between selection on the genomes of the symbiont and the host , which we did not test here . We have demonstrated before that the Virus-Selected population had a higher survival upon DCV infection than the Control populations even when Wolbachia was removed from these populations [25] . This indicates that , overall , the selected alleles confer an advantage in the presence of viruses independently of the presence of Wolbachia . However , it was recently shown that the strength of selection on host genetic variation is decreased in the presence of these protective symbionts [34] . Therefore , the presence or absence of Wolbachia interacts with selection at the host level . However , this does not address interactions between selection acting both at the level of the symbiont and the host . We show differences between the wMel variants using isofemale lines established from the Control populations , and therefore not evolved under Virus challenge . This indicates that the virus susceptibility phenotypes associated with the wMel variants are not dependent of selection at the level of the host genome . Moreover , we compared the phenotypes of the progeny of several independent reciprocal crosses between lines carrying different wMel variants . This setup controls for differences in host genetic background . It will be interesting in the future to investigate how genetic variation in the host impacts on the phenotypes of Wolbachia variants , and vice-versa . Other wMel variants were shown to differ in survival upon viral infection [14] . wMel variants from clade VI confer more protection to viruses than variants from clade III or clade VIII [14] . Here clade V variants are more protective than clade III variants ( and clade I variants are also counter-selected in the Virus-Selected populations ) . These results indicate that clade V and VI are more protective against viral infections and clade I , III and VIII , less protective . It will be important in the future to make a direct comparison of the antiviral protection conferred by these different variants and understand their dynamics in natural populations . Previous work showed that variants that differ in protection to viruses also differed in the cost to the host in the absence of infection , indicating a trade-off between the two traits [14 , 15] . This led to the suggestion that the frequencies of different variants in natural populations might depend on the prevalence of viruses [14] . Here we demonstrate that an increase in viral burden does lead to changes in wMel variant frequencies . Moreover , the selection coefficient for specific wMel variants can be very high and promote their rapid fixation . wMel variants are strictly maternally transmitted and show no sign of recombination [13 , 29 , 30] . Therefore , as in these conditions specific haplotypes are fixed , the overall genetic diversity of wMel is strongly reduced ( since mitochondria are co-inherited with Wolbachia this selection may also impact on their genetic diversity ) . Viruses seem to impose strong natural selective pressure , as demonstrated by the fast evolutionary rates and signatures of positive selection in D . melanogaster genes involved in antiviral resistance [35] . Wolbachia can protect hosts against several positive sense single-stranded RNA viruses [3 , 4 , 23 , 24] , including DCV , a natural pathogen of D . melanogaster [36–39] . However , approximately 25 different viruses have been found to infect natural populations of D . melanogaster [38 , 40 , 41] . Although most of them are positive sense single-stranded RNA viruses we do not know which represent the biggest burden to natural populations . Moreover , the effect of Wolbachia against most of these viruses is unknown , although it protects against the few that were tested ( DCV , Cricket Paralysis virus , and Nora virus [3 , 4] ) . Different wMel variants also have different costs in the absence of infection and this is most probably an important factor in the dynamics of wMel in natural populations [14] . Our particular experimental evolution setup , with all the individuals being infected with DCV at every generation before reproduction , demonstrates that wMel selection upon viral infection is possible . In which conditions and to which degree this occurs in natural populations remains to be determined . We can explain the strong selection coefficient for clade V over clade III wMel variants with the differences in the protection to viruses they confer to their hosts . Previous analyses of virus-infected hosts carrying different wMel variants or Wolbachia strains have shown differences in viral titers and survival [14 , 15 , 23 , 32 , 33 , 42] . Here we also show that flies carrying clade V variants have lower viral titers and higher survival when compared to flies carrying clade III variants . This higher survival most likely contributes to the selection of clade V variants . However , there is also a much higher fertility of flies carrying clade V wMel variants , upon viral infection , which likely also determines the strong selection coefficient . In fact , in natural populations this parameter might be more important for the protective effect of Wolbachia against viruses and the differential selection of wMel variants , than the effect on host survival . Here , using experimental evolution , we provide direct proof that endosymbiont and host can form an evolutionary unit with adaptation relying on the evolution of both genomes . It is straightforward to extrapolate our results with maternally transmitted Wolbachia to interactions involving other defensive endosymbionts such as Spiroplasma , Regiella , and Hamiltonella [6 , 7 , 16] . The tight association between endosymbionts and their hosts make it probable that it is common for selection at the host phenotypic level to impact symbiont population genetics . It will be interesting in the future to assess to which degree this phenomenon occurs in interactions between hosts and microbes with different modes of transmission . One obvious example is the gut microbiota of mammals , which can protect the host against gut pathogens [8] and show some degree of vertical transmission [43] . As research on microbiota-induced phenotypes and potential co-evolution with hosts increases , a central question arises on how selection on the microbiota-induced phenotypes impacts the population genetics of the microbes .
We used an outbred population of D . melanogaster established in 2007 from 160 fertilized females , as described in [25 , 26] . The population was kept in laboratory conditions for more than 50 non-overlapping generations at high census . Before the initiation of experimental evolution , this population was serially expanded for two generations to allow the establishment of 36 new populations of which 12 were used in this work . All individual founders were naturally infected with Wolbachia wMel and the initial populations were 100% infected with Wolbachia ( checked individually by PCR with wsp primers , as described in [44] ) . Flies were kept in laboratory cages at constant temperature ( 25°C ) and humidity ( 70% ) in a light-darkness cycle ( 12h:12h ) . Flies were raised in standard cornmeal-agar medium . Each generation took three weeks and egg density per food cup was controlled . Virus-Selected populations were infected every generation by pricking flies in the thorax with DCV ( 2 x 107 median tissue culture infective dose ( TCID50 ) per ml ) ) [25] . DCV was grown and titrated as described in [3] . This dose caused in the initial population an average mortality of 66% 10 days after infection . Three hundred and ten males and 310 females were infected with DCV at every generation . Surviving individuals mated randomly in population cages and eggs were collected five to seven days post-infection . This selection protocol proceeded for 20 generations before Pool-Seq analysis . Control populations were pricked at every generation with sterile solution . These populations were controlled to 600 adults at every generation . Bacteria-Selected populations infection and selection protocol at every generation was the same as for the Virus-Selected populations . Flies were infected by pricking with Pseudomonas entomophila at a dose that causes an average mortality of 66% in the initial populations ( OD600 = 0 . 01 ) [26] . DNA extraction , library preparation and whole genome sequencing of pools of individuals was described in [25] . Briefly , 12 populations were sequenced ( four per regime ) : Ancestral ( generation 0 ) , Control and Virus-Selected populations , the latter two at generation 20 . Genomic DNA was extracted from a homogenate pool of 200 individuals of each population using a high-salt extraction protocol . Genomic DNA was sheared using a Covaris S2 device ( Covaris , Inc . ) and paired-end 100bp libraries were prepared using the TruSeq v2 DNA Sample Prep Kit ( Illumina ) . Libraries were sequenced on a HiSeq 2000 ( Illumina ) . Raw reads were trimmed using Trimmomatic [45] ( leading and trailing bases clipped if quality < 20 , 3’ clipped if average quality of a window ( 4 bp ) dropped below 20 , minimum read length = 50 ) and then realigned to the reference Wolbachia genome ( NC_002978 . 6 [46] ) using bwa 0 . 6 . 2 [47] , with the following parameters: maximum differences = 1% , maximum number of gaps = 2 , maximum gap or deletion size = 12 , seeding disabled . Alignments were converted to the sam/bam format using samtools [48] and sorted , filtered for quality , proper pairs and duplicate reads using bamtools [49] . Afterwards , SNPs were called simultaneously in all populations using freebayes ( v 9 . 9 . 2 ) [50] , in positions with a minimum count of the alternate allele of 2 and a minimum global alternate allele frequency of 2% . Only biallelic SNPs were considered . Effects of the polymorphisms on putative coding sequences were predicted using SnpEff 4 . 11 [51] , based on the ENSEMBL GCA_000008025 . 1 . 26 genome annotation . We analyzed the frequency of clade V wMel variants by testing individual flies in Ancestral , Virus-Selected ( at generations 5 , 10 , and 20 ) , Control ( at generation 20 ) , and Bacteria-Selected ( evolved against Pseudomonas entomophila , at generation 20 ) populations . We extracted DNA from 96 individual female flies of each replicate population following the protocol in ( http://www . drosdel . org . uk/molecular_methods . php#prep ) [52] . Briefly , single flies were squashed in 100 mM Tris-EDTA-NaCl buffer ( pH 7 . 7 ) , 0 . 5% SDS and incubated at 65°C for 30 minutes . After protein and RNA precipitation with 6M LiCl / 5M KAc , DNA was precipitated using ice-cold isopropanol followed by ethanol cleaning . PCR amplification of the genomic region surrounding position 805 , 011 was performed using the primers 805011F ( 5’-AGTCGGGAGCATGAGGGAAAAGT-3’ ) and 805011R ( 5’-TTTCAGCATCAGTCGCCTCCGC-3’ ) . The polymorphism was detected by differential cleavage of amplified product with the enzyme BtsCI ( NEB ) . Digestion was performed at 50°C for 60 minutes and the digestion product visualized in an agarose gel . The polymorphism at this position distinguishes wMel variants of clades I , II , III and IV from variants of clades V and VI . In our populations this SNP allows distinguishing clade V variants from clade I/III variants . Ninety-six isofemale lines were founded from Control populations . The Pool-Seq data show that these populations only had wMel variants from clades III and V . Each line was tested for three different wMel SNPs . Position 805 , 011 was tested as above . The SNPs at positions 655 , 839 and 1 , 027 , 577 distinguish clades I , II and III from clades IV , V , VI and VIII . PCR amplification of the genomic regions surrounding these positions were performed using the primers 655839F ( 5’-AGCAGCTCTAGCAATCGCAGCA-3’ ) , 655839R ( 5’-GGCGTTTTAGGGGTGTGGTTGGT-3’ ) , 1027577F ( 5’-TCCTGCATCAGTCCTGCCACCA-3’ ) , and 1027577R ( 5’-GGCAGCACTGTAGGCTTGACCA-3’ ) . The PCR products were digested at 37°C for 60 minutes using the restriction enzymes MscI and HindIII ( NEB ) for positions 655 , 839 and 1 , 027 , 577 , respectively . The results of the three enzymes were congruent allowing us to identify isofemale lines carrying clade V or clade III wMel variants . We also tested for the insertion IS5-WD1310 by PCR , as described in [12] . This insertion is present in clade VI variants , absent in clade III and VIII variants , but unknown for variants of other clades , including clade V [12 , 14] . All flies were negative for this insertion . After these analyses we selected eleven independent isofemale lines carrying clade V wMel variants and eleven independent isofemale lines carrying clade III wMel variants . Isofemale lines were kept in vials in similar conditions to the D . melanogaster populations . Eleven independent pairs of isofemale lines with wMel variants of clades III and V were crossed in a reciprocal scheme ( female clade V x male clade III and female clade III x male clade V ) . The female progeny of these two crosses have an equivalent genetic background but different wMel variants ( which is maternally transmitted ) . This female progeny was used for the phenotypic characterization and each reciprocal pair was considered a random effect in the statistical analysis ( “cross genotype” , see below ) . Reproductive time-window and general husbandry conditions of these crosses were the same as for the experimental evolution protocol . To analyze the contribution of mitochondria associated with different wMel clades to the fitness-related phenotypes we established Wolbachia-free lines derived from the above selected isofemale lines carrying different wMel variants . We treated ten clade III isofemale lines and ten clade V isofemale lines with tetracycline ( as in [14] ) . Lines were raised in fly food with 0 . 05 mg/ml of tetracycline hydrochloride ( Sigma ) for two generations . After antibiotic treatment each treated line had their microbiota reconstituted with the microbiota associated with their original line . 150 μl of a bacterial inoculum of each of the original lines was added to each tetracycline-treated lines . Each inoculum was constituted of 5ml of sterile water mixed with 2 g of food from a 10 days old vial of the original stock , filtered to remove eggs and larvae . All stocks were confirmed to be free of Wolbachia by PCR using primers specific for the Wolbachia gene wsp; wsp-81F ( 5’-TGGTCCAATAAGTGATGAAGAAAC-3’ ) and wsp-691R ( 5’-AAAAATTAAACGCTACTCCA-3’ ) , as in [3] . Flies were raised without antibiotics for two generations before assays . To compare the phenotype of different cytotypes in the presence or absence of Wolbachia we set up reciprocal crosses between lines carrying different wMel variants and reciprocal crosses between their matching isofemales lines after tetracycline treatment . Only ten reciprocal crosses of each kind were performed in this assay . The phenotypic assays were performed on the progeny of these crosses . For the survival assays , 100 females ( 3–6 days old ) from each reciprocal cross , infected with DCV , were placed in vials ( 10 vials with 10 individuals each ) , at 25°C . The mortality was monitored daily for 20 days . For the progeny assays , 20 couples ( 3–6 days old ) from each reciprocal cross were infected with DCV and placed in vials 5 days after infection ( 1 couple per vial ) . Flies were allowed to lay eggs for two days and then removed ( this protocol matches that of the experimental evolution ) . The progeny of each female corresponds to the number of pupae per vial . The same protocol was used for progeny quantification with females not exposed to DCV . For the quantification of Wolbachia and viral titers in the progeny of the reciprocal crosses , we used three DCV-infected females of the progeny of each matched pair . Seven days post-infection total nucleic acid was extracted using MasterPure Complete DNA and RNA Purification Kit ( Epicentre ) , according to manufacturers' protocol , with some modifications . To purify DNA , 10 μl of each sample was treated with 1 μl of 10 mg/ml RNAse A ( Roche ) . To purify RNA , samples were treated with 1U DNAse ( Promega ) per μg of total nucleic acid , in a total volume of 10 μl , at 37°C for 30 min; the reaction was stopped by adding 1 μl of RQ1 DNAse stop solution , and incubated at 65°C to inactivate the DNAse . RNA samples were then reverse transcribed to cDNA using M-MLV Reverse Transcriptase ( Promega ) , according to manufacturers’ instructions . DNA and cDNA samples were used to quantify Wolbachia and DCV levels , respectively . Quantification of Wolbachia levels and viral titers was performed by qPCR as described in [14] . For each reaction we used 6 μl of iQTM SYBR Green supermix ( Bio-Rad ) , 0 . 5 μl of each primer solution at 3 . 6 μM and 5 μl of diluted DNA . Each plate contained three technical replicates of every sample for each set of primers . Relative amounts of wsp and DCV were calculated using the Pfaffl method [53] and Drosophila Rpl32 as a reference . Levels of wsp and DCV are relative to the wMel clade V samples . Trimmed fastq and assembled bam files are available via the European Nucleotide Archive ( http://www . ebi . ac . uk/ena/about/search_and_browse ) , as project PRJEB8815 , with reads accession numbers ERS684186-ERS684197 and ERS764859–ERS764870 , respectively .
|
Animals live in close association with microbial partners that can shape many aspects of their lives . For instance , several insects carry bacteria that defend them against parasites and infectious diseases . The intracellular bacterium Wolbachia protects the fruit fly Drosophila melanogaster against viral infection . Natural populations of Drosophila carry different variants of Wolbachia , which differ from one another in the strength of this protection . Here we show that a population of Drosophila infected with viruses during several generations adapts to this challenge through turnover in Wolbachia composition . The Wolbachia variants that give higher protection to viruses , by increasing fly survival and fecundity upon infection , are strongly selected . This work demonstrates that the interaction of an animal with a pathogen can shape its associated microbial populations . We show that adaptation to pathogens can be achieved not only through selection of resistance on the host proper but also through the evolutionary shaping of its microbial community .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"invertebrates",
"organismal",
"evolution",
"medicine",
"and",
"health",
"sciences",
"pathology",
"and",
"laboratory",
"medicine",
"viral",
"transmission",
"and",
"infection",
"parasite",
"evolution",
"population",
"genetics",
"microbiology",
"animals",
"wolbachia",
"parasitology",
"animal",
"models",
"drosophila",
"melanogaster",
"model",
"organisms",
"microbial",
"evolution",
"population",
"biology",
"bacteria",
"drosophila",
"research",
"and",
"analysis",
"methods",
"pathogenesis",
"insects",
"viral",
"evolution",
"arthropoda",
"host-pathogen",
"interactions",
"natural",
"selection",
"virology",
"genetics",
"biology",
"and",
"life",
"sciences",
"evolutionary",
"biology",
"evolutionary",
"processes",
"organisms"
] |
2016
|
Drosophila Adaptation to Viral Infection through Defensive Symbiont Evolution
|
Metagenomic sequencing has contributed important new knowledge about the microbes that live in a symbiotic relationship with humans . With modern sequencing technology it is possible to generate large numbers of sequencing reads from a metagenome but analysis of the data is challenging . Here we present the bioinformatics pipeline MEDUSA that facilitates analysis of metagenomic reads at the gene and taxonomic level . We also constructed a global human gut microbial gene catalogue by combining data from 4 studies spanning 3 continents . Using MEDUSA we mapped 782 gut metagenomes to the global gene catalogue and a catalogue of sequenced microbial species . Hereby we find that all studies share about half a million genes and that on average 300 000 genes are shared by half the studied subjects . The gene richness is higher in the European studies compared to Chinese and American and this is also reflected in the species richness . Even though it is possible to identify common species and a core set of genes , we find that there are large variations in abundance of species and genes .
Metagenomic sequencing of the human microbiome has contributed to our understanding of the microbial communities that live in symbiosis with humans and their genomic capabilities [1] , [2] . The human gut microbiome is associated with a range of metabolic diseases and likely influences our physiology and nutrition [3] , [4] , [5] , [6] . To discern the associations between the gut microbiome and human health , metagenomic sequencing by generating millions of short reads from community genomes is a very powerful tool that generates vast amounts of information about the microbiome . To analyze the functional content of a metagenomic data set , its diversity and content , bioinformatics tools together with computational resources are necessary . By aligning the reads to a database of reference genomes or genes assembled de novo from the reads themselves and counting the reads on each reference sequence , a quantitative measure of the microbiome composition can be obtained . The analysis also involves preprocessing such as quality assessment and filtering out human reads . Several methods exist for either performing de novo assembly of the metagenomic data to predict gene sequences from longer contigs such as SOAPdenovo [7] , velvet [8] and MOCAT [9] which is a dedicated pipeline for metagenomic de novo assembly . The de novo assembly tools are important because the available genomic databases do not yet include complete genomes for many organisms present in metagenomic samples . Tools for taxonomic assignment of metagenomic reads have been developed and these include Phylophytia [10] , PhymmBL [11] and MetaPhlAn [12] . These tools rely on a database of reference genomes that is either used for training a classifying model or for direct alignment of sequence reads . To address the problem of quantitative characterization of a metagenome data set , we have developed a tool for quality control , filtering reads and counting alignments to reference genomes and a gene catalogue database in one step . Furthermore , downstream tasks such as handling a large number of samples and annotating the alignment counts to taxonomic and functional databases are handled . Handling an abundance table of several hundred samples and millions of gene features puts special requirements on efficient implementation . This requires a machine with a large amount of RAM and efficient data management codes . We have tested MEDUSA on four gut metagenomic datasets from three continents and evaluate its performance by mapping to two databases , one reference genome catalogue made up of 1747 bacterial and archaeal genomes and a gene catalogue constructed in this study . One important question in the field of the human gut microbiome is whether there is a common core of species and genes and how variable the microbiome is between different individuals . A core of gene functions was identified in an American population of 18 individuals but using 16S rRNA sequencing on 154 individuals did not identify a core at the species level [13] . By using metagenomic sequencing on 124 individuals from Denmark and Spain , a species core was identified and as well a core of almost 300 000 genes was identified in at least half the population [2] . An unanswered question is whether there is a core microbiome across continents . Is there a core at the species level and at the gene level ? To address these questions we used the data from four studies and found core species and genes . The core genes are also the most abundant genes but each individual also carries a large number of genes that are not shared with a majority of the population or are unique . Interestingly we found that the abundance of core species varies substantially between the studies .
MEDUSA is an integrated pipeline for analysis of short metagenomic reads , it contains modules for mapping reads to reference databases , combining output from several sequencing runs and manipulating the tables of read counts and testing for differential abundance ( Figure 1a ) . Python was used for creating a pipe to stream metagenomic reads stored in fastq files ( can be compressed with gz , bzip2 or in SRA archives ) through a quality control step , filtering out human reads and mapping reads to two databases simultaneously , without the need for writing intermediate files ( Figure 1b ) . By streaming reads in a pipe , time consuming disk IO is eliminated and disk space is saved . MEDUSA also contains tool for combining and analyzing a table of counts in numpy which facilitates a fast framework for manipulating a table that had several hundred by several millions entries . These tools include performing rarefaction to sample the reads to the same depth of sequencing , testing for differential relative abundance and plot relative abundance for selected features . The reference catalogues used can be a gene catalogue and a genome catalogue and this approach has been used previously [2] , [3] . MEDUSA can merge count tables of genes and genomes with annotation information to generate a KEGG ortholog abundance and taxonomic table . In this study , four of the largest published gut metagenomic datasets to date were included and compared . The subjects are from United States of America ( Human microbiome project , HMP ) [1] , China [4] , Denmark , Spain ( MetaHIT ) [2] and Sweden [3] , all together containing 40 billion metagenomic reads and 782 samples . All samples were sequenced on the Illumina platform with read lengths from 44 to 100 base pairs . A non-redundant catalogue of species genomes was constructed based on the results of a method using 40 universal single copy phylogenetic marker genes used for clustering prokaryotic genomes into species [14] . The catalogue contains 1747 species genomes downloaded from NCBI Genbank and the full list of genomes is presented in Table S1 . The quality controlled and filtered reads were aligned to the genome catalogue and the number of aligning reads to each contig in the database was counted . Reads files from the four studies were used as input to the function streamAligner . This function can take a number of compressed fastq files as input and will produce a count file for each input file and reference database . The function produces a log file for each input file with mapping statistics and output from the various software used in the stream such as fastx and Bowtie2 . The function streamAligner can easily be parallelized by starting many instances of the function; each instance will look in the list of files supplied and start working on unprocessed files given that all instances have access to the same file system . The input number of reads for each study were on average 40±12 , 102±28 , 45±18 and 31±18 million single end reads per sample for the studies China , HMP , MetaHIT and Sweden , respectively . Most of the sequencing runs have a high quality with almost 98% of the reads passing the quality cutoff ( Figure 1c , Table S2 ) . Out of the high quality reads , on average only 0 . 023% aligned to the human genome although the HMP data had been cleaned for human reads before submission to a public database . It is worth to note that the degree of human reads in a sample is highly variable with a few samples with considerable fraction of human reads and therefore the filtering of human reads is important even in gut metagenome datasets where the fraction of human reads is low compared to data from other body sites [1] . Out of the HQ non-human reads , 75% could align to the gene catalogue while 39% could be aligned to the genome catalogue which is similar to previous results or alignment to gene and genome catalogues [2] , [3] . This indicates that there are still species in the gut that have not yet been identified . The function combineCounts takes a range of input files and a file mapping sequence runs to a sample since some samples could be sequenced in several runs . The output of combineCounts is a large abundance matrix which has aligned features as rows and samples as columns . We compared our results of the genus abundance to another tool , Metaphlan [12] which uses clade specific marker genes from reference genomes for taxonomic profiling of metagenomes . HMP samples profiled with Metaphlan were compared to the results using MEDUSA on the genus level and the comparison accounts on average for 99 . 5±0 . 46% and 98 . 1±2 . 1% of the reads aligned reads , respectively . Comparing the 137 samples that were shared , we find that the Pearson correlation between the profiles are 0 . 95±0 . 06 ( Table S3 ) , indicating that the two methods produce very similar results . Performance of Metaphlan has been reported to be 450 reads per second on a single CPU [12] . MEDUSA was here performing with a throughput of 938 reads per second ( AMD Opteron 6220 ) , but then quality control , human filtering and alignment to the reference genomes and gene catalogue were done simultaneously . The taxonomic profiles at the species and genus level of all samples were determined by analyzing the aligned reads to reference genomes . The most abundant genus in the cohort was Bacteroides but the inter-individual variation was large spanning from almost 1 to 0 ( Figure 2a ) , the top 20 most abundant genera account for 93±8% of the annotated reads . The most abundant species were from Bacteroides , Faecalibacterium and Eubacterium with inter-individual variations in abundance spanning several orders of magnitude ( Figure S1 ) . The abundance of Bacteroides was higher in HMP and Chinese samples compared to Metahit and Swedish samples and the latter had higher abundance of Ruminococcus ( Figure 2b ) . The abundance of other genera also varied across study populations and in general the Swedish and to some extent the Metahit population had more Firmicutes , e . g . Faecalibacterium , Eubacterium , Clostridium and Dorea ( Figure S2 ) . Analyzing the diversity of the species found in the samples shows that the diversity is highest in the Swedish samples followed by MetaHIT which are also less dominated by Bacteroides . Heatmaps of species and genera abundance are shown together with a clustering of samples in Figure S3 and S4 . Using the species abundance profiles to calculate the diversity of species shows that MetaHIT and Swedish samples have a higher diversity compared to American and Chinese . The higher diversity in these samples is likely due to a smaller dominance by Bacteroides which is not replaced by one species or genera but several different Firmicutes species . To address whether there is a core of species that is shared by subjects from the different cohorts , we looked at species with a relative abundance above 0 . 0001 across subjects and found 116 species above this threshold in 50% of the subjects and 71 species above the threshold in 90% of the subjects ( Figure 2d and Table S4 ) . This indicates that there is a common core of species shared across all cohorts but their abundance differs extensively . Since the size of the species core have been shown to be affected by the depth of the analysis using the HITChip [15] we investigated the sensitivity using metagenomic sequencing . The performed analysis shows that the size of the core is relatively insensitive to the cutoff used for abundance ( Figure S5 ) . Three enterotypes or clusters of stratified intestinal microbiota composition were suggested [16] and here we investigate the existence of enterotypes in the combined cohorts . The strongest support was found for three clusters with an average Silhouette width of 0 . 29 ( Figure S6 ) . The driver genera were Bacteroides , Prevotella and Ruminococcus as originally proposed ( Figure S7 ) . However , the three enterotypes were strongly associated with the 4 study cohorts , China and HMP samples were enriched in enterotype 1 , Metahit evenly distributed among the three and Sweden enriched in enterotype 3 ( Table S5 and Figure 2b ) . When studying only the Danish samples from the Metahit cohort and comparing to the outcome in the original population , there is a 96% agreement between the clustering results ( Table S6 ) . Ranking the subjects according to their relative abundance of Bacteroides indicates that there is a smooth gradient but Prevotella shows a bimodal distribution indicating that subjects fall into primarily two categories with the abundance either being >10% or <1% ( Figure S8 ) . We extended the human gut microbial gene catalogue by merging data from the four different gut metagenome studies . Contigs from each study were downloaded and genes were predicted , in total 72 . 5 million genes were predicted . 67 million genes were predicted from the individual assemblies of samples and 5 . 5 million genes were predicted from the global assemblies that were performed on unassembled reads ( Figure S9 ) . Genes from each individual study were then clustered based on their sequence similarity using Uclust [17] and a 95% identity and 90% coverage cutoff . In a final step , the NR genes from each study were then clustered using the same criteria as above and a global human gut microbial gene catalogue was obtained containing 11 million genes . Each study showed a substantial number of unique genes while the common genes to all studies was 488 482 and 2 . 7 million genes were shared between any two studies whereas almost 9 million genes were unique to a single study ( Figure 3 ) . The largest number of unique genes was found in the HMP samples and these were also the deepest sequenced . The lowest number of unique genes was found in the Chinese cohort on which a global assembly of unassembled reads from individual assemblies was not done . The largest overlap between two studies was found between the Swedish and HMP studies with over 1 . 5 million shared genes . Although each study contained many unique genes from de novo assembly , we wanted to study the abundance of the shared and unique genes in each subject . To get a quantitative measure of gene abundance , reads were mapped back to the gene catalogue as described above and in Methods . On average 38±8% of reads in each sample mapped to the core genes ( 488 482 ) found in all studies ( Figure 3 ) . A similarly large part of reads mapped to study-unique genes ( 36±4% ) . This indicates that there is a substantial part of the microbiome that is shared but also that low abundant genes are unique to individuals . If the abundance is also normalized to the number of genes in each category it is clear that the most abundant genes are shared ( Figure S10 ) . To determine the richness of the microbiota using the gene catalogue , aligned reads were counted and two reads were required to call a gene present in a sample . Comparison of the gene richness in the 4 studies shows that the European samples have a higher gene count compared to Chinese and HMP samples ( Figure 4a ) . When counting genes , all samples were rarefied to the same number of reads , 11 million , in order to remove the effect of different sequencing depth and 23 samples were removed because of limited sampling depth . Regardless of rarefaction , European samples showed a higher gene richness compared to Chinese and HMP samples . Recently the gene richness has been associated with lower BMI and favorable metabolic markers in a study of Danish subjects [6] . All HMP subjects are reported to be healthy but still show a markedly lower gene richness compared to the two European cohorts . Since the gene richness is so closely associated with the different studies , we did not investigate any associations between gene richness and health status , as methodological differences cannot be ruled out . In a study of American twins , the association between gut microbiota richness and obesity has also been reported previously using 16S rRNA sequencing [13] . Low diversity of the microbiota has been reported to be associated with inflammatory bowel disease [18] and inflammation in elderly [19] . A comparison of the diversity between populations also found that American subjects had a less diverse microbiota compared to Amerindians from Venezuela and Malawians [20] . The differences became evident after 3 years of age , but not in younger subjects . Despite differences in diversity , there is a core of genes found in a majority of the subjects . By counting the genes present in at least 50% of the population we found 283 705 genes which indicated that a large portion of the genes carried by an individual is shared . In the original MetaHIT study of 124 subjects , each individual carried just above 536 112 genes on average [2] . A core of genes was identified of 294 110 genes being present in at least half the MetaHIT population which also means that a large number of genes were only found in one or a few subjects . However , there are only 3 genes shared by all subjects of this study ( Figure 4b , c ) . The number of genes shared by at least 50% of the subject is stable when more subjects are added and it can therefore be expected that this number will be stable also when more subjects are included . However , the number of core genes is highly dependent on the fraction of subjects required to carry the gene ( Figure 4c ) e . g . there are 1 . 3 million genes shared by at least 20% of the population . The pan genome is quickly increasing by the number of subjects which also means that most genes are shared by at least 2 individuals and in fact over 10 million genes are found in at least 2 individuals . The genus origin and functional potential of the core genes were compared to those of all genes in the catalogue . The fraction of genes with an unknown genus origin is lower in the core genes compared to all genes in the catalogue ( 13% compared to 31% , respectively ) ( Table S8 ) . The core genes were 20% from Bacteroides and 13% from Clostridium origin and these two genera were also the most common annotated genera in the full gene catalogue . At the functional level , a higher fraction of genes could be assigned to a gene in KEGG . A wide set of KEGG KOs had a higher annotation frequency to the core genes ( Table S9 ) . These functions include biosynthesis of secondary metabolites , amino acids and starch and sucrose metabolism . In summary , on average there is a shared common pool of genes but there is also a large number of genes in each individual that is shared with very few but are not completely unique .
The higher abundance of Bacteroides in the HMP and Chinese subjects compared to the European subjects can be due to differences in lifestyle , age , disease state , antibiotic use and diet . Bacteroides abundance has been associated with a diet high in animal protein , amino acids and saturated fats suggesting high meat consumption , Prevotella was found to be associated with high intake of carbohydrates and simple sugars [21] . It has also been observed that a diverse diet is associated with a diverse microbiota in an elderly population [19] . The gene catalogue presented here could be used for mapping of metagenomics sequence reads in future studies as it spans a large and diverse population . It clearly shows that there is a common core of genes across continents and populations although there are a many genes that are only found in few subjects . This indicates that more genes will be found when new subjects are studied but it is likely that these genes will have a very low abundance as the core genes found here have a high relative abundance . Possibly , some of the genes found in few individuals are transient genes whereas the core genes are more stable over time . The stable species of the microbiota has been found to be also the most abundant part by a 16S rRNA study using low error prone sequencing technology [22] . Differences in microbiota richness seen here between the European and Chinese and HMP studies can be due to a number of reasons . Antibiotic use , diet and other lifestyle effects are possible reasons for this difference . Also , methodological differences in sample collection and DNA extraction could influence sample richness and composition . The effect of antibiotics at subtherapeutic levels in mice is reduced diversity [23] and also in humans antibiotic use have been shown to have a major impact on the microbiota and reduced diversity [24] . The difference in diversity between the MetaHIT and HMP samples have also been seen in a previous study using phylogenetic marker genes [25] . In this study , this trend was seen both in species and gene richness and especially pronounced in the gene richness . It is likely that HMP samples which were sequenced to a greater depth have a higher proportion of their microbiome represented in the assemblies; this is also reflected in the large number of genes assembled from the HMP samples . However , the number of genes seen with a normalized number of reads is still substantially less than in the European samples . In conclusion , we here present the MEDUSA pipeline , a tool for metagenomic data analysis with possibility for simultaneous taxonomic and gene annotation and handling of large data sets . We have applied this tool to perform the first comparison of four large studies from three continents and found a common species and gene core although the abundances of core components differ between populations . Furthermore , we provide a gene catalogue spanning over 11 million genes constructed from the different populations .
MEDUSA was implemented in python programming language and requires the numpy package ( http://www . numpy . org/ ) . MEDUSA makes use of standalone tools such as FASTX , bowtie2 [26] and GEM [27] that need to be callable from the Unix command line . The MEDUSA pipeline together with databases and results are available at http://www . metabolicatlas . com/medusa . A non-redundant catalogue of genomes from prokaryotic species was constructed by using the results from grouping of prokaryotic genomes into species [14] . For each species , the longest of its member genomes was chosen as representative and the genome downloaded from NCBI Genbank . 8 genomes from the list were excluded as these records had been changed or retracted since the creation of the list of non-redundant species . All downloaded contigs were merged into a single fasta file and indexed by gem-indexer . The catalogue was annotated to NCBI taxonomy using the function annotateToNCBITaxonomy which creates an output file with taxonomy ids and taxnomomic names to each record in the reference catalogue . Four large metagenome studies were included in the construction of a global gut microbial gene catalogue . Assembled contigs were downloaded for the four studies [1] , [2] , [3] , [4] . Genes were predicted on the contigs using Metagenemark [28] . Usearch [17] was used for constructing non-redundant sets of genes with 95% sequence identity and 90% coverage of the shorter sequence . This cutoff groups homologous genes from strains of the same species together but does generally not group more distantly related genes such as a protein family . A catalogue for each study was first constructed and then these were merged into a global catalogue . In this study , 782 human gut metagenomes were analyzed from four different studies , Sweden [3] , MetaHIT [2] , HMP [1] and China [4] . All samples were analyzed with the Illumina sequencing technology and a total of 40 billion reads were analyzed ( Table S2 ) . Some of the HMP subjects were sequenced on up to three occasions ( Table S7 ) . Each sequencing run was analyzed using the streamAligner function in MEDUSA and paired end reads were treated independently . Sequencing runs were merged into samples with the function combineCounts using a mapping file linking sequence runs to samples . The function annotateCounts was used on the gene count table to annotate counts to NCBI taxonomy and creating species and genus abundance tables . Genes were considered present if two reads from the same sample aligned to it which is the same criteria used in by Qin et al . [2] . To normalize the sampling depth , the MEDUSA function rarefy was used to sample 11 million aligned reads from each subject . In the analysis of core species and genes , HMP samples from visit 2 and 3 were removed to make sure that the core is defined on the individual basis and this reduced the number of samples from 782 to 719 . The minimum relative abundance of a species to be counted as present in the core was 10−4 and the sensitivity to this cutoff for core species is shown in Figure S5 . Enterotypes were determined using the genus abundance with the methods suggested in http://enterotype . embl . de/ and in the paper by Arumugam et al [16] , the analysis was performed in R using the package ade4 . Data and software tools can be accessed through http://www . metabolicatlas . com/medusa .
|
Our bodies are home to a myriad of microbial cells and our intestinal tract is especially densely populated with bacteria . Alterations in the composition of the gut microbiota have been associated with common human diseases . By sequencing the genomes of the microbes , the metagenome , detailed information about who is there and their capabilities can be obtained . In this paper , a method for analyzing metagenomic data is presented together with an analysis of gut metagenomes from 4 different studies and 3 different continents . We identify a core set of genes and species were identified but the abundance of core components differs between study populations . A catalogue of gut microbial genes from the 4 studies was constructed containing more than 11 million genes .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"metagenomics",
"genetics",
"biology",
"and",
"life",
"sciences",
"genomics",
"computational",
"biology"
] |
2014
|
Metagenomic Data Utilization and Analysis (MEDUSA) and Construction of a Global Gut Microbial Gene Catalogue
|
Myosin-I molecular motors are proposed to function as linkers between membranes and the actin cytoskeleton in several cellular processes , but their role in the biosynthesis of fungal secondary metabolites remain elusive . Here , we found that the myosin I of Fusarium graminearum ( FgMyo1 ) , the causal agent of Fusarium head blight , plays critical roles in mycotoxin biosynthesis . Inhibition of myosin I by the small molecule phenamacril leads to marked reduction in deoxynivalenol ( DON ) biosynthesis . FgMyo1 also governs translation of the DON biosynthetic enzyme Tri1 by interacting with the ribosome-associated protein FgAsc1 . Disruption of the ATPase activity of FgMyo1 either by the mutation E420K , down-regulation of FgMyo1 expression or deletion of FgAsc1 results in reduced Tri1 translation . The DON biosynthetic enzymes Tri1 and Tri4 are mainly localized to subcellular structures known as toxisomes in response to mycotoxin induction and the FgMyo1-interacting protein , actin , participates in toxisome formation . The actin polymerization disruptor latrunculin A inhibits toxisome assembly . Consistent with this observation , deletion of the actin-associated proteins FgPrk1 and FgEnd3 also results in reduced toxisome formation . Unexpectedly , the FgMyo1-actin cytoskeleton is not involved in biosynthesis of another secondary metabolite tested . Taken together , this study uncovers a novel function of myosin I in regulating mycotoxin biosynthesis in filamentous fungi .
Fusarium head blight ( FHB ) caused predominately by Fusarium graminearum is an economically devastating disease of small grain cereal crops [1] . This disease not only reduces yield and seed quality but also poses a great risk to human and animal health owing to its ability to contaminate grains with mycotoxins . The common mycotoxins associated with F . graminearum are deoxynivalenol ( DON ) , nivalenol ( NIV ) and zearalenone ( ZEA ) [2] . Among them , DON is the most frequently detected mycotoxin in cereal grains throughout the world [3] . DON can inhibit protein synthesis by binding to the ribosome , and cause emetic effects , anorexia and immune dysregulation as well as growth , reproductive and teratogenic effects in mammals [4] . To minimize human and animal exposure to DON , regulatory organizations have established maximum permissible levels for DON in cereals and their products in many countries [5 , 6] . However , DON contamination has become a challenging social issue because of the increased frequency and severity of FHB epidemics [7 , 8] . DON contamination is closely linked to the severity of FHB disease in the field . The best way to prevent DON contamination would be to manage FHB in the field during crop cultivation . Currently , application of chemical fungicides is still a major approach against F . graminearum infection due to the lack of highly resistant wheat cultivars . However , application of several commercialized fungicides at sub-lethal concentrations could trigger DON biosynthesis [3 , 9–11] . Recently , a novel cyanoacrylate fungicide phenamacril ( JS399-19 ) has been marketed for FHB management and sale of phenamacril in China was approximately $40 million in 2016–2017 . Interestingly , this small molecule compound ( S1 Fig ) exhibits highly specific antifungal activity against mycelial growth of a few Fusarium species including F . graminearum , F . asiaticum , F . verticillioides and F . oxysporum but not other fungal pathogens [12] . It shows excellent efficacy in controlling FHB in field trials [12 , 13] . Combining inferences from genetic and biochemical results , we recently discovered that this compound acts on a novel target , the class I myosin ( FgMyo1 ) in F . graminearum , which is homologous to Myo3p and Myo5p in Saccharomyces cerevisiae [12] . FgMyo1 is essential for F . graminearum growth . At the beginning of this study , we found that phenamacril not only suppressed the mycelial growth of F . graminearum , but also significantly inhibited DON production . These preliminary results suggested that the myosin I might also be involved in the secondary metabolism . Class I myosins are widely expressed , single headed and membrane-associated members of the myosin superfamily that participate in regulating membrane dynamics and structure in nearly all eukaryotic cells [14 , 15] . However , the underlying function of myosin I in mycotoxin biosynthesis was totally unknown . Enzymes for secondary metabolite synthesis may be compartmentalized at conserved sub-cellular sites in fungi , potentially channeling precursors , sequestering intermediates and products from the rest of the cell , thus promoting the efficiency of biosynthesis pathways [16] . In Penicillium chrysogenum , the major facilitator-type secondary transporter PenM promotes translocation of isopenicillin N from the cytosol to the peroxisomal lumen where it could be further metabolized to penicillin [17] . In Aspergillus , aflatoxin biosynthetic enzymes flow from peroxisomes to the motile vesicles termed aflatoxisomes in which aflatoxin biosynthesis takes place [18] . In F . graminearum , and other Fusarium spp , the biosynthetic pathway leading from the isoprenoid intermediate farnesyl pyrophosphate to DON involves 15 genes encoding the biosynthetic enzymes , a DON transporter and regulatory proteins , which are located on different chromosomes: the 25 kb Tri5 cluster containing 12 genes on chromosome 2 , the Tri1-Tri16 locus with two genes on chromosome 1 and the single gene locus for Tri101 on chromosome 3 [19–21] . Recent studies suggested that there is a cellular compartmentalization of biosynthetic enzymes for DON biosynthesis in F . graminearum [22] . Hydroxymethylglutaryl ( HMG ) CoA reductase ( Hmr1 ) is a key enzyme in the mevalonate pathway for generating farnesyl pyrophosphate and indispensable for DON production . Fluorescent labeled Hmr1-GFP localized to the reticulate peripheral and perinuclear endoplasmic reticulum ( ER ) in toxin non-inducing conditions , while the ER was remodeled to form spherical and ovoid structures in the trichothecene biosynthesis inducing ( TBI ) conditions [16 , 22] . In addition , the enzymes trichodiene oxygenase ( Tri4 ) and calonectrin oxygenase ( Tri1 ) catalyzing the early and late steps in the DON biosynthetic pathway were co-localized and showed the same localization patterns as Hmr1 in TBI medium [22 , 23] . These novel cellular structures containing DON biosynthesis enzymes were named "Fusarium toxisomes" ( “toxisomes” in shorter form in this study ) [16 , 22 , 23] . However , the molecular mechanism of toxisome formation remains elusive . The object of this study was to uncover the underlying mechanism of a myosin I inhibitor in regulating DON biosynthesis . Our results showed that myosin I plays critical roles in the translation of a Tri enzyme and in toxisome formation in F . graminearum . The importance of myosin I in the development of the mycotoxin biosynthetic machinery in F . graminearum may apply to other toxigenic pathogens .
TRI1 encodes calonectrin oxygenase that catalyzes calonectrin to 7 , 8-dihydroxycalonetrin , which is a late step of DON biosynthesis in F . graminearum [24] . To characterize expression patterns and the sub-cellular localization of Tri1 protein under various conditions , the TRI1 open reading frame tagged with GFP ( green fluorescent protein ) was introduced into a ΔTri1 F . graminearum PH-1 background , and the complemented strain expressing the Tri1-GFP ( ΔTri1::Tri1-GFP ) was used in the following study . In the toxin non-induction minimal ( MM ) or potato dextrose broth ( PDB ) media , Tri1-GFP displayed faint signals and was mainly associated with cell endomembrane ( Fig 1A , left and middle panels ) . Tri1-GFP was highly induced and localized at the spherical structures ( toxisomes ) after 48 hours of incubation in the trichothecene biosynthesis induction ( TBI ) medium ( Fig 1A right panel; S2 Fig ) and in planta ( Fig 1G , left panel ) . In addition , ER ( endoplasmic reticulum ) -tracker red staining indicated that Tri1-GFP was mainly localized to the ER in TBI cultures ( Fig 1B ) , which is consistent with a previous finding that the toxisomes were identified as reorganization of the endoplasmic reticulum [22] . To determine whether the spherical structures were associated with the nucleus , we visualized nuclei by tagging the histone1 protein encoded by the FGSG_10800 locus with red fluorescent protein ( RFP ) , which was designated as H1-RFP in the PH-1::Tri1-GFP strain . The H1-RFP/Tri1-GFP dual labeled strain was grown in TBI for 48 h , and localization of H1-RFP with Tri1-GFP was examined . As shown in Fig 1C , Tri1-GFP surrounded the H1-RFP labelled nuclei when the strain was cultured in the TBI medium . Moreover , the trichodiene oxygenase ( Tri4 ) catalyzing the early step of DON biosynthesis had the same localization pattern as Tri1 ( S3 Fig ) . Taken together , several lines of evidence suggested that trichothecene biosynthetic enzymes were clustered and localized to toxisomes derived from ER under the toxin inducing conditions . Since the toxisomes are important for DON biosynthesis , a compound disrupting the toxisome formation may be very well effective against DON biosynthesis . To test this hypothesis , we established a "toxisome formation inhibitor screening" assay to quickly screen active compounds for their ability to restrict toxisome formation ( see Material and methods ) . Briefly , the reporter strain expressing Tri1-GFP was grown in 24-wells plates supplemented with TBI medium . After 24 h incubation , individual compounds were added to wells . After incubation for another 24 h , the fluorescent intensity in each well was scanned with the plate-reader for the first round of screening . The wells with low or no fluorescent signals were further observed by microscopy . A total of 131 compounds including 11 commercial fungicides were tested for their activity against toxisome formation . Phenamacril was found to be the most efficient compound to inhibit toxisome formation and DON production ( S4B and S4C Fig ) . The Tri1-GFP fluorescent signals were reduced dramatically and no typical toxisomes were observed in the mycelia treated with 0 . 5 μg/ml ( approximately EC90 against mycelial growth ) phenamacril for 6 h ( S5B Fig ) or 24 h ( Fig 1D ) in comparison with those in the non-treatment control . In addition , the beta-tubulin inhibitor , carbendazim , did not inhibit toxisome formation ( Fig 1D ) . As shown in Fig 1E , the translation levels of Tri1-GFP were further verified by immunoblot assay using an anti-GFP antibody . Consistent with the microscopic observation , the intensity of the Tri1-GFP band from the strain treated with carbendazim increased more than 2-fold as compared with the non-treated control . In contrast , a faint immunoblot band was detected in the same strain treated with phenamacril ( Fig 1E ) . Correspondingly , DON in the mycelia treated with phenamacril was below the level detectable by LC-MS ( liquid chromatography-mass spectrometer ) ( Fig 1F ) . Furthermore , we tested the efficiency of phenamacril against DON production in planta and in the field . As shown in Fig 1G , phenamacril also clearly inhibited toxisome formation in hyphae of F . graminearum inoculated on wheat leaf . In the field trials , this antifungal compound was very effective against FHB and DON production in comparison with the control chemical carbendazim ( Fig 1H ) . The class I myosin ( named FgMyo1 ) of F . graminearum has been identified as the target of phenamacril [12] . Taken together , these results strongly indicated that the myosin I inhibitor phenamacril was able to inhibit DON biosynthesis in F . graminearum . Given that the myosin I inhibitor significantly reduces DON biosynthesis , myosin I may be critical for toxisome formation . In order to verify this , we tagged FgMyo1 with RFP to determine its subcellular localization . In toxin non-induction media MM and PDB , FgMyo1-RFP protein was detected as diffuse fluorescent signal in the cytoplasm , mainly localized at hyphal tips ( Fig 2A ) . However , in the TBI medium , most FgMyo1-RFP fluorescence accumulated in subapical spherical structures ( Fig 2A , right panel ) . To determine whether myosin I was localized to the toxisomes , a strain labeled with FgMyo1-RFP and Tri1-GFP was constructed and cultured in TBI . As indicated in Fig 2B , both proteins were mainly co-localized at the toxisomes . Additionally , Co-IP and BiFC ( Bimolecular Fluorescence Complementation ) assays showed that FgMyo1 interacted with Tri1 in toxin inducing condition ( Fig 2C and 2D ) . Affinity capture mass spectrometry ( ACMS ) was then used to identify interacting proteins upon toxin-induction conditions using the dual tagged protein ZZ-Tri1-Flag as the bait . In the ACMS assay , FgMyo1 was captured by Tri1 ( S1 Table ) . Furthermore , ten of the 30 Tri1-interacting proteins were described previously [22] as components of the toxisome , including the three cytochrome P-450 enzymes Tri1 , Tri4 and Tri11 as well as HMG-CoA reductase . These results indicated that FgMyo1 interacts with Tri1 and thus has the potential for involvement in toxisome formation . To verify the role of FgMyo1 in toxisome formation , we used a knock-down approach because FgMYO1 is an essential gene in F . graminearum [12] . First , we took the advantage of the RNA interfering ( RNAi ) pathway to induce FgMYO1 silencing with hairpin RNA ( hpRNA ) , which has been proven to be efficient in knockdown of mRNA expression for target genes in F . graminearum [25] . The recombinant plasmid pSilent-FgMYO1 , designed for generating the hpRNA of an FgMYO1 fragment ( 540 bp ) , was introduced into the wild-type PH-1 . Predicting that transformants with reduced expression of FgMYO1 may grow poorly on the medium supplemented with phenamacril , we screened for transformants with increased sensitivity towards this compound and then verified the FgMYO1 expression level by reverse transcription-PCR . Among the 20 transformants tested , four showed increased sensitivity to phenamacril , and the expression levels of FgMYO1 were decreased 65%-90% in these silencing transformants in comparison with the wild type . The FgMYO1-S2 transformant , having the lowest FgMYO1 expression ( 10% of the parent strain PH-1 ) , was selected for further characterization . It had normal growth rate on PDA but failed to grow on PDA supplemented with phenamacril at 0 . 3 μg/ml ( approximately EC50 against mycelial growth ) ( Fig 2E ) . As expected , the toxisome formation indicated by Tri1-GFP was significantly impaired and only faint fluorescent signals were observed in the mycelia of FgMYO1-S2 harboring Tri1-GFP ( Fig 2F ) . Next , we replaced the native promoter of FgMYO1 with the zearalenone ( ZEA ) -inducible promoter ( Pzear ) [26] to generate a transformant that conditionally expressed FgMYO1 . The resulting transformant ( termed as Pzear-FgMYO1 ) without ZEA induction was unable to grow on PDA supplemented with 0 . 3 μg/ml phenamacril ( Fig 2E ) . Consistently , this strain formed very faint toxisomes in TBI without the inducer as compared to the wild type ( Fig 2F upper panel ) . The defects in mycelial growth and toxisome formation of Pzear-FgMYO1 were partially recovered by adding the inducer β-estradiol ( Fig 2E and 2F , upper panel ) . In addition , translation levels of Tri1-GFP protein in above strains quantified by the western blotting assay were consistent with fluorescent signals ( Fig 2F , lower panel ) . All of the above mutants , whether constructed by silencing or conditional expression , revealed significantly reduced DON production in TBI ( Fig 2G ) . Since DON is a critical virulence factor and plays a significant role in the spread of pathogen within host tissues [27–29] , it follows that each of these strains was severely attenuated in virulence toward flowering wheat heads ( S6 Fig ) . These results confirmed that FgMyo1 plays an important role in toxisome formation . To gain an insight into the function of FgMyo1 in toxisome formation , we further conducted an ACMS assay using the dual tagged protein ZZ-FgMyo1-Flag as the bait . In the ACMS assay , the ribosome-associated protein Asc1 ( hereafter named FgAsc1 , ) was captured by FgMyo1 . Unexpectedly , FgAsc1 was also pulled down by Tri1 ( S1 Table ) . In addition , the interaction of FgMyo1 and FgAsc1 was confirmed by Co-IP assay ( Fig 3A , left panel ) , while the directly interaction between these two proteins was not verified by BiFC . Given that the translation level of Tri1-GFP protein was inhibited dramatically by phenamacril ( Fig 1D ) , and Asc1 is a conserved ribosomal protein and is required for efficient protein translation [30 , 31] , we inferred that FgMyo1 might regulate Tri1 translation via interacting with Asc1 . To test this hypothesis , we first examined co-localization of FgMyo1-GFP and FgAsc1 tagged with RFP . In toxin non-inducing conditions , FgAsc1-RFP was detected as diffuse fluorescent signal in the cytoplasm ( Fig 3B ) . However , in the TBI medium , most FgAsc1-RFP accumulated in spherical structures and co-localized with FgMyo1 ( Fig 3A , right panel ) and also with Tri1 at the perinuclear positions ( Fig 3B , lower panel ) . Since Asc1 is ribosome-associated protein , to further visualize localization of ribosomes , FgRpL25 ( an essential component of 60S subunit of ribosome [32 , 33] ) was tagged with mCherry under the control of its own promoter , and transformed into the wild type . Confocal microscopic examination showed that most FgRpL25-mCherry accumulated at the perinuclear positions in the toxin inducing conditions ( S7 Fig , bottom panel ) . In contrast , FgRpL25- mCherry was mainly localized in the cytoplasm in the toxin non-inducing conditions ( S7 Fig , upper panel ) . These results indicated that FgMyo1 interacts with the ribosome protein FgAsc1 in toxin inducing conditions . To further understand the role of FgAsc1 in Tri1 translation , we constructed a deletion mutant of FgAsc1 . As expected , the mutant exhibited dramatically reduced hyphal growth ( Fig 3C , upper panel ) . The translation level of Tri1-GFP in this mutant was decreased markedly in comparison with that in the wild type ( Fig 3D , right panel ) , and subsequently , toxisome formation and DON production was not detected in this mutant cultured in TBI medium ( Fig 3D , left panel; Fig 3E ) . It is very interesting that the translation level of FK506-binding protein Fg_Fkbp54 ( FGSG_01059 ) was not altered in ΔFgAsc1 as compared to that in the wild type ( Fig 3F , right panel ) , suggesting that FgAsc1 controls the translation of some proteins ( at least Tri1 ) but not all proteins , which is agreement with a role for Asc1 in regulating the translation of specific mRNAs in S . cerevisiae [31 , 34] . Taken together , these results indicated that FgMyo1 was indispensable for translation of Tri1 protein by interacting with the ribosome protein FgAsc1 . In a previous study , we found that the ATPase activity of FgMyo1 is dependent on actin . FgMyo1E420K bearing a mutation at the actin interacting domain of FgMyo1 , which caused the actin-activated ATPase activity of FgMyo1E420K was reduced to 5% as that of the wild-type FgMyo1 [12] . Correspondingly , toxisome formation in this strain was markedly decreased in comparison with that of the wild type ( Fig 2E , upper panel ) . Moreover , we found that components of the actin cytoskeleton were enriched in the ACMS with Tri1 and FgMyo1 as the bait ( S1 Table ) , suggesting that actin cytoskeleton may be associated with toxisome formation in F . graminearum . To address this possibility , we further constructed a strain bearing actin-RFP and Tri1-GFP . Then , the interaction between actin-RFP and Tri1-GFP was further verified by Co-IP assay ( Fig 4A ) . In S . cerevisiae , the myosin I interacts with the actin and is required for polarization of the actin cytoskeleton [35] . Consistent with what is known in S . cerevisiae , actin was also associated with FgMyo1 in the ACMS assay using FgMyo1 as the bait ( S1 Table ) . In addition , the interaction of FgMyo1-GFP and actin-RFP was further confirmed by the Co-IP assay ( Fig 4B ) . Since actin is essential for F . graminearum growth , we were unable to obtain a knockout mutant of the ACTIN gene . Thus , to further investigate the function of actin in DON biosynthesis , the actin polymerization inhibitor latrunculin A was used to mimic impaired function of the actin cables . After treatment with latrunculin A at 0 . 1 μg/ml ( approximately EC90 against mycelial growth of F . graminearum ) , the typical toxisome structures could not be observed , and Tri1-GFP was detected as diffuse fluorescent signal in the cytoplasm ( Fig 4C , left panel ) . In addition , Tri1-GFP was noticeably decreased in the western blot assay upon latrunculin A treatment ( Fig 4C , right panel ) . Subsequently , latruncunlin A showed strong inhibition of DON production ( Fig 4D ) . These results indicated that the actin cytoskeleton is involved in toxisome formation in F . graminearum . In S . cerevisiae , Prk1 and End3 are involved in the organization of the actin cytoskeleton [36 , 37] . To better understand the roles of the myosin I-actin cytoskeleton in toxisome formation , we therefore were interested in constructing deletion mutants of their orthologs FgPrk1 ( FGSG_05586 ) and FgEnd3 ( FGSG_09721 ) . Toxisome formation in mycelia of these two gene deletion mutants harboring the tagged Tri1-GFP was examined . The Tri1-GFP signals decreased noticeably in both ΔFgPrk1 and ΔFgEnd3 mutants ( Fig 5A , left panel ) . In addition , western blot assays confirmed the amount of Tri1-GFP protein in these mutants was considerably lower than that of the wild type under the toxin inducing condition ( Fig 5A , right panel ) . Furthermore , these mutants produced significantly less DON as compared with the wild type ( Fig 5B ) and both mutants showed increased sensitivity to the myosin I inhibitor phenamacril ( Fig 5C ) . Taken together , these results strongly indicated that the myosin I-actin cytoskeleton is essential for the toxisome formation in F . graminearum . To test whether or not the myosin I-actin cytoskeleton is also necessary for biosynthesis of other secondary metabolites ( SM ) , we examined aurofusarin biosynthesis because aurofusarin is a red polyketide pigment and easily visualized . As shown in Fig 6A , the FgMyo1 point mutation ( FgMyo1E420K ) and FgMYO1 knockdown mutants had similar red pigmentation in comparisons with the wild type PH-1 , as well as ΔTri1 and ΔTri4 mutants after incubation for 3 days on PDA or 5 days in liquid PDB . Consistent with these observations , phenamacril and latrunculin A did not inhibit aurofusarin biosynthesis in the wild type ( Fig 6A ) . As controls , deletion mutants of aurofusarin biosynthesis genes AurJ and AurF did not produce the red pigment ( Fig 6A ) . These results suggest that the myosinI-actin cytoskeleton is dispensable for aurofusarin pigmentation . To further confirm this finding , the aurofusarin biosynthesis gene AurJ was tagged with RFP and transformed into the wild type bearing Tri1-GFP or the peroxisomal structural protein FgPex3-GFP . As indicated in Fig 6B ( left panel ) , AurJ-RFP was mainly located in the cytoplasm and presented in a punctuate pattern that was different from the Tri1-GFP localization . However , AurJ-RFP was clearly co-localized with FgPex3-GFP . These results indicate that aurofusarin might be synthesized in peroxisomes . In addition , the cellular localization and fluorescent intensity of AurJ-RFP was not discernibly affected by treatment with phenamacril or latrunculin A ( Fig 6C ) . In summary , the myosinI-actin cytoskeleton is not involved in aurofusarin pigmentation in F . graminearum .
Trichothecenes are synthesized from acetyl-CoA as the basic precursor though the isoprenoid intermediate farnesyl pyrophosphate ( FPP ) and ultimately the trichothecene biosynthesis pathway [38] . The enzymes Tri1 and Tri4 are delivered to the specific cellular compartment known as the toxisome under the DON induction condition ( Fig 1A , S2 Fig ) . This process is largely dependent on various environmental factors or stimuli , including nitrogen and carbon sources [10 , 11 , 39] , amines [40] , pH [41] , light [42] , and reactive oxygen species ( ROS ) [3] . Accumulating evidence indicates that some fungicides also stimulate DON biosynthesis . Milus and Parsons reported that propiconazole and tebuconazole treatments could result in a 50% increase in DON contamination in field trials [9] . Application of fluquinconazole or azoxystrobin reduced disease incidence on wheat spikes but led to a significant increase in DON production by F . culmorum or F . graminearum in the harvested grains [10] . The fungicides epoxyconazole and propiconazole could also stimulate DON production in vitro and in wheat grains [11] . Therefore , the effects for disease management by application of fungicides may not be consistent with the impacts on mycotoxin biosynthesis . In this study , we tested the effect of 131 antifungal compounds on DON biosynthesis and found that phenamacril showed significant inhibition against DON biosynthesis . In agreement with previous studies , other fungicides including the carbendazim and azoles at sub-lethal concentrations could stimulate DON biosynthesis . Therefore , the chemical fungicides for FHB management should be carefully considered to avoid stimulating mycotoxin biosynthesis . In eukaryotic cells , myosins participate in a wide variety of cellular processes , including cytokinesis , organellar transport , cell polarization , transcriptional regulation , intracellular transport , and signal transduction [43 , 44] . They bind to the filamentous actin or other binding partners , and produce physical forces by hydrolyzing ATP , therefore converting chemical energy into mechanical force [12–14 , 44 , 45] . The conserved head domain is accompanied by a broad diversity of N-terminal or C-terminal domains that bind to different molecular cargos , providing the functional specificity of myosin proteins [46] . A total of 31 defined myosin classes have been identified in eukaryotes based on genomic surveys and phylogenetic analyses [15 , 46] . Three myosins: an essential class II myosin FgMyo2 ( FGSG_08719 ) , a class V myosin FgMyo2B ( FGSG_07469 ) , and the essential class I myosin FgMyo1 ( FGSG_01410 ) are recognized in F . graminearum [47] . FgMyo2 is specifically localized to the delimiting septum of phialides and conidia , and required for septation [48] . In addition , the expression levels of TRI5 and TRI6 were obviously higher in the FgMyo2B heterokaryotic disruption mutant than those in the wild type [16 , 49] . These studies indicated that FgMyo2 and FgMyo2B may not be involved in mycotoxin biosynthesis directly . In the current study , we found that FgMyo1 is necessary for toxisome formation . Moreover , we further proved that FgMyo1 was not essential for the biosynthesis of the polyketide secondary metabolite , aurofusarin . These data suggest that the myosin I , but not other myosin motors , participates in DON biosynthesis in F . graminearum . The cellular compartmentalization ( toxisome ) for DON biosynthesis in F . graminearum was first described though the dynamic localization of fluorescent labeled Tri1 and Tri4 [23] . More recently , the toxisome was further identified as reorganization of the endoplasmic reticulum with pronounced expansion at perinuclear-and peripheral positions [22] . Consistent with that , results of the current study further confirmed that Tri1 and Tri4 are often localized in the perinuclear ER under the toxin inducing condition ( Fig 1B ) , and that the ER was remodeled from thin reticulate ER ( S8 Fig ) in the toxin non-inducing conditions to thickened ER in the TBI conditions ( Fig 1A ) . In addition , the ER remodeling is further supported by accumulation of the perinuclear ribosomes under the TBI conditions ( S7 Fig ) since ribosomes are often attached the rough ER . The toxisome structures were predicted to confer multiple beneficial biological functions including clustering of DON biosynthetic enzymes , promoting the efficiency of DON biosynthesis , as well as serving as a self-protection system against the self-toxicity of the Tri products and reaction intermediates [9 , 22] . To date , four proteins including Tri1 , Tri4 , Tri14 and Hmr1 were validated to be localized to toxisomes [16 , 22 , 23] . However , the molecular mechanism for the ER remodeling to toxisome remains unknown . In eukaryotic cells , structures and functions of ER are dynamically changed by various intercellular and extracellular stimuli . For example , the ER network of Arabidopsis undergoes extensive remodeling , which is critically depended on a myosin-actin cytoskeleton system [50] . The plant specific myosin XI provides the force to propel ER streaming and the dynamic rearrangement of the ER network depends on the propelling action of myosin-XI over actin coupled with a SYP73-mediated bridging [51] . Since F . graminearum doesn’t contain a myosin XI homologous protein , we infer that the FgMyo1-actin cytoskeleton may be involved in the ER remodeling for toxisome formation in F . graminearum . This inference is supported by multiple lines of evidence . First , FgMyo1 is comprised of the motor domain that binds to and interacts with actin [12 , 18] , an isoleucine and glutamine ( IQ ) motif , and a C-terminal tail . The tail domain contains a pleckstrin homology ( PH ) motif that is known to bind the anionic phospholipids in cellular membranes ( S9 Fig ) [52 , 53] . The presence of a lipid-binding domain in the tail and an actin binding region in the motor domain equips the myosin I for cellular roles that link membranes to the actin cytoskeleton [54] . Second , dysfunction of FgMyo1 and actin by inhibitors disrupts the toxisome formation ( Figs 1D and 4C ) . Third , knockdown expression of FgMyo1 or the deletion of actin cytoskeleton organization related genes FgPrk1 and FgEnd3 resulted in a defect in toxisome formation and a reduction in DON production ( Figs 2F , 5A and 5B ) . Finally , the point mutation FgMyo1E420K allowing only 5% of the wild-type ATPase activity also affected toxisome formation ( Fig 2F ) , which is in agreement with the interpretation that the hydrolysis of ATP in FgMyo1 coverts the chemical energy into mechanical force and might provide the physical forces for ER remodeling . In addition to providing the force for membrane dynamics , the myosin I motors have also been suggested to function as anchors or tethers between membranes and other proteins . In opossum kidney epithelial cells , Myo1b was found to tether amino acid transporters to the apical plasma membrane , thereby facilitating neutral amino acid transport across the membrane [55] . Similarly , Myo1a is important for the retention and localization of sucrose isomaltase in the intestinal brush border membrane [56] . Furthermore , the spatial association of nuclear myosin I with the ribosome protein S6 plays an important role in the export of small ribosomal subunits through the nuclear pores [57] . In current study , we found that FgMyo1 interacts with the ribosome-associated protein Asc1 , thereby facilitating translation of toxin biosynthesis enzymes , and further contributing to toxisome formation in the toxin inducing conditions . In eukaryotic cells , the myosin-actin system also plays important roles in endocytosis [58–60] . Consistent with that , deletion mutants of actin cytoskeleton organizing gene orthologs , Prk1 and End3 resulted in the defects in both endocytosis and toxisome formation in F . graminearum . However , the mutants of two conserved endocytic components ( Apm4 and Abp1 ) still formed typical toxisomes in TBI ( S10C Fig ) . Importantly , the FgMyo1E420K mutant that exhibits the defect in toxisome formation ( Fig 2F ) retains the capability of endocytosis ( S10A Fig ) , while the actin-activated ATPase activity of FgMyo1E420K is very low ( circa 5% as that of the wild-type FgMyo1 ) [12] . This finding is similar to a previous report that the Myo1 mutants of Aspergillus nidulans with no more than 1% of the actin-activated ATPase activity of wild-type Myo1 in vitro and no detectable in vitro motility activity can support fungal cell growth , albeit with a delay in germination time and a reduction in hyphal elongation [61] . Therefore , the myosin I mediated endocytosis process is not connected with the toxisome formation in F . graminearum . The myosin-actin system also involves in the movement of organelles within cells , including the organelles for secondary metabolites organization . For instance , the short transportation of melanosomes for the skin pigment melanin biosynthesis at the peripheral region of the mammalian cell is largely dependent on the Rab27a , melanophilin , myosinV-actin filament complex [62] . In Fusarium spp . Tri12 is suggested to play a role in export of trichothecene mycotoxins , which forms vacuoles and vesicles during the mycotoxin inducing condition [20 , 21] . A previous study suggested that Tri12 interacted with toxisomes and may transfer the trichothecenes from toxisomes into the vesicles and vacuoles for further export [23] . The motility of vesicles containing Tri12 was reversibly inhibited by latrunculin A , indicating that movement was dependent upon the filamentous actin [21 , 23] . The motor proteins are needed for the cellular motility of Tri12 by mechanical driving force on the filamentous actin . There are three major super-families of motor proteins: kinesins , dyneins , and myosins . The first two act as motors on microtubule filaments , while myosins function on actin [63] . Thus , it would be interesting to further study the functions of myosins in the transport of toxins that may accumulate in Tri12-linked vacuoles and vesicles in F . graminearum and in other toxigenic fungi . Taken together , our data support a model in which FgMyo1 is essential for toxisome formation under the DON induction conditions in F . graminearum by interacting with FgAsc1 indirectly for regulating the Tri protein biosynthesis and by directly participating in the endoplasmic reticulum ( ER ) remodeling via the myosin-actin cytoskeleton system . In addition , the small molecule phenamacril is able to suppress the toxisome formation by inhibiting the ATPase activity of FgMyo1 ( Fig 7 ) .
The F . graminearum wild-type strain PH-1 ( NRRL 31084 ) was used as a parental strain . The wild-type strain and transformants generated in this study were grown on potato dextrose agar ( PDA ) and minimal medium ( MM ) for hyphal examination . The carboxymethyl cellulose ( CMC ) liquid medium was used for conidiation assays [64] . For toxisome observation and trichothecene production analysis , each strain was grown in liquid trichothecene biosynthesis inducing ( TBI ) medium [38] at 28 °C in a shaker ( 150 rpm ) in the dark . Each experiment was repeated three times . The strains ΔFgPrk1 , ΔFgEnd3 , ΔFgTri1 , ΔFgTri4 , ΔFgAsc1 , ΔFgAurJ and ΔFgAurF were constructed using the protocol described previously [65] . Briefly , the open reading frame ( ORF ) of each gene was replaced with hygromycin resistance cassette ( HPH ) and subsequent deletion mutants were identified by PCR assays with relevant primers ( S2 Table ) . For complementation , each ORF fused with a tag and geneticin resistance gene was introduced into corresponding mutant , and transformants were selected with geneticin . To construct FgMyo1 silenced mutants , a 540 bp fragment was amplified and inserted forward and reverse into the pSilent-1 plasmid , and the recombination hairpin RNA silencing plasmid was introduced into PH-1 as previous described [25] . To replace the FgMYO1 promoter with Pzear , the HPH and Pzear fragments were amplified respectively and fused by overlap PCR . Subsequently , the “HPH-Pzear” fragment was further fused with the 5′ and 3′ flanking regions of the FgMYO1 gene . The resulting fusion fragment was purified and transformed into PH-1 . To induce the Pzear replacement , the inducer β-estradiol at 30 μM was added to the medium during the regeneration and mutant selection processes [66] . To construct the FgTri1-GFP fusion cassette , the FgTri1 fragment containing the native promoter and ORF ( without stop codon ) was amplified with primers A15 + A16 ( S1 Table ) . The resulting PCR products were co-transformed with Xho1-digested pYF11 into XK1-25 . The alkali-cation yeast transformation kit ( MP Biomedicals , Solon , USA ) was used to generate the recombined FgTri1-GFP fusion vector . Subsequently , the FgTri1-GFP fusion vector was recovered from the yeast transformant by using the yeast plasmid extract kit ( Solarbio , Beijing , China ) and then transferred into E . coli strain DH5α for amplification . Using the same strategy , other GFP or RFP fusion cassettes were also constructed . Each recombination plasmid was transformed into PH-1 or the corresponding mutant for generating fluorescent label strains . The strain expressing the FgTri1-GFP in the ΔTri1 background was used as the fluorescent reporter strain for anti-toxisome formation screening . The TBI medium supplemented with 104 conidia/mL was added into a 24-well plate ( 2 . 0 mL/well ) . After 24 h static incubation at 28 °C , each tested compound was added into a well and the plate was incubated for another 48 h . Then , the fluorescent intensity in each well was scanned with the Varioskan Flash Multimode Reader ( Thermo Scientific , MA , USA ) for first round screening . The wells with lower or no fluorescent signals compared with that of the control treatment ( the same volume of solvent dimethyl sulfoxide , DMSO ) were further observed by a confocal microscopy . A total of 131 antifungal compounds including 11 commercialized fungicides were tested for the activity against toxisome formation . For each compound , there were three-well replicates , and the experiment was repeated three times . The fluorescent intensity and localization of tagged proteins were observed with a Zeiss LSM780 confocal microscopy ( Gottingen , Niedersachsen , Germany ) . For observation of toxisome formation patterns in PH-1 and derived mutants , each strain labeled with FgTri1-GFP was cultured in TBI for 48 h before examination . All samples were mounted on glass slides and sealed with cover glasses . The following parameter sets of the confocal microscopy were used: Plan-Neofluar 40x/1 . 30 Oil DIC M27 objective; laser: at 488 nm at 50% power for green fluorescence; dimension of X = 70 . 78 μm , Y = 70 . 78 μm; pinhole: 90 μm; digital gain: 1 . 00 . To observe toxisomes in planta , fresh mycelial plugs of the fluorescent reporter strain were inoculated on the leaves of wheat seedlings of a susceptible cultivar Jimai 22 . After incubation at 25°C and 100% RH ( relative humidity ) for 5 days , the infected leaves were taken for toxisome examination observed under Plan-Neofluar 20x/0 . 50 M27 objective . The following filter sets were used for other fluorescent or dye staining: the laser excitation wavelength was set at 405 nm for DAPI ( blue fluorescence ) , at 561 nm for FM4-64 or RFP/mCherry ( red fluorescence ) , at 514 nm for YFP ( yellow fluorescence ) . The endoplasmic reticulum ( ER ) was stained with ER-Tracker Red ( Beyotime technology Co . , Ltd ) , and laser was set at 587 nm for red fluorescence . The intensity of fluorescence was acquired using the Zeiss ZEN 2010 software . For BiFC assays , the final plasmid constructs of pYFPN-FgTri1 and pFgMyo1-YFPC were verified by sequencing and then co-transformed into the protoplasts of PH-1 in pairs . Transformants resistant to both hygromycin and neomycin were isolated and confirmed by PCR . The recombination plasmid pYFPN-FgTri1 or pFgMyo1-YFPC was transformed into PH-1 , and resulted transformants were used as negative controls . YFP signals in the mycelia grown in TBI for 48 h were examined under a Zeiss LSM780 confocal microscope ( Gottingen , Niedersachsen , Germany ) . The protein isolation was performed as described previously [67] . The resulting proteins were separated by 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis ( SDS-PAGE ) and transferred to Immobilon-P transfer membrane ( Millipore , Billerica , MA , USA ) . The polyclonal anti-Flag A9044 ( Sigma , St . Louis , MO ) and monoclonal anti-GFP ab32146 ( Abcam , Cambridge , UK ) antibodies were used at a 1:5000 to 1:10 000 dilution for immunoblot analyses . The samples were also detected with monoclonal anti-GAPDH antibody EM1101 ( Hangzhou HuaAn Biotechnology Co . , Ltd . ) as a reference . The intensity of immunoblot bands were quantified using the ImageQuantTL software . To quantify the mycotoxin production , each strain was grown in TBI medium or inoculated on wheat kernels . DON was extracted , and then purified , and quantified using the LC-MS/MS system as described previously [5 , 68] . The bait protein FgMyo1 was dual labeled with ZZ tag and 3×Flag at its N-terminus and C-terminus , respectively . The resulting fusion cassette was transferred into PH-1 . The resulting transformant ( PH-1::ZZ-FgMyo1-3×Flag ) was used for protein extraction as previous described previously [65] and the affinity capture was conducted by the following procedures . After protein extraction , supernatant ( 25 ml ) was transferred into a sterilized tube . The first run affinity capture was conducted using rabbit IgG agarose beads ( Haoran Biotech Co . , Shanghai , China ) , which was immuno-interacted with the ZZ tag . A total of 500 μl IgG agarose beads were added into the above supernatant to capture ZZ-FgMyo1-3×Flag interacting proteins , following the manufacturer’s instructions ( General Electric Company , GA , USA ) . Then , the washed beads were subjected for the second run capture with anti-Flag agarose beads according to the manufacturer’s instructions ( Abmart , NJ , USA ) . The final ZZ-FgMyo1-3×Flag interacting proteins captured by the anti-Flag agarose beads were eluted with TBS supplemented with 10% SDS . In addition , the ZZ-FgTri1-3×Flag was constructed and the interacting proteins were captured using the same strategy . The captured proteins were digested with trypsin and further analyzed by mass spectrometry using a previous published protocol [69] . Enrichment for proteins assigned to particular functional categories ( FunCat ) was calculated as described previously [20 , 28] . The GFP , RFP , 3× Flag , or mCherry-fusion constructs were verified by DNA sequencing and transformed in pairs into PH-1 . Transformants expressing pairs of fusion constructs were confirmed by western blot analysis . In addition , the transformants expressing a single fusion construct were used as references . For Co-IP assays , total proteins were extracted and incubated with the anti-GFP ( ChromoTek , Martinsried , Germany ) or anti-Flag ( Abmart , Shanghai , China ) agarose as described above . Proteins eluted from agarose were analyzed by western blot detection with a polyclonal anti-Flag A9044 ( Sigma , St . Louis , MO ) , or an anit-GFP antibody ( Abcam , Cambridge , UK ) . The protein samples were also detected with monoclonal anti-GAPDH antibody EM1101 ( Hangzhou Huaan Biotechnology Co . , Ltd . ) as a reference . Each experiment was repeated twice .
|
The mycotoxin deoxynivalenol ( DON ) is the most frequently detected secondary metabolite produced by Fusarium graminearum and other Fusarium spp . To date , relatively few studies have addressed how mycotoxin biosynthesis occurs in fungal cells . Here we found that myosin I governs translation of DON biosynthetic enzyme Tri1 via interacting with the ribosome-associated protein FgAsc1 . Moreover , the key DON biosynthetic enzymes Tri1 and Tri4 are mainly localized to the toxisomes derived from endoplasmic reticulum under toxin inducing conditions . We further found that the FgMyo1-actin cytoskeleton was involved in toxisome formation but not for the biosynthesis of another secondary metabolite tested . Taken together , these results indicate for the first time that myosin I plays critical roles in mycotoxin biosynthesis .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"methods"
] |
[
"medicine",
"and",
"health",
"sciences",
"toxins",
"pathology",
"and",
"laboratory",
"medicine",
"endoplasmic",
"reticulum",
"cell",
"processes",
"fungal",
"structure",
"toxic",
"agents",
"toxicology",
"molecular",
"motors",
"actin",
"motors",
"cellular",
"structures",
"and",
"organelles",
"motor",
"proteins",
"cytoskeleton",
"contractile",
"proteins",
"actins",
"mycology",
"proteins",
"biochemistry",
"cytoskeletal",
"proteins",
"cell",
"biology",
"secretory",
"pathway",
"myosins",
"biology",
"and",
"life",
"sciences",
"biosynthesis"
] |
2018
|
The fungal myosin I is essential for Fusarium toxisome formation
|
Substantial evidence suggests that the phasic activity of dopamine neurons represents reinforcement learning’s temporal difference prediction error . However , recent reports of ramp-like increases in dopamine concentration in the striatum when animals are about to act , or are about to reach rewards , appear to pose a challenge to established thinking . This is because the implied activity is persistently predictable by preceding stimuli , and so cannot arise as this sort of prediction error . Here , we explore three possible accounts of such ramping signals: ( a ) the resolution of uncertainty about the timing of action; ( b ) the direct influence of dopamine over mechanisms associated with making choices; and ( c ) a new model of discounted vigour . Collectively , these suggest that dopamine ramps may be explained , with only minor disturbance , by standard theoretical ideas , though urgent questions remain regarding their proximal cause . We suggest experimental approaches to disentangling which of the proposed mechanisms are responsible for dopamine ramps .
In the main class of TD models of the phasic dopamine response [1 , 2 , 16] , the computational goal of learning is to predict from each state s the expected discounted sum V ( s ) of the rewards that will be encountered during a trial V ( s ) = E { γ 0 r t + γ 1 r t + 1 + γ 2 r t + 2 + … | s t = s } , ( 1 ) where rt is the reward delivered at time t , and 0 ≤ γ ≤ 1 is a discount factor that controls how much weight is given to future relative to immediate rewards . Crucially , the definition of this state value function satisfies a ( Bellman ) consistency condition with respect to each possible next state s′: V ( s ) = E { r t + γ V ( s ′ ) } . ( 2 ) This leads to the idea of using local discrepancies in the value of sampled successive states to drive learning [3 , 4 , 17 , 18] . Thus , the TD error δt is defined as δ t = r t + γ V ( s t + 1 ) - V ( s t ) , ( 3 ) and can be used to improve estimates of V ( s ) . It is exactly this TD error that phasic dopaminergic activity has been hypothesized to represent . As noted , in this paper , we consider data on dopamine concentrations in target structures ( denoted [DA] ) rather than the phasic activity of dopaminergic neurons . These quantities are known to be related [19]; we assume this relationship is simple—a ‘dopamine response function’ ( DRF ) based qualitatively on the signal evoked in NAc by VTA stimulation ( Fig 2 ) . We model the DRF using an alpha function f ( t ) = t ξ e 1 - t ξ , ( 4 ) with time constant ξ = 0 . 7s set to match experimental observations [10] . In other words , dopaminergic activity at time t , which we tendentiously denote δ t p— a phasic TD error —causes an increase in dopamine concentration that peaks after a delay of ξ seconds and then decays with time constant ξ . Thus , changes in dopamine concentration levels relative to baseline , Δ[DA] , are acquired by convolving time-varying activity δ t p with the DRF described in Eq ( 4 ) : Δ [ DA ] ∝ δ p * f ≡ ∫ - ∞ + ∞ d s δ p ( s ) f ( t - s ) . ( 5 ) We should note two important caveats to this model . First , there is evidence for richer temporal and non-linear structure in the DRF [20] , albeit perhaps most affecting timescales and strengths of responding that are different from those considered here . Of more immediate note is that while there is evidence that fluctuations in dopamine concentration within NAc symmetrically encode positive and negative prediction errors [21] , other studies do not show such clear negative deviations from baseline corresponding to a negative prediction error ( e . g . [22] ) . Indeed , evidence suggests that negative prediction errors are represented differently from positive prediction errors in the activity of midbrain dopaminergic neurons: while positive prediction errors appear to correlate positively with the firing rates of dopaminergic neurons , the magnitude of negative prediction errors correlates rather with the duration of a pause in burst firing [23 , 24] , though this itself generates additional complexities . To incorporate the possibility of an asymmetry in how positive and negative prediction errors affect dopamine concentration , below we also examine the effect on dopamine concentration of first asymmetrically scaling negative prediction errors by a factor of d = 1/6 [25] . The second caveat is that modulation of striatal dopamine concentrations can occur independently of changes in the observed firing rates of dopaminergic cells . Thus , tonic levels of striatal dopamine are thought to be controlled by the number of active dopaminergic cells rather than by the firing rates of a fixed pool of neurons [26] . Furthermore , a range of mechanisms local to the striatum are known to play a role in regulating dopamine release , including a host of other neurotransmitters such as glutamate , acetylcholine , and GABA ( for recent reviews , see [27 , 28] ) . In a case more general than that of learning purely to predict , animals may be allowed to select actions to achieve desired outcomes . A mapping from states to actions is usually referred to as a policy , denoted π , and the more general problem is to find a policy which maximizes some measure of reward . The TD error signal defined in Eq ( 3 ) can be used to evaluate state values with respect to a given policy , Vπ ( s ) . Given this value function , the agent can potentially improve on its current policy by selecting actions that lead to successor states of higher value . Iteration between successive steps of policy evaluation and policy improvement characterises the policy iteration algorithm [29 , 30] which is a cornerstone of RL methods [4] . The actor-critic algorithm [31] , an asynchronous version of policy iteration , is just one of a number of TD-based suggestions for RL [4] . However , it has played a particularly salient role in neural RL modelling [16 , 32–34] . In the actor-critic architecture , state values and policy are explicitly represented in different memory structures . The policy structure is known as the actor , since it is responsible for selecting actions; and the value structure is known as the critic , since it criticizes actions taken by the actor , where this critique takes the form of the TD error described above . In terms of neural substrate , it has been suggested that the dual learning functions of the actor-critic map to a fundamental division in the functional anatomy of striatum into dorsal and ventral subregions [1 , 33 , 35 , 36] . In particular , the ventral striatum ( NAc ) is implicated in reward and motivation [37] , while the dorsal striatum is implicated in motor and cognitive control [38] . This dissociation is consistent with an implementation of actor and critic components in the dorsal and ventral striatum , respectively [1 , 36] . Initial theorizing in neural RL focused on tasks involving a simple action or choice between different discrete actions in response to an explicit experimental cue . More recent modelling work has sought to extend standard RL models to other dimensions of choice , thereby making contact with the large experimental literature on free operant tasks in which subjects not only choose between different actions but also when and how quickly to act [39–41] . Two key differences from previous work have been involved in the first collection of models of free operant tasks . Firstly , the agent not only chooses an action a to perform , but also an associated latency τ with which to perform it . Formally , this entails moving from the usual discrete Markov decision process ( MDP ) model , in which agent-environment interactions progress at fixed time intervals , to a semi-Markov decision process ( SMDP ) [42] , which permits the time spent in a particular state to follow an arbitrary probability distribution . Secondly , rather than assuming that the agent aims—at least approximately—to maximize an expected sum of discounted future rewards , models have assumed an average reward criterion . In this case , the aim is to find a policy that maximizes the long-run average reward rate ρ π ≡ lim n → ∞ E π 1 n ∑ t = 0 n r t , ( 6 ) which is independent of starting state , assuming ergodicity . The value of a state under policy π is now defined relative to the long-run average reward under that policy , ρπ , and can be denoted V ˜ π ( s ) to highlight that this is a relative value [4]: V ˜ π ( s ) = E π ∑ k = 0 ∞ ( r t + k - ρ π ) | s t = s . ( 7 ) Similarly , the relative action value Q ˜ π ( s , a ) of taking action a in state s is defined as Q ˜ π ( s , a ) = E π ∑ k = 0 ∞ ( r t + k - ρ π ) | s t = s , a t = a . ( 8 ) For example , consider the case in which there is just a single action—a lever press—to perform , and the decision concerns the latency τ with which to perform it . For consistency with earlier results , we temporarily consider the case of continuous time . Assume that τ is selected , following presentation of an explicit cue , in an initial state ‘1’ . After the selected time τ , there is a transition to a second state ‘2’ in which the lever press completes and reward is delivered . Subsequent transition back to state 1 follows immediately , and the process begins anew ( Fig 3A ) . Niv et al . [39] considered a hyperbolic cost structure in which a lever press of latency τ is more costly depending on its speed . In particular , they adopted the function form for the cost: a/τ + b , where b ≤ 0 is a unit cost for the press , and a ≤ 0 is a factor which determines the magnitude of hyperbolic dependence on τ . Each lever press is assumed to yield an immediate reward of utility r > 0 . As shown by Niv et al . [39] , the theory of average reward RL tells us to select the optimal lever-press latency τ* in state 1 that maximizes the optimal relative Q-value , τ * = argmax τ Q ˜ * ( 1 , τ ) = argmax τ a τ + b + r - ρ * τ + V ˜ * ( 2 ) , ( 9 ) where asterisks are used to indicate values corresponding to an optimal policy . As noted in [39] , the optimal latency here is controlled by the opposing forces of the ( negative ) utility of acting quickly , a/τ , and the opportunity of cost of acting slowly , −ρ*τ . This latter term arises from Eq ( 8 ) since ρ* ( which is ρπ when executing the optimal policy ) is accumulated over all the timesteps comprising latency τ . Indeed we have τ * = - a ρ * , ( 10 ) which shows that the optimal latency decreases as the average utility rate ρ* increases . Since ρ* also depends on τ* , the problem is recursive , but techniques for finding the optimal solution exist [42 , 43] . The connection to current concerns is the proposal that the tonic level of dopamine , especially in NAc , represents the long-run average rate of reward ρπ , effectively signalling an opportunity cost of sloth [39 , 44] . This suggestion is based on a long literature implicating dopamine in the modulation of behavioural vigour [45] . It has been further supported by recent human studies [46 , 47] , albeit assuming that this long-run average rate arises as a slowly-changing running estimate . The equivalent of the dopamine response function for this signal is unexplored . Dayan [41] has recently broadened the theoretical study of instrumental vigour to include the case of acting to avoid punishment . A first example of the phenomena of interest comes from an experiment by Roitman et al . [10] very similar in structure to the lever pressing case considered above . Following presentation of an explicit cue , a rat could press a lever at a time of its own choosing to receive a sucrose reward ( Fig 4A ) . Cue presentation evoked an increase in dopamine concentration in NAc ( Fig 4B , upper trace ) , but not in control animals for which a lever press did not yield reward ( Fig 4B , lower trace ) . The apparent decrease in signal in the latter case was found not to be caused by a change in dopamine concentration [10] . However , Roitman et al . also observed that , when aligned to the time of lever pressing , average dopamine concentration began to increase a short time before the time of the lever press itself , reaching peak concentration around the time of pressing ( Fig 4C and 4D ) . Crucially , this occurred not only on the majority of the trials ( 83% ) in which animals pressed the lever at relatively short latencies following the initial cue ( <5 s; Fig 4C ) , but also on the smaller number of trials in which animals responded at longer latencies ( >5 s; Fig 4D ) . Similar increases in extracellular dopamine just prior to response have been reported in other FSCV studies [8 , 9 , 11–13] . Roitman et al . also reported that while cue-aligned and press-aligned peak dopamine concentrations were indistinguishable for short-latency trials ( 68±19 nM vs . 73±23 nM ) , press-aligned peak dopamine was significantly larger than cue-aligned peak dopamine on long-latency trials ( 54±17 nM vs . 110±20 nM; Fig 4D , inset ) . A second , perhaps more dramatic , example of dopamine ramping has recently been reported by Howe et al . [14] ( Fig 5 ) . In this study , dopamine concentrations in the striatum were measured using FSCV while rats navigated mazes to obtain remote rewards . It found a gradual increase in dopamine concentration that began at trial onset and ended after reaching the goal ( Fig 5A ) . Whether rats took a relatively short or long time to reach the goal , dopamine peaked at similar concentrations at the goal ( Fig 5B , upper ) . Similarly , dopamine peaked at comparable concentrations at the goal for mazes of different length ( Fig 5B , lower ) . Single-trial examples in which rats paused mid-run showed a remarkable correspondence between proximity to the goal and dopamine concentration ( Fig 5C ) . Furthermore , dopamine ramps scaled with size of reward , so that peak dopamine was higher for larger than smaller rewards ( see [14] , figure 3 ) . While we take both of these examples to be instances of dopamine ramping , their explanations may not be identical . Nevertheless , neither case seems to fit neatly with standard RL models because apparently reliable activity is not predicted away by earlier reliable cues .
Whether an animal faces a task in which it is free to respond as often and as quickly as it likes , or is limited to a single response within an interval following a cue , it typically has at least some freedom to choose its time of response . In the case of Roitman et al . [10] described above ( Fig 4A ) , rats were free to lever press at a time of their own choosing following a cue marking the start of a new trial . As reported in a number of similar studies , ramp-like increases in NAc dopamine concentration which preceded the time of lever-pressing were observed ( Fig 4C and 4D ) . From a conventional TD perspective , phasic dopaminergic activity reflects a prediction error . Such errors can be occasioned by changes in latent states associated with the subject’s internal execution of the task , provided that there is some uncertainty associated with these changes . Such uncertainty can be generated by two forms of ignorance: what the critic fails to know about the actor’s choice of when to act , and what both actor and critic fail to know about the passage of time [48 , 49] . Consider first the critic’s knowledge about the temporal decisions of the actor ( Fig 6 ) . We assume , reminiscent of studies by Libet and colleagues [50] , and consistent with both patterns of cortico-striatal connectivity [51–53] and observed patterns of discharge [54 , 55] , that internal information proximal to the action , such as some form of motor preparation , is communicated to the critic via efference copy just before it is evident to the experimenter ( a′′ in Fig 6 ) . This resolves any uncertainty the critic may have about the time of the impending action . The question is what happens at the time that the actor makes its decision about the latency of lever pressing following the initial cue . There are two natural possibilities . One is that the actor also intimates its decision about when to act directly to the critic at that time , e . g . , via a more indirect form of efference copy ( a′ in Fig 6 ) which could be transmitted via interacting cortico-striatal loops or some more direct means [56] . This would then influence the critic’s predictions about future events . The other is that the critic has no such privileged access to the actor’s initial decision , implying that its predictions could be based only on its experience of downstream signals resulting from the actor’s choices . A second , related issue concerns the realization of timing . If the actor communicates its choice to the critic and the two share the same clock , then there seems to be little room for timing uncertainty to affect the critic’s predictions: regardless of whether the clock is fast , slow , or variable , actor and critic will be in synchrony . On the other hand , if the actor does not specify an exact time of action , or its decisions are subject to additional sources of what the critic will experience as uncontrolled variability ( for instance if actor and critic employ different clocks ) , timing uncertainty may play a role in the critic’s predictions and resulting prediction errors [48 , 49] . To explore these issues , we consider the same lever-pressing task described previously ( Fig 3A ) , though with a state space that is augmented to reflect the assumption that the critic may receive internal information about the lever press just before it occurs ( Fig 3B ) . As before , an initial cue ( ‘1’ ) is observed , prompting selection of a latency τ with which to press the lever . After the selected duration τ , which may or may not be known by the critic , the animal transitions to a state of preparedness to press , assumed to be communicated to the critic via efference copy ( ‘2’; this corresponds to the time at which the critic receives signal a′′ in Fig 6 ) . Note that this latter state is distinct from that corresponding to consummation of the lever press itself ( ‘3’ ) which is assumed to occur only after a further interval τpost . We set τpost = 500 ms to correspond roughly with the time with which the so-called ‘readiness potential’ is detected prior to self-initiated action [57 , 58] . A reward of utility r = 1 is delivered on press completion . Completion of the lever press and reward delivery is followed by a fixed inter-trial interval τI = 30 s , after which the process begins anew . Given the models described in the previous section , we consider results from three different cases: two in which the critic only receives information about the lever press indirectly , and one in which the critic additionally receives direct information from the actor . In each case , we consider the effect of the critic receiving notice of impending action at different times—T = {1 , 3 , 10} seconds—on the TD error δ t p , and evaluate the resulting change in dopamine concentration Δ[DA] under both symmetric and asymmetric encoding assumptions . Results for all the cases are summarized together in Fig 8 . Our first possible account of ramping , the TD account of pre-response signals described above , assigns dopamine a passive role in decision-making: increases in dopamine reflect a latent state transition arising from a decision to act which has already been made . However , experimental evidence suggests that accumbens dopamine could also play a more causal role . For example , Phillips et al . [9] found that electrically-evoked dopamine transients in NAc increased the probability that rats would lever press for cocaine immediately afterwards , further commenting that videotaped behavioural records showed that stimulation led to immediate changes in behaviour , notably behavioural sequences up to and including lever approach . Relatedly , Nicola [64] found that blocking dopamine signalling in NAc impaired rats’ ability to approach and press a lever for food , but only when animals were likely to have to re-engage with the task by following a novel sequence of actions to approach the lever . Such findings have led to the suggestion that accumbens dopamine is necessary for ‘flexible approach’ [64] . In fact , models associating phasic dopamine with a TD error signal have long considered a dual role for dopamine in which indirect effects on behaviour , involving learning , are accompanied by direct ones [1 , 65–67] . We next explore a second potential mechanism for ramping signals . In particular , we show that a particular decision-making scheme which couples dopamine directly to the decision process also generates dopamine ramps . A rich vein of work in psychology and neuroscience revolves around the idea that the brain implements some version of the sequential probability ratio test ( SPRT ) , a sometimes optimal procedure for two-alternative forced-choice decisions under uncertainty [68 , 69] . While the SPRT and its close associates are usually considered in relation to decision making under state uncertainty , as when there is doubt about whether the overall motion of a random dot field is to the left or right [70] , such models have also been applied with some success to memory-based [71] or value-based [72] decisions in which sensory information is absent or unambiguous . We consider the possibility that this arises from accumulation of value information , in which information stored in synapses is read out via spike trains in a temporally extended manner . A prominent realization of the SPRT is the so-called drift-diffusion model ( DDM ) which we describe in detail below [68 , 69 , 71 , 73–75] . In particular , this can be shown to be a suitable abstraction of a particular sort of neural circuit involving competition between two ( or sometimes more ) populations of neurons representing the choices [73 , 76 , 77] . One of the earliest computational suggestions for the role of dopamine and other catecholamines was that by influencing the excitability of neurons [78] , they could influence gain control in such circuits , and thereby influence the course of decision-making [79 , 80] . Such models were originally conceived of in terms of cortical decision-making circuits; however , for instance , Frank’s [81] neural network model of the basal ganglia assumes that dopamine controls the relative excitability of direct ( ‘Go’ ) and indirect ( ‘Nogo’ ) pathways via different dopamine receptor subtypes , thereby influencing both the propensity and latency to act . Specifically , higher levels of dopamine shift the balance of activity in favour of the ‘Go’ pathway , leading to a greater propensity to act and faster reaction times . Dopaminergic modulation of excitability in this model can also be interpreted in terms of gain-setting [81] . Here we bring together these two ideas—of an accumulative decision-making process and dopaminergic gain control—to explore how a more direct coupling between dopamine and decision-making may explain ramping dopamine signals in striatum . We now consider a third account of dopamine ramps based on a new model of discounted vigour . Incorporating the observations and suggestions of Howe et al . [14] , together with a partially free-operant experiment of his own , Berke and colleagues ( personal communication , [85] ) suggested that the concentration of dopamine measured by FSCV in the accumbens might be strongly influenced by the discounted value function V ( s ) of Eq ( 1 ) . This will show evidence of ramping towards final goal states when the discount factor is less than 1 , consistent with the observations of Howe et al . ( Fig 5 ) . We describe this signal as being quasi-tonic since , when there is no reward , it is a form of integral of the TD prediction error , which is phasic . However , one should bear in mind that when the state changes abruptly , the value can change abruptly too . The key question , though , is why we should expect to see any such quasi-tonic signal in this context ? We consider the possibility that this signal is the equivalent in the discounted case of the average reward ρ ( for convenience , in this section , we omit the superscript π ) which , as we have seen , has previously been argued to be ( a ) the comparison point for the phasic prediction error or the immediate reward; ( b ) the spur to instrumental vigour; and ( c ) represented by tonic levels of dopamine [39] . Indeed , consider afresh an apparent inconsistency in the definition of the TD prediction error between the cases of average and discounted reward . In the average case , the phasic component of the full prediction error ( c . f . Eq ( 16 ) ) , now denoted δA ( st ) , is δ A ( s t ) = r ( s t ) + V ˜ ( s t + 1 ) - V ˜ ( s t ) , ( 21 ) and we expect the mean of this over the long run to be the overall mean reward rate 〈 δ A ( s t ) 〉 = ρ , ( 22 ) which is a tonic signal that therefore acts as a comparison point for the phasic prediction error . Eqs ( 21 ) and ( 22 ) can also be seen as arising from the observation that the relative values V ˜ are expected undiscounted sums of the differences between r ( st ) and ρ . Unfortunately , even if the relationship in Eq ( 22 ) actually holds , ρ , because it is stationary , is formally hard to measure with FSCV , whose measurements are typically referenced to a potentially ever-changing baseline . By contrast , in the discounted case , the phasic prediction error δγ ( st ) is normally written as δ γ ( s t ) = r ( s t ) + γ V γ ( s t + 1 ) - V γ ( s t ) , ( 23 ) now writing the discounted value function as Vγ ( s ) , and is expected on average to be 0: 〈 δ γ ( s t ) 〉 = 0 . However , two considerations encourage us to write this expression slightly differently , with an undiscounted phasic TD prediction error just as in Eq ( 21 ) : δ A γ ( s t ) = r ( s t ) + V γ ( s t + 1 ) - V γ ( s t ) , ( 24 ) which should , on average , take the value 〈 δ A γ ( s t ) 〉 = ( 1 - γ ) 〈 V γ ( s t + 1 ) 〉 . ( 25 ) Here , ( 1 − γ ) 〈Vγ ( st + 1 ) 〉 , by analogy with the truly stationary signal ρ , would be represented as a quasi-tonic signal which acts as a target for a phasic TD prediction error signal that involves a discounted value function . Assuming that this baseline signal is represented in a quasi-tonic concentration signal would thus licence ramping . The two considerations that encourage this interpretation of phasic and quasi-tonic dopamine signals are: ( i ) continuity between average and discounted cases as γ → 1; ( ii ) something of particular pertinence in the current context , namely the determinants of vigour for discounted problems . We discuss these in turn . There is also a rough analogy with the Hamilton-Jacobi-Bellman ( HJB ) equation [86 , 87] , but as this requires considering continuous space and time , as well as a different sort of transition structure , we do not discuss it further . Insight into discounted vigour comes from numerical calculations of the optimal latencies τ* as a function of γ and for different values a < 0 that control the hyperbolic cost of vigour c ( τ ) = a τ in two cases: a terminating chain with Vγ ( st + τ ) = 1 , ∀τ , γ ( roughly as in [14] ) and a continuing chain as in Fig 3 . Fig 14A shows optimal latencies τ* in the terminating case . Generally , as γ decreases , the faster the weight given to future value Vγ ( st + τ ) decays with time , encouraging quicker latencies . This tendency is balanced by the greater cost of acting quickly that is then incurred . In fact , one can show that there there is a limit on the cost of acting of amin = 4/ ( e2 log γ ) below which there is no solution for τ* — crudely , the cost of acting quickly deems such a long latency that the resulting discounted value of the reward ( from Vγ ( st + τ ) = 1 ) is insufficient to warrant action at all ( Fig 14A , solid red line ) . Fig 14B and 14C show differences in τ* in the continuing compared to the terminating case . In the continuing SMDP , the result of pressing the lever includes a further opportunity to press the lever ( without which , the infinite horizon average reward ρ would formally be 0 ) . When the inter-trial interval τI is large relative to −1/log γ , there is little difference from the terminating case ( Fig 14B ) ; however , when it is not , the prospect of accelerating not only the immediate reward but also future rewards further hastens lever pressing , visible in greater decreases in τ* compared to the terminating case ( Fig 14C ) . Given the preceding analysis , it is straightforward to show that a quasi-tonic dopamine signal reflecting the quantity ( 1 − γ ) Vγ ( st + 1 ) would lead to the sort of ramping observed by Howe et al . [14] in their spatial reward task ( c . f . Fig 5 ) for γ < 1 . Indeed , just as observed by Howe et al . , [DA] gradually ramps up as the goal is approached and peaks at the same value regardless of the time taken to reach the goal or the distance travelled to reach it , assuming a fixed reward size ( Fig 15A ) . Further , as observed experimentally , increasing the reward size leads peak [DA] to increase ( Fig 15B ) and , given a lack of progress towards the goal—for instance if the agent remains stationary or moves away from the goal—[DA] remains approximately stationary or decreases , respectively , as observed by Howe et al . on such trials ( Fig 15C ) . One should note in this latter case that the single-trial examples shown by Howe et al . find dopamine concentrations tracking spatial proximity remarkably closely ( see Fig 5C ) , while convolution of ( 1 − γ ) Vγ ( st + 1 ) with the DRF that we have assumed leads to a signal which looks comparatively over-smoothed ( Fig 15C , right ) . However , given the heterogeneous nature of striatal dopamine release [88] , how rapidly [DA] is observed to change may well depend on the exact positioning of the voltammetric sensor . Examination of further single-trial examples could help clarify this issue .
What TD accounts of pre-response dopamine signals predict depends on the assumptions made about the relationship between actor and critic . We considered three possibilities associated with different predictions of how a TD error occurring just prior to pressing , and the resulting change in dopamine concentration , should change as response latencies increase: remain constant , decrease , or increase . The model in which the critic receives both direct and indirect information , but suffers from timing uncertainty , yielded results most consistent with the experimental data reported by Roitman et al . [10] . In particular , this case replicated the observation that peak dopamine concentration around time of pressing was larger than at time of cue for long latency trials . This result relied on the assumption that the critic’s uncertainty about the time of action increases with choices of longer press latencies . This is consistent with the finding , in the equivalent Pavlovian circumstance , that the responses of dopaminergic neurons to a cue predicting reward delivery after a long delay are smaller than responses to cues predicting shorter delays; conversely , dopamine responses to the reward itself increase with longer delays [62 , 89] , a finding that indeed has been suggested to arise through timing uncertainty . This finding is apparently opposite , though , to an observation also mentioned above . This is that for the case of a single , non-exponential hazard function , which mandates a range of possible times at which a reward-related cue might be presented , relatively late presentations inspire smaller dopamine responses than early ones [63] . An obvious explanation of this finding is that as time goes by , presentation of the cue is more and more likely , and so less and less unexpected . This does not contradict our finding , which depends on many possible hazard functions , one for each choice of lever-press latency . Two assumptions in the proposed TD account merit further comment . Firstly , while we assumed that the actor’s choice of when to press the lever immediately follows cue presentation , one can imagine variability in when the actor makes decisions about when to act . For example , it might be that the animal initially fails to notice the cue , or is otherwise engaged ( even in instrumental leisure; [40] ) when the cue arrives , only later resolving to engage with the lever . Secondly , and perhaps relatedly , while it was convenient to assume that latencies τ , and therefore times T , follow a gamma distribution , the reported distribution of press times appears to have heavier tails than we would expect if they were drawn from a single gamma distribution . Thus , Roitman et al . reported mean response times of 1 . 2 s and 26 . 2 s for short-latency ( 83% ) and long-latency ( 17% ) trials , respectively . Closer examination of the empirical response distribution in such studies would be of interest for future work . A more general problem for a TD account of pre-response signals is that while there is abundant evidence of a systematic temporal relationship between the time at which a cue indicating reward availability is presented and subsequent phasic activity in dopaminergic neurons , there is little or no evidence of such a relationship between the time of the phasic response and when a subsequent instrumental action—necessary to obtain the reward—is emitted . For example , Ljungberg et al . [90] found that when monkeys were exposed to cues that predicted when they could obtain food by reaching into a box , activity of dopaminergic neurons was time-locked to the predictive cue rather than movement onset . Whether this is also true for the timescale and nature of rodent movements is unclear . Even in monkeys , Romo and Schultz [91] reported gradual increases in the firing rates of some putative dopaminergic cells ( 12 out of 104 recorded ) up to 1500 before onset of self-initiated arm movements to obtain food . However , this slow change in activity does not resemble the sort of bursting activity that might be associated with a phasic TD signal . Suggestively , striatal ( and cortical ) neurons in monkeys show various patterns , including ramp-like increases in activity , before self-initiated movements [55 , 92 , 93] . Similarly , some neurons in rat ventral striatum show anticipatory increases in activity when approaching or waiting for food delivery [94 , 95] . Furthermore , simultaneous electrophysiological and FSCV recordings from the same electrode have revealed that changes in dopamine concentration and activity of specific subsets of accumbal cells can be temporally correlated [96 , 97] . Suppression of phasic activity in VTA dopaminergic cells appears to disrupt such time-locked activity , perhaps indicating that it is phasic activity of dopaminergic cells which drives such correlated activity [98] . So at the mechanistic level at least , there are multiple possibilities for the origins of pre-response signals beyond phasic dopaminergic activity: they may reflect the sort of slow change in activity of dopaminergic neurons observed in [91] , or they may reflect increased dopamine release instigated more directly by the activity of other cells , such as reflected in cortico-striatal inputs . A range of previous work has considered the vagaries of the representation and processing of time . We noted that possible sources of uncertainty included partial observability of the actor’s choices in the case where the critic does not have direct access to this information , and possible timing uncertainty in the case where it does . Implications of partial observability for TD models of dopamine have been explored in previous work , notably by Daw and colleagues [48] , though that did not address the possibility of partial observability arising between distinct internal agencies , nor the possible relevance to self-initiated action envisaged here . The same study and a number of others [48 , 49 , 99–101] have addressed the issue of the representation of time , and how this representation may influence timing uncertainty ( see [102] for a recent review ) . The implications for TD ( and indeed different models of discounting ) of the possible distinction between the animal’s ‘internal’ time and the experimenter’s ‘conventional’ time have been worked out in detail by Nakahara and Kaveri [49]; we also considered the possibility of separate internal clocks for actor and critic . Additional complexity , which we leave to future work , arises from the putative connection between dopamine and the speed of an internal clock , as inferred , for example , from the effects of dopamine manipulations on behaviour in interval-timing tasks [102 , 103] . We showed that ramping dopamine signals can be generated by a mechanistic decision-making model in which dopamine sets the gain of value-based accumulation . Furthermore , we saw that this direct coupling of dopamine to decision-making could generate a negative correlation between the size of TD error and decision time , consistent with the experimental observation that a larger phasic response of dopaminergic cells to a start cue is associated with a shorter latency of behavioural response [83] . This route to ramping signals is primarily statistical , arising from trial-averaging . On any individual trial , dopamine ramping towards the time of decision may or may not occur , though it is certainly more typical when dopamine fluctuations incorporate a strong , TD-related phasic component . To the best of our knowledge , whether pre-response transients in NAc reliably precede the animal’s response on individual trials , or may reflect trial-averaging , is unknown . Mathematical analysis of the time-varying gain DDM that we described is given by Moehlis et al . [82] and , indeed , the idea that dopamine could set this gain follows directly from previous work by Cohen and colleagues on catecholaminergic gain control [79 , 80] . For example , Shea-Brown et al . [80] suggested that noradrenergic activity of cells in the locus coeruleus may help to optimize decision-making by adjusting the gain of an integrative decision process . Furthermore , they showed that their model could replicate the experimental finding that phasic responses of the locus coeruleus correlate more closely with time of behavioural response than with time of stimulus onset in a decision-making task [104] . Also of relevance is the biologically-detailed neural network model of the basal ganglia proposed by Frank [81] in which dopamine modulates the balance between direct and indirect pathways . Ratcliff and Frank [105] have recently explored the links between the latter’s neural network model and more abstract diffusion models , though without exploring a possible direct role for dopamine in the latter . Nevertheless , it is interesting to consider that , depending on the form of the DDM used to fit the data , dopaminergic modulation of a temporally-extended decision process may be manifest in different parameters . For example , a positive correlation between increased tonic dopamine levels and faster responding may also be captured by the assumption that dopamine modulates the threshold of a DDM where the gain is fixed [106] , rather than modulating the gain under a fixed threshold . Additionally , one may consider potential effects of dopamine not only on the latency of response , but also on which choice is made , for instance due to asymmetries in how dopamine modulates direct and indirect pathways ( M . J . Frank , personal communication; [107] ) . More generally , it would be of interest to know whether dopamine ramps would also be observed in Frank’s comparatively detailed model of the basal ganglia . We reconciled an apparent inconsistency between the definitions of TD errors in the cases of average and discounted reward via an analysis in which ramp-like signals would be expected to emerge . In particular , we suggested that the quantity ( 1 − γ ) 〈Vγ ( st + 1 ) 〉 in the discounted reward model plays an equivalent role to the average reward rate ρ in the average reward model . Since values often ( though not always ) change modestly as a result of the passage of time , this signal is quasi-tonic , and thus a candidate for what would be recorded using a technique such as FSCV . This signal can explain the ramping phenomena observed by Howe et al . [14] and also those observed in more recent experimental work [85] . We speculate below on its network or biophysical realization . Potentially at odds with our suggestion that the quantity ( 1 − γ ) 〈Vγ ( st + 1 ) 〉 is appropriate for controlling vigour , changes in the running speed of rats in Howe et al . ’s study do not show a close match to the temporal profile of dopamine concentrations . However , one would not necessarily expect a straightforward relationship between these variables , given that the subjects must negotiate environments without crashing into walls . Howe et al . used T- , M- , and S-shaped mazes , whose turns , unsurprisingly , led to decreases in velocity ( see [14] , figure 3h–k ) . Our analysis suggests that ramps are scaled by the discount factor γ , prompting the question of how this discount factor is set , whether it is variable or fixed , and indeed , whether it is unique . There is substantial evidence that human and animal discounting takes a hyperbolic form [108 , 109] rather than being exponential as considered here ( which is rather ubiquitous in engineering and economic settings ) . This can arise from a combination of two or more exponentials , and it would be most interesting to extend our analysis to this case . From a formal viewpoint , the discount factor can be seen as the probability per unit time of task termination or , indeed , as a means of simplifying a problem en route to an ultimate solution [110] . In humans , there is evidence that discount rates can be manipulated experimentally [111] and that individuals can flexibly vary their discount rates to suit task demands [112] . It has also been suggested that some regions , notably the striatum , display a graded map of discount rates which serve reward prediction at different timescales [113 , 114] . Howe et al . observed ramping in dopamine concentration in both ventromedial and dorsolateral striatal areas , though ramping responses were reported to be more common in ventromedial striatum . Hints of steeper ramping are perhaps discernible in the average signals reported in ventromedial as opposed to dorsolateral striatum ( [14] , figure 1 and extended data figures 3a and 4 ) . However , whether such ramping signals display systematic , graded differences across the striatum or otherwise change in response to experimental manipulation of discount factors remains an open question . Whereas the TD account of pre-response transients naturally attributes the observed signal to the phasic activity of dopaminergic neurons [2 , 5–7 , 19] , the sources of tonic and particularly ‘quasi-tonic’ dopamine signals are less clear . One long-standing suggestion is that phasic and tonic modes of firing in dopaminergic cells provide independent control of phasic and tonic dopamine levels within NAc [26 , 115 , 116] . Thus , burst firing of dopaminergic neurons is thought to mediate a fast , high-amplitude dopamine transient which is spatially-restricted to a region within or proximal to release terminals by dopamine reuptake . By contrast , the comparatively slow , irregular , ‘tonic’ mode of activity exhibited by a pool of dopaminergic neurons , potentially of varying size , is thought to control the more stable , tonic levels of extrasynaptic dopamine . If average reward rate is represented in tonic levels of dopamine [39] , then a natural suggestion is that representation of this quantity is controlled by this tonic mode of activity . Where does a quasi-tonic dopamine signal fit into this picture ? It is not clear that the relatively short timescale of change of the ramping signals reported by Howe et al . could arise through mechanisms thought to modulate tonic activity . On the other hand , ramping in the phasic activity of dopaminergic neurons has seldom been reported . Fiorillo et al . [117] reported ramp-like increases in between-trial averaged activity under conditions of uncertain reward delivery , though interpretation of this result has been controversial [25 , 118] . While the paucity of such reports may simply be due to a lack of appropriate electrophysiological recordings in spatial tasks—which may also explain why ramping of dopamine concentrations has not been observed prior to [14]—an interesting alternative is that the gradual increase in dopamine concentration is partially- or fully- independent of the activity of dopaminergic cells [15] . As mentioned above , a number of local regulatory mechanisms are known to gate the probability of dopamine release [27 , 28] , and there is evidence that striatal dopamine release can occur independently of dopamine cell firing [119] . An understanding of how these different mechanisms of dopamine release interrelate is of clear experimental and theoretical interest . It should be noted that although we have referred throughout to dopamine signals in the nucleus accumbens generally , this should not be taken to suggest that dopamine release is homogeneous within this region . Indeed , FSCV measurements suggest substantial spatial heterogeneity [88] . Subregions of NAc have been segregated according to various anatomical features , classically into core and shell subregions [120 , 121] . Pre-response transients have typically been observed in NAc core [8–13] . Much interest centres on the functional significance of this core-shell distinction [37 , 122–125] and , indeed , distinctions at a finer grain [126] , including in relation to possible differences in dopaminergic release [127] . We noted above that ramping ostensibly disrupts TD’s explanation for dopaminergic release , since it would have , oxymoronically , to be a predictable prediction error . Alternative accounts have been suggested according to which prediction errors indeed persist . Gershman [128] considered the consequences of an unsuitable state representation . The idea is that the exponentially discounted value signal Vγ ( s ) cannot be captured in an error-free manner if the state ( i . e . , the position of the animal ) is represented in particular , over-generalizing manners , for instance by units whose activity is governed by the square , rather than linear , distance to the goal . In this case , a ramping prediction error turns out to arise via persistent representational error . Place cells [129–131] provide an accessibility-sensitive representation of space , and the generalization afforded by the coarse-coding they imply is often useful [132] . However , it is also known that Bayesian decoding of even a modest number of such cells leads to surprisingly accurate localization of animals in their environments [133] , and thus what would amount to a table-lookup representation that would not lead to persistent error . Of course , one must remember that this sort of decoding is in silico , rather than in vivo . Morito and Kato [134] have also also suggested that the Howe et al . ramping signal reflects persistent prediction errors . In their proposal , these arise out of the assumption of a time-dependent decay of learned state values . One challenge for this model is that its generation of ramping signals qualitatively similar to that observed experimentally appears to be unstable to changes in reward magnitude [134] , and indeed to the passage of more substantial periods of time . The most pressing consideration is a set of experiments that can test and refine or reject these various mechanisms , and understand how they might work together . Perhaps the most straightforward to test is the last suggestion , since it is unique in its dependence on discounting . Given that the rate of this should be sensitive to things like the reliability of the environment [135] , it would be interesting to manipulate these factors , determine the extent to which behaviour changes appropriately , and concurrently measure ramping . Similarly , it may be that individual differences in discounting , as measured by choices between immediate , smaller rewards and delayed , larger rewards , can be predicted by the rate of ramping . Although behaviour generally follows hyperbolic rather than exponential discounting [109] , this would only make a modest difference at the timescales that appear relevant for the sort of ramping behaviour observed by Howe et al . Testing the second suggestion could be accomplished using photo-uncaging of dopamine in the accumbens ( for instance , using RuBi-Dopa [136] ) , since of the three mechanisms , it suggests the strongest coupling between dopamine and immediate behaviour . Optogenetically-stimulated release ( using TH-CRE or DAT-CRE lines ) could also be employed , although it would then be hard to distinguish the specifically dopaminergic component from any other influences of the ( potentially antidromically-stimulated ) activity of the dopamine neurons . It would be interesting to contrast the results of this with direct stimulation of D1-receptor-containing and D2-receptor-containing neurons [137] to try to assess downstream mechanisms . Testing the relationship between actor and critic is particularly tricky , since we know so little about the implementation ( or indeed existence ) of either and , in particular , the micro- or nano-scopic nature of choice over time [40] . Nevertheless , it would certainly be interesting to compare the nature and magnitude of ramping when subjects are made to wait for shorter or longer times , with and without cues for the precise passage of time that could be exploited . More generally , key issues surround the relationships between the number of dopamine cells that are active , the phasic and tonic activity of those neurons , the spatiotemporal profile of the concentration of dopamine at receptor targets in the accumbens , and the action of this dopamine on those receptors ( along with the action on target neurons of other neurotransmitters co-released by the same neuronal activity ) . This information is key for making qualitative and ultimately quantitative progress .
|
Dopamine has long been implicated in reward-motivated behaviour . Theory and experiments suggest that activity of dopamine-containing neurons resembles a temporally-sophisticated prediction error used to learn expectations of future reward . This account would appear to be inconsistent with recent observations of ‘ramps’ , i . e . , gradual increases in extracellular dopamine concentration prior to the execution of actions or the acquisition of rewards . We explore three different possible explanations of such ramping signals as arising: ( a ) when subjects experience uncertainty about when actions will be executed; ( b ) when dopamine itself influences the timecourse of choice; and ( c ) under a new model in which ‘quasi-tonic’ dopamine signals arise through a form of temporal discounting . We thereby show that dopamine ramps can be integrated with current theories , and also suggest experiments to clarify which mechanisms are involved .
|
[
"Abstract",
"Introduction",
"Models",
"and",
"Results",
"Discussion"
] |
[] |
2015
|
Tamping Ramping: Algorithmic, Implementational, and Computational Explanations of Phasic Dopamine Signals in the Accumbens
|
The propensity of segmental duplications ( SDs ) to promote genomic instability is of increasing interest since their involvement in numerous human genomic diseases and cancers was revealed . However , the mechanism ( s ) responsible for their appearance remain mostly speculative . Here , we show that in budding yeast , replication accidents , which are most likely transformed into broken forks , play a causal role in the formation of SDs . The Pol32 subunit of the major replicative polymerase Polδ is required for all SD formation , demonstrating that SDs result from untimely DNA synthesis rather than from unequal crossing-over . Although Pol32 is known to be required for classical ( Rad52-dependant ) break-induced replication , only half of the SDs can be attributed to this mechanism . The remaining SDs are generated through a Rad52-independent mechanism of template switching between microsatellites or microhomologous sequences . This new mechanism , named microhomology/microsatellite-induced replication ( MMIR ) , differs from all known DNA double-strand break repair pathways , as MMIR-mediated duplications still occur in the combined absence of homologous recombination , microhomology-mediated , and nonhomologous end joining machineries . The interplay between these two replication-based pathways explains important features of higher eukaryotic genomes , such as the strong , but not strict , association between SDs and transposable elements , as well as the frequent formation of oncogenic fusion genes generating protein innovations at SD junctions .
In humans , segmental duplications ( SD ) cover up to 5 . 2% of the genome [1] and are responsible for numerous gene-dosage imbalances [2] , gene fusions and disruption events 3 , 4 , 5 . Together with large insertions/deletions , SDs lead to gene copy number variations ( CNVs ) which represent a major source of polymorphism between individuals [6] . They have been associated with the development and evolution of both cancers [7] , [8] , [9] , [10] , [11] and genetically complex phenotypes such as predisposition to autism [12] , epilepsy [13] , Alzheimer disease [14] , glomerulonephritis [15] , systemic autoimmunity [16] and susceptibility to HIV/AIDS infections [17] . A specific mapping of CNVs on human chromosome 22 revealed that more than 2/3 of the breakpoints intersect with SDs [18] . This strong correlation reflects the similar nature of CNVs and SDs and suggests tightly coupled co-evolution mechanisms [19] . We previously designed a gene dosage assay in Saccharomyces cerevisiae to screen for the spontaneous duplication of a single gene , RPL20B [20] . Although the size of this gene is relatively small ( 1 . 6 kb ) , no single gene duplication was ever found . Instead , only intra- and inter-chromosomal duplications of large DNA segments , encompassing dozens of neighboring genes , were recovered ( 88% and 12% , respectively , Figure 1A ) [20] . These findings showed that spontaneous SDs can compensate for gene dosage imbalance by altering gene copy number in the yeast genome and that CNVs can encompass numerous genes . Approximately half of the SD junctions involved dispersed repeats such as Long Terminal Repeats ( LTRs ) from Ty retroposons , while the other half consisted of low complexity DNA sequences ( poly A/T , trinucleotide repeats ) , as well as microhomologous sequences whose identity spans only over a few nucleotides in length . The location and the type of sequences found at the breakpoints suggested that SDs might result from replication accidents improperly repaired through both homologous and non-homologous recombination events [20] . In order to explore the mechanisms of SD formation , we deciphered how perturbations of the replication process and of double strand break ( DSB ) repair pathways affect rates , types , sizes and breakpoint sequences of duplications . Providing the largest set of experimentally generated de novo duplications , the present study describes 338 independent SDs recovered in different mutant backgrounds and culture conditions . We show that replication-generated DNA ends are converted into large SDs through both homology-dependent and -independent replication-based mechanisms .
Two highly similar paralogous genes , RPL20A ( YMR242c ) and RPL20B ( YOR312c ) , encode the Rpl20 yeast ribosomal protein . The deletion of RPL20A results in a marked slow-growing phenotype which can be compensated by the spontaneous duplication of RPL20B [20] . Slow growing parental strains ( rpl20AΔ ) are propagated through serial transfer into rich medium . Rapidly growing revertants among slow growing populations are isolated by regularly plating aliquots of the cultures at each transfer step . Using this assay , we re-estimated the spontaneous duplication rate of RPL20B to be 1×10−7 SD/cell/division ( Luria-Delbruck fluctuation tests using the 0 term of the Poisson law ( p = 1−elnf0/ndiv; see Methods; Table 1 ) . This value is higher than previously estimated ( between 2×10−9 and 10−10 SD/cell/division [20] , [21] ) due to an initial underestimation of the time needed for a duplication-carrying cell to overtake the population of the slow growing parental cells ( see Methods ) . To confirm this surprisingly high value , an independent estimation of the duplication rate was achieved by designing a new selection assay based on the recovery of uracil prototrophy instead of growth recovery . In this system , RPL20A is not deleted and therefore both parental and duplicated strains show the same growth rate . Two truncated copies of the URA3 gene , overlapping by only 58 bp , were introduced in place of the two Ty3 LTRs , YORsigma3 and YORsigma4 located on either side of RPL20B and separated by 115 kb ( YKFB614 , Figure 1B ) . In the original growth-assay , approximately half of all SDs ( 48% , [20] ) , corresponds to an intra-chromosomal 115 kb direct tandem duplication between these two LTRs ( Figure 1B ) . The size of the URA3 overlapping sequences ( 58 bp ) is comparable to the largest identity region shared by the two LTRs ( 44 bp ) . Thus , recovery of a functional URA3 gene at the duplication breakpoints is indicative of direct tandem ura3-mediated SDs , mimicking the 115 kb LTR-mediated SDs . In this system , the duplication rate was evaluated to 0 . 9×10−7 event/cell/division ( using the median method [22] , Table 1 ) . To further test this rate , we created a rpl20AΔ derivative of YKFB614 ( YKFB605 , Table S1 ) and examined its duplication rate using the growth recovery assay . We found a rate of 1 . 7×10−7 , consistent with the fact that the rate derived from the URA3 assay represents only half of real duplication rate and close to our present estimate of 1×10−7 . This rate only accounts for duplications encompassing the RPL20B reporter gene , located on the right arm of chromosome XV . Therefore , extrapolation to the whole genome would lead to a much higher rate , suggesting that spontaneous SD events must be extremely common in yeast populations . For instance , a very high rate of histone gene amplification , compensating for decreased level of histones , was shown to result from recombination events between two Ty1 retroelements leading to supernumerary circular chromosomes [23] . However , our present estimate of SD rate is several orders of magnitude higher than that of other types of chromosomal rearrangements characterized in different studies using native yeast chromosomes [24] , [25] . This discrepancy could be explained by the absence of spatial constraints imposed on the boundaries of the SDs in our screen while in the other studies , the location of one end of the rearrangements is restricted within a narrow chromosomal region . To investigate the molecular mechanisms involved in SD formation , we used our selection system in conditions where replication is altered . Clb5 is a B-type cyclin known to activate late replication origin: in a clb5Δ strain S-phase duration is increased and the replication pattern modified [26] , [27] , [28] . The rate of SD formation is greatly increased in clb5Δ ( 730x compared to the control strain , Table 1 ) , unveiling the broad genomic instability induced by the perturbation of replication origin firing . Interestingly , the relative proportions of intra- versus inter-chromosomal SDs are conserved compared to the wild-type ( WT ) strain ( Table 1 ) . Although this is at the limit of statistical significance , the proportion of the 115 kb LTR-mediated duplications ( between the two Ty3 LTRs , YORWsigma3 and YORWsigma4 , Figure 1B ) is slightly increased in clb5Δ ( 62% compared to 48% in the WT , P = 0 . 05 Fisher's exact test , Table 1 ) . It is noteworthy that these LTR sequences lie next to tRNA genes whose transcription by PolIII is known to stall the progression of replication forks [29] . The size distribution of the intra-chromosomal SDs in clb5Δ remains globally similar to that of the WT ( see Figure 1C , in which WT and clb5Δ strains have radically different SD distributions as compared to rad52 and rad1 mutants ) . The breakpoint sequence of a non-LTR mediated SD was characterized through comparative genomic hybridization ( CGH ) and PCR amplification , revealing the presence of microhomologies at the junction ( Figure 2 ) . This junction is identical to the one found in the strain YKF1080 strain isolated in our original control screen [20]: the same two copies of a 9 nucleotide microhomologous sequence ( ACTTTTTTT ) have been involved in the formation of two independent SDs , recovered in two different genetic backgrounds . There are 2367 copies of this sequence in the genome , 47 of which interspersed between the two recombining sequences . It is unlikely that this repetitive use occurred by chance and therefore must be indicative of a chromosomal rearrangement hotspot . Interestingly , the centromere-proximal sequence lies next to an autonomous replicating sequence ( ARS1524 ) and the centromere-distal one corresponds to a replication termination site ( Figure 2 ) , which could explain their recurrent use in SD formation ( below ) . Altogether these results suggest that the mechanisms of SDs formation are similar in WT and clb5 strains . In addition , the dramatic SD rate increase associated with the clb5 mutation could be related either directly to the perturbed S-phase origin firing and/or to indirect effects of this perturbation onto replication . In this regard , the reported Rad9-dependent activation of the replication checkpoint key protein Rad53 , by late S-phase , strongly suggests that a CLB5 deletion results in the formation of replication-induced DNA breaks [30] . Such breaks could therefore represent the precursor lesions leading to SDs . In order to test whether broken replication forks could correspond to these precursor lesions , we monitored SD formation in cells treated with camptothecin ( CPT ) , a topoisomerase I inhibitor . CPT stabilizes the covalent intermediate that forms during the catalytic DNA nicking-closing cycle of Top1 , and CPT cytotoxicity results from the conversion of single strand nicks into double-stranded DNA ends when a moving replication fork collides with a CPT-Top1 complex [31] . The rate of SD formation is strongly increased in an exponential culture treated for 3 hours with 10 µg/ml of CPT ( x 320 , Table 1 ) . This observation could be explained if the precursor lesions leading to SDs were indeed double-strand DNA ends that in standard conditions would result from replication accidents . Several other lines of evidence support this hypothesis . First , SD breakpoints often correspond to sequences known to interfere with the replication forks progression ( Figure 2 and [20] ) . Moreover , replication-induced DNA damages in a clb5Δ strain [30] would explain the massive increase in SD formation observed in the absence of this cyclin . These lesions are likely to impede fork progression and trigger the activation of the replication checkpoint . Besides preventing fork collapse and the subsequent formation of DNA breaks , the replication checkpoint also regulates a large variety of cellular events including repression of late-replicating origins , inhibition of mitosis and induction of DNA repair genes [32] . This trans-acting branch of the replication checkpoint relies on the hyperphosphorylation of Rad53 that can be specifically abrogated with a mrc1AQ allele [33] . To determine whether SDs result primarily from S-phase induced DSBs rather than being secondary byproducts of the checkpoint activation , we characterized SD formation in a mrc1AQ mutant in presence of hydroxyurea ( HU ) . By inhibiting the ribonucleotide reductase activity , HU slows down replication fork progression and promotes the formation of ssDNA at the forks , which is sufficient to activate the checkpoint in normal cells [32] , [34] . In a mrc1AQ strain and in the presence of HU ( 100 mM for 3 hours; Methods ) , the integrity of stalled replication forks is maintained while the trans-acting branch of the replication checkpoint is suppressed . In these conditions we found no significant differences between HU-treated and untreated cultures in either checkpoint competent or deficient cells ( 2 to 5 fold increase , Table 1 ) . These results demonstrate that neither stalled replication forks nor the Rad53 hyperphosphorylation-mediated functions of the replication checkpoint are sufficient to stimulate SD formation . Altogether , the above findings strongly suggest that broken forks are the precursor lesions that are directly processed into SDs . Free DNA ends generated at broken forks are thought to be repaired primarily by strand invasion of the sister-chromatid , followed by the assembly of a new fork and subsequent replication up to the chromosome end ( or to the next replication fork ) . This break-induced replication ( BIR ) mechanism can occur through successive rounds of strand invasion and dissociation , and lead to chromosomal rearrangements if reinvasion occurs within ectopic repeated sequences [35] . We explored a potential role for BIR-related mechanisms by investigating SD formation in a pol32Δ strain . Pol32 is a non-essential subunit of S . cerevisiae major replicative DNA polymerase Polδ and is required for the replication fork assembly that initiates the BIR reaction [36] . Absence of Pol32 completely abolishes the formation of SDs . No mutant carrying duplications of any type were isolated out of 184 independent pol32Δ cultures . Thus , although the true duplication rate cannot be calculated , the occurrence of a single event , out of the 184 cultures , would have lead to a reversion rate of 6 . 9×10−9 . Although this value is an overestimate of the true duplication rate , it represents a 14-fold reduction compared to the WT control ( <0 . 07 , Table 1 ) . These data reveal the crucial role played by Pol32 in the generation of all types of SDs . Given that Pol32 is not required for repair by gene conversion ( GC ) events , SDs must therefore result principally from replication-based mechanisms rather than from unequal crossing-overs ( UCO ) between sister-chromatids . In addition , since only half of SDs contain repeated homologous sequences at their junctions [20] , classical BIR mechanism involving Rad52-mediated interactions between large sequences of homology could only account for half of all of the events: the other half might result from a Pol32-dependent replication-based mechanism involving microhomologous or low complexity sequences at the site of strand invasion ( see below ) . These two Pol32-dependant replication-based mechanisms must rely on an initial step of ectopic strand invasion . The Rad1/Rad10 complex possesses an endonuclease activity required for the removal of non-homologous tails during GC events [37] . This complex is also essential for the Rad52-independent microhomology mediated end-joining ( MMEJ ) DNA repair pathway [38] and was shown to promote the production of gross chromosomal rearrangements ( GCRs ) [39] . A deletion of the RAD1 gene results in a 5-fold reduction of SD formation ( x 0 . 2 as compared to WT , Table 1 ) , suggesting that the endonuclease activity is required to generate duplications . We also noted a substantial ( although not highly significant ) decrease in the proportion of LTR-mediated SDs compared to WT ( 14% vs . 48% , respectively; P = 0 , 06 , Table 1 ) , suggesting that Rad1 is directly involved in the generation of BIR-mediated SDs rather than of microhomology-related ones . Consistently , the proportion of small ( <115 kb ) intra-SDs is increased and reminiscent of the distribution of rad52-independent duplications ( Figure 1C; below ) . One SD breakpoint from a rad1Δ mutant was sequenced , revealing an eight-nucleotide homology at the junction ( Figure 2 ) and implying that , despite its predominant role in MMEJ , Rad1 is not required for the generation of microhomology-mediated SDs . Similar microhomologies were reported at the junction of GCRs recovered in rad1Δ and rad10Δ strains [39] . Interestingly the centromere-proximal microhomologous sequence involved in this rearrangement lies within the tRNA ( tA ( UGC ) O; Figure 2 ) that flanks YORWsigma3 ( the LTR recurrently used in the 115 kb intra-chromosomal SDs ) . Given that tRNAs transcription is able to stall incoming replication forks , these sequences were proposed to exhibit spontaneous fragility and thus promote chromosomal instability [40] . The eight-nucleotide microhomology sequence could therefore represent the recurrent breaking site which initiates the formation of the common 115 kb LTR-mediated SD: in the presence of Rad1 , the 3′ flap sequence between this break site and the LTR sequence would be excised so that a BIR-mediated SD could occur . It is generally believed that most SDs must result from non-allelic recombination events between dispersed repeats , but so far no demonstration for the involvement of the homologous recombination ( HR ) pathway in SD formation has been clearly established . In a rad52Δ strain where HR is abolished the class of 115 kb LTR-mediated SDs is completely suppressed ( 0 out of 71 independent events compared to 23 out of 48 duplications in the WT , P<10−6 , Table 1 ) . This result clearly demonstrates that this class of SDs results from Rad52-dependent recombination events between interspersed repeats . These duplication events are most likely resulting from a BIR reaction , since they are also dependent on the presence of Pol32 ( see above ) . Furthermore , while pol32Δ exhibits a limited reduction in GC efficiency [36] , absence of Rad51 restricts both BIR and GC events , although BIR occurs more frequently than GC among the remaining events [41] , [42] , [43] . In a rad51Δ strain , the rate of SD formation is increased ( x 7 . 7 , Table 1 ) . This increase suggests that the lesions that were repaired in wild type through gene conversion or allelic BIR are channeled into non-allelic BIR in rad51 mutants . In addition , the proportion of inter-chromosomal SDs increases up to 32% ( 10 out of 31 events ) as compared to 12% in the WT ( 6 out of 48 , P = 0 . 02 ) . All types of LTR-mediated SDs are favored in the absence of Rad51 ( 71% vs . 48% for the control , P = 0 . 02 , Table 1 ) . These findings suggests that Rad51 prevents recombination events between diverged sequences , such as the two LTR repeats YORWsigma3 and YORWsigma4 which share only 76% identity over 319 bp ( largest identical domain: 44 bp ) . This is consistent with the fact that Rad51-independent BIR requires shorter identical regions to achieve strand invasion than Rad51-dependent repair ( ∼30 bp vs . ∼100 bp , respectively [44] ) . Therefore , it might be that RAD51 does not simply suppress recombination between diverged sequenced , but normally promotes gene conversion ( and allelic BIR ) that usually outcompetes ectopic BIR . Altogether , the above results strongly suggest the following scenario for the formation of the class of 115 kb LTR-mediated SDs: ( i ) a DNA free end would arose from a broken replication fork in the vicinity of LTR YORWSigma3 ( potentially stalled within the tA ( UGC ) O tRNA gene ) , ( ii ) repair of the DSB occurs through a Rad51-independent strand invasion of the non-allelic LTR sequence , YORWSigma4 , ( iii ) followed by a Rad1-dependent 3′ flap removal and ( iv ) a Pol32-dependent conversion of this strand annealing intermediate into a replication fork generating a large intra-chromosomal SD through BIR ( Figure 1D ) . To further explore the contribution of homologous recombination to SD formation , a system where SDs result principally from HR events was designed . The two LTR sequences , YORWsigma3 and YORWsigma4 , were replaced in this strain YKFB608 by two truncated copies of the URA3 gene , overlapping with a 401 bp region of perfect identity , such that a URA3-mediated intra-chromosomal duplication would restore uracil prototrophy ( Figure 1B , Table S1 ) . As expected , all growth revertants isolated in an rpl20AΔ background resulted from duplication events corresponding to URA3-mediated SDs ( data not shown ) . Although the size of the URA3 overlapping sequences is similar to the size of the LTRs ( 401 and 319 bp , respectively ) , the rate of SD formation showed a 56 time increase compared to the original strain with intact LTRs ( 5 . 6×10−6 vs . 1×10−7 , respectively ) and a 62 time increase compared to a strain carrying only a 58 bp overlap ( 5 . 6×10−6 vs . 0 . 9×10−7 , Table 1 ) . These results confirm that the accumulation of divergence between dispersed repeats suppresses genome rearrangements , while increasing the length of sequence identity between these repeats promotes genomic instability . Indeed , the mismatch repair system can trigger an anti-recombination activity thereby limiting chromosome rearrangements between diverged repeats [45] . In addition , we monitored the effect of the POL32 deletion in this HR-based assay . In the absence of Pol32 , and in the absence of mismatches between repeated sequences , only a 23-fold increase is observed , as compared to the 62-fold increase characterized in the presence of this protein ( Table 1 ) . This corresponds to a 2 . 7-fold decrease ( 63/23 ) in the rate of uracil-prototroph formation in pol32Δ , a lesser effect that the >14-fold decrease observed in the growth recovery assay ( above ) . It also shows that in the absence of mismatches between repeated sequences , not all SDs require Pol32 . These Pol32-independent SDs likely result from UCOs between the repeated identical URA3 sequences . In the original assay , similar UCO events involving the flanking LTRs are probably suppressed due to divergence between the sequences . The rate of SD formation in a rad52Δ strain is slightly higher than in WT ( 2 . 8-fold increase , Table 1 ) , revealing that duplications can form even when HR is abolished ( as suggested previously in [46] , although using a very different system ) . The SDs recovered in a rad52Δ background appear radically different from those obtained in the WT strain . First , there is a significant decrease in the proportion of inter-chromosomal events , since all 71 SDs but one correspond to intra-chromosomal duplications ( versus 6 out of 48 events in the WT , P = 0 . 02 , Table 1 ) . Second , the size distribution of intra-chromosomal SDs is significantly biased towards smaller segments as most of them ( 57 out of 71 ) are smaller than 115 kb ( Figure 1C ) . Third , sequencing of the breakpoints revealed that only microhomologous ( between 8 and 9 nt ) and low complexity sequences ( polyT ) are now used to generate SDs ( Figure 2 ) . Interestingly , a recent report proposed that the large SDs in the human genome that cause the dysmyelinating PMD disease might result from replication fork stalling followed by homology-independent template switching , relying instead on the presence of microhomologies [47] . Our sequenced breakpoints once again coincide with replication-related elements , such as ARS , termination sites and tRNAs ( Figure 2 ) . Given the location and the nature of the initiating lesions , as well as the strict dependency to Pol32 ( see above ) , we conclude that the non-HR mediated SDs result from a new mechanism that would rely on an initial Rad52-independent recombination event , occurring between 5 to 10 bp of microhomology or stretches of low-complexity DNA sequences such as microsatellites , followed by a Pol32-dependent fork assembly initiating DNA synthesis ( Figure 1D ) . Therefore , we propose to designate this new mechanism MMIR for microhomology/microsatellite-induced replication . All of the above data clearly show that spontaneous SDs result from replication-based mechanisms . Nevertheless , the putative contribution of NHEJ to SD formation was addressed . NHEJ is strictly dependent on the activity of the ATP-dependent DNA ligase , Dnl4 ( also named Lig4 ) , as well as that of the Yku70/Yku80 DNA binding complex [48] , [49] When DNL4 is deleted , SDs arise at a slightly lower frequency ( x 0 . 8 , Table 1 ) , and present a similar proportion of LTR-mediated events ( Table 1 ) . Among the non-LTR mediated events , two junctions were sequenced . One lies next to a microsatellite ( GTT ) 14 identical to the one found in the WT strain YKF1057 [20] , again corresponding to the recurrent use of a particular sequences at SD boundaries . The other corresponded to a 10 bp-long sequence of microhomology ( TGACGCAAAT ) , repeated 109 times in the genome , in which the two recombining copies lie next to a tRNA gene and a replication termination site ( Figure 2 ) . Although all of these characteristics are very similar to SDs generated in the WT strain , there is , however , a significant decrease in inter-chromosomal duplications ( 0 out of 51 in dln4Δ versus 6 out of 48 in WT , P = 0 . 01 , Table 1 ) , suggesting that Dnl4 is required for inter-chromosomal SD formation . However , the junction sequences of the 6 inter-chromosomal events in WT were indicative of either LTR-mediated or microsatellite-mediated events ( 3 occurrences each , respectively ) [20] , [21] . These sequences differ strongly from those usually found at NHEJ-mediation junctions ( 1–4 nucleotides complementary sequences , [50] ) , suggesting that , in addition to its well-described role in NHEJ , Dnl4 might participate in the replication-based mechanisms of inter-chromosomal SD formation . In the double mutant rad52Δ dnl4Δ the rate of SDs formation is moderately increased ( x 4 . 3 ) compared to WT ( Table 1 ) . It is noteworthy that in the GCR assay , developed by Kolodner and collaborators , the concomitant deletion of RAD52 and DNL4 completely abolished the formation of non-reciprocal translocations since all GCRs observed resulted from telomere additions [51] . This discrepancy underlines the differential genetic requirements between SD and other GCR mechanisms . Deletion of RAD1 in the rad52Δ dnl4Δ strain reduced the SD rate to a level similar to that of the WT ( Table 1 ) , as expected since Rad1 promotes SDs formation ( above ) . The type of SDs , the size distribution as well as the breakpoint sequences isolated in the progenies of these double and the triple mutants strains , are similar to the ones characterized in the rad52 single mutant ( Figures 1C , Table 1 and Figure 2 ) . Therefore , when both HR and NHEJ are abolished and when MMEJ is , at least , severely compromised ( as in rad52Δ dnl4Δ rad1Δ strain ) , SDs still occur at a WT rate . Since SDs would mainly result from the replicative-repair of a one-ended DSB generated at a broken fork , the concomitant mutations of the 3 major DSB repair pathways should severely reduce if not abolish SD formation . The maintenance of a rate of formation similar to WT and the physical characteristics of SDs in this background suggest that MMIR could represent a new DSB repair pathway . Alternatively , these SDs could be formed by template switching , in the absence of any DSB , as suggested for the formation of PLP1-encompassing SDs in the human genome [47] . Altogether , 26 SD breakpoints were sequenced ( this work and [20] , [21] ) allowing the identification of 13 different chimerical Open Reading Frames ( ORF ) containing either microhomologies or trinucleotide repeats at their junctions ( Figure 2 ) . Microhomologies were found at breakpoint junctions in rad52Δ , dnl4Δ and rad1Δ backgrounds , where HR , NHEJ and MMEJ are impaired , respectively ( Figure 2 ) . This shows that these sequences can be used in the absence of all known DSB repair pathways . Because of their extremely high genomic density , the impact of microhomologies in SD formation , and more generally in genome dynamics , is likely to be important . For instance , the 8 to 10 nucleotide breakpoint sequences characterized in the rad52Δ , dnl4Δ and rad1Δ backgrounds are found in the S . cerevisiae genome from 109 times for the less frequent ( TGACGCAAAT ) , and up to 793 times for the most common ( TAGAGGA , Figure 2 ) . Chimerical genes arise either from in- or out-of-frame ORF fusions ( 3 occurrences each ) , from 3′ or 5′ ORF truncations ( 1 and 5 occurrences , respectively ) or from the fusion between an ORF and a tRNA ( Figure 2 ) . These fusions can generate new proteins and thus represent a potential mechanism of protein evolution . Whereas chimerical ORFs resulting from translocation and inversion events are associated with the concomitant lost of the original gene integrity , SD-mediated chimerical genes formation leave intact the original copies of the genes involved at the breakpoint . For instance , in addition to the original full-length gene a truncated copy of SGS1 ( homolog of human BLM ) has been found in the pathogen yeast species Candida glabrata [52] . This powerful mechanism allows SD-mediated chimerical genes to explore new combinations that might be counter-selected for in the cases of classical translocation- or inversion-mediated events . In-frame ORF fusions ( 3 cases ) might result in new protein architectures by combining previously existing domains . In addition , SD-mediated frameshift fusions and ORF truncations may result in true protein innovations at the junctions by promoting the transcription of otherwise non-coding sequences . The corresponding transcripts would encode entirely new amino acid combinations . For instance , the frameshift chimerical ORF generated in strain YKF1114 comprises a coding sequence whose last 47 amino acids ( from the breakpoint to the stop codon ) represent a truly new protein segment that shows no similarity to the rest of the yeast proteome . Such peptides were found in 5 out of the 13 chimerical ORFs characterized ( Figure 2 ) . Although relatively small ( average size of 28 amino acids ) , these peptides are new genomic features and may generate new protein domains . Despite their known association with diseases and genome rearrangements , it has been proposed that SDs have been fixed in the human genome to increase copy number of fusion genes originating from initial duplications of gene-rich core regions , eventually leading to the emergence of new gene families that are either unique to hominoids or considerably diverged when compared with other mammalian species [53] .
Given their close association with various genomic disorders and cancers and their broad evolutionary impact , SDs and CNVs represent one of the most important discoveries that stem from the human genome project . Careful computational characterization of SD breakpoints in the genomes of human and other primates has suggested an important role for Alu-mediated recombination in the production of intra- and inter-chromosomal SDs [54] . However , Alu elements are found in only 30% of the SD breakpoints and sequences presenting the physicochemical properties of “fragile sites” were shown to play an important role as well [55] . In addition , recent studies have proposed that SDs and other complex rearrangements associated with genomic disorders would result from replication-based mechanisms rather than from more classically invoked recombination-based models such as non-allelic homologous recombination between dispersed repeats [47] , [56] , 57 . Although essentially based on breakpoint analyses , these studies reach conclusions similar to those drawn here from experimental evidences . We found a massive SD rate increase both in a clb5Δ strain where origin firing is perturbed , S-phase is lengthened and DNA damages are detected by late S-phase [26] , [27] , [28] , [30] and in CPT-treated cultures in which single-strand nicks are converted into broken forks [31] . The recurrent use of genetic elements known to interfere with replication forks progression at SD breakpoints ( tRNA , microsatellites , ARS , replication slow zones and termination regions , Figure 2 ) also points towards the involvement of replication and the use of broken forks as the initiating lesions in the pathways leading to SDs . In addition , the finding that all SD formation requires the nonessential Polδ subunit Pol32 shows that duplications results from replication-based mechanisms rather than from UCOs , which are suppressed by the natural DNA divergence between dispersed repeats such as LTRs . It also suggests that BIR , which also requires Pol32 to initiate new DNA synthesis [36] would be the mechanism by which SDs are formed . However , BIR is a homologous recombination process which implies an initial Rad52-dependent invasion step necessitating large sequences of homology between the recombining molecules ( reviewed in [58] ) . These requirements imply that BIR cannot be the unique pathway leading to SDs , because only half of the SDs are generated through a Rad52-dependent recombination event between homologous sequences ( Table 1 ) . The remaining SDs occur independently from both Rad52 and large homologous regions and are generated through recombination between short identical/low complexity sequences . A Rad52-independent half-crossover pathway was previously described [59] , [60] and unequal half-crossovers in G2 could also generate tandem duplications . However , the class of Rad52-independent SDs described here involves only microhomology/microsatellite sequences at breakpoints and requires Pol32 , two characteristics that are hardly compatible with the half-crossover pathway . Given its unique substrate and genetic requirements , this new mechanism of SD formation has been called microhomology/microsatellite-induced replication , or MMIR , because it brings together characteristics from both MMEJ ( ie . recombination between microhomologies in a Rad52-independent manner , [38] ) and BIR ( ie . a Pol32-dependent DNA synthesis step , [36] ) . In addition , we show that MMIR-mediated SDs still form in the absence of all known DSB repair pathways ( HR , NHEJ and MMEJ ) suggesting that MMIR could represent a new repair pathway . Alternatively , one cannot exclude that MMIR-mediated SDs would arise in the absence of any DSB as a result of template switching events as it has been suggested for the large PLP1 duplications that cause the dysmyelinating PMD disease in human [47] . Altogether , our results provide the first experimental deciphering of the molecular pathways leading to SDs , demonstrating that two alternative replication-based mechanisms , BIR and MMIR , are responsible for the spontaneous SD formation in the yeast genome ( Figure 1D ) . While these two pathways probably use similar precursor DNA lesions and share the Pol32 requirement , they differ from one another by their recombination substrate and their dependency to HR proteins ( Rad52 , Rad51 and Rad1 ) . To our surprise , the Dnl4 ligase seems to contribute to the formation of inter-chromosomal SDs resulting from either BIR or MMIR . A similar Dnl4 requirement has been described for the formation of non-reciprocal translocations in S . cerevisiae [61] . Dnl4 has a preponderant role in NHEJ and also participates in MMEJ [38] , [48] , [49] . However , the sequences characterized at inter-chromosomal SD breakpoints ( LTRs and microsatellites ) are very different from typical signatures of either NHEJ or MMEJ events [38] , [50] . These results suggest that the role played by Dnl4 in inter-chromosomal SD formation would be different from the other known functions of this protein . Discrete microhomology/microsatellite sequences are recurrently used at SD breakpoints although hundreds , even thousands , of other identical copies are dispersed within the genome . These particular regions thus behave as duplication hotspots . Interestingly , they often correspond to genetic elements linked to replication initiation , progression and termination ( e . g . ARS , termination regions , tRNAs , replication slow zones; Figure 2 ) . Such correlation suggests that genomic architectural constraints may favor interactions between specific loci , for instance through promoting spatial proximity during replication . In yeast , two replication forks originating from the same replicon co-localize in the nucleus within a replication factory , a spatial location likely to harbor other forks as well [62] . The tight link between replication and SD formation raises interesting questions with regard to the influence of these factories on eukaryotic genome stability ( Figure 3 ) . A single broken fork could be repaired either in a Rad52-dependent or -independent manner ( Figure 3i or ii , respectively ) . The invading broken strand would presumably correspond to the lagging strand template where more ssDNA is exposed at the forks [32] . Given that SD formation requires Pol32 , the displacement of the lagging strand would also be compatible with the recent finding that lagging strand replication is performed by Polδ [63] . SDs recovered in the absence of Rad52 present a relatively smaller size ( median = 60 kb ) , reminiscent of the size of a replicon bubble in yeast . This may proceed from the possibility for a DNA free-end to interact spontaneously in a Rad52-independent manner with a sequence present in its vicinity within in the same replication factory ( Figure 3ii ) . In contrast , in a WT background where Rad52 is present , homology search would promote strand invasion between more distant sequences possibly located in different replication bubbles/factories , and thus generate larger duplications . Interestingly , in highly aggressive cases of neuroblastoma , an heterogeneous pediatric cancer , segmental chromosome instability results in unbalanced chromosome translocations , sometimes associated with additional aneuploidies [64] . These genomic profiles are formally similar to the different classes of inter-chromosomal duplications characterized in S . cerevisiae [20] . Whereas BIR is the mechanism usually invoked to account for the development of such chromosomal alterations [65] , the absence of repeated sequences at the breakpoints of many of these rearrangements suggests that MMIR may be an important path towards development of cancer .
All strains are derivatives of S . cerevisiae BY4743 ( MATa/α , his3Δ1/his3Δ1 , leu2Δ/leu2Δ , met15Δ/MET15 , lys2Δ/LYS2 , ura3Δ/ura3Δ ) [66] . Strain names and their corresponding genotypes and origins are summarized in Table S1 . Mutations were obtained either directly through a PCR-based deletion strategy or from EUROSCARF strains where the original geneticin resistance cassette KanMX4 was replaced by another resistance cassette . All constructions were verified by PCR and Southern blot analysis . For each mutation monitored , a diploid parental strain heterozygous for both the YMR242c ( RPL20A ) deletion and the deletion of the tested gene ( s ) was constructed then sporulated . Spores from the progeny carrying both the YMR242c deletion and the tested deletions were analyzed . In the growth-recovery assay , duplication rates were calculated from Luria-Delbruck fluctuation tests , either by using the 0 term of the Poisson law ( p = 1−elnf0/ndiv ) when a small subset of all cultures contained revertant cells ( see [20] for details ) , or using the median method when most of the cultures were overtaken by revertants [22] . In previous studies , the doubling time of a revertant culture was estimated to be twice as fast as the slow growing parental strain [20] , [21] . However , in the culture conditions where the selection assay was performed ( serial dilutions in 6 ml YPD in 24-wells plates ) , careful measurements revealed that the time needed for revertant cells to overtake slow growing populations was longer than predicted and was strain dependant: the doubling time of a duplicated strain is actually 1 . 3 to 1 . 4 times smaller than that of the slow growing parent , depending on the mutant background . This discrepancy resulted in a strong effect on the duplication rate estimation compared to our former studies ( from 2×10−9 to 1×10−7 per cell per division in control strain ) . In the strains used for the uracil-prototrophy recovery assay , the RPL20A gene is not deleted ( see Table S1 ) and both parental and duplicated strains show the same growth rate . Two truncated copies of the URA3 gene , covering either the 5′ or 3′ half of the gene and overlapping by either 58 bp or 401 bp , were introduced in place of YORWsigma3 and YORWsigma4 ( strains YKFB614 and YKFB608 , respectively , Figure 1B ) . The rate of appearance of uracil autotrophic colonies was determined by a fluctuation test analysis using the median method [22] . Briefly , ten independent YPD cultures , inoculated with ∼200 cells , were grown at 30°C to ∼3×108 cells/ml . Cells were plated on uracil lacking medium , incubated at 30°C for 2 days and [ura+] colonies were counted . The breakpoint junction indicative of a 115 kb ura3-mediated direct tandem duplication was sought through PCR amplification of the region . All [ura+] colonies analyzed carried such duplications , resulting from the fusion of the two URA3 overlapping sequences . Independent colonies ( 2×107 cells ) from strains YKF120c and YBaG398 were inoculated in 24 wells plates containing 6ml YPD , and cultivated under agitation during 6 generations at 30°C . Approximately 2×106 cells from each well were then inoculated into either fresh YPD medium , YPD supplemented by 100 mM Hydroxyurea ( HU , Sigma ) or YPD supplemented by 10 µg/ml Camptothecin ( CPT , Sigma ) , and incubated for 3 hours . After wash approximately 2×106 cells from each well were inoculated into fresh YPD medium . Every 10–11 generations , similar aliquots from each well were re-inoculated into fresh YPD medium . Between every cycle , a sample of the culture was plated onto YPD plates at a density of ∼2×102 to 5×103 cells/plate and incubated at 30°C ( above; [20] ) . Electrophoretic karyotypes of parental and revertant strains , as well as genomic DNA extraction and labelling , were performed as described [20] . Labelled DNA was hybridized against either PCR product-based ( Ecole Normale Superieure , Paris France and MWG Biotech ) or oligo-based yeast whole-genome arrays ( Affymetrix , YG-S98 ) . Arrays were analyzed with the GenePix Pro5 . 0 or with the Affymetrix GeneChip software , respectively . A genomic ratio for each ORF was defined as the ratio between normalized spot intensity of the revertant and parental strains , from which the mean of all spot intensities ratios was subtracted . SD junctions were PCR amplified . Products were purified using gel extraction columns ( NucleoSpin , Macherey Nagel ) and sequenced by the Genome Express company ( Cogenics ) .
|
Duplications of long segments of chromosomes are frequently observed in multicellular organisms ( ∼5% of our genome , for instance ) . They appear as a fundamental trait of the recent genome evolution in great apes and are often associated with chromosomal instability , capable of increasing genetic polymorphism among individuals , but also having dramatic consequences as a source of diseases and cancer . Despite their importance , the molecular mechanisms of formation of segmental duplications remain unclear . Using a specifically designed experimental system in the baker's yeast Saccharomyces cerevisiae , hundreds of naturally occurring segmental duplications encompassing dozens of genes were selected . With the help of modern molecular methods coupled to detailed genetic analysis , we show that such duplication events are frequent and result from untimely DNA synthesis accidents produced by two distinct molecular mechanisms: the well-known break-induced replication and a novel mechanism of template switching between low-complexity or microhomologous sequences . These two mechanisms , rather than unequal recombination events , contribute in comparable proportions to duplication formation , the latter being prone to create novel gene fusions at chromosomal junctions . The mechanisms identified in yeast could explain the origin of a variety of genetic diseases in human , such as hemophilia A , Pelizaeus-Merzbacher disease , or some neurological disorders .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"Methods"
] |
[
"evolutionary",
"biology/microbial",
"evolution",
"and",
"genomics",
"molecular",
"biology/chromosome",
"structure",
"microbiology/microbial",
"evolution",
"and",
"genomics",
"genetics",
"and",
"genomics/chromosome",
"biology"
] |
2008
|
Segmental Duplications Arise from Pol32-Dependent Repair of Broken Forks through Two Alternative Replication-Based Mechanisms
|
Pathogens , which alternate between environmental reservoirs and a mammalian host , frequently use thermal sensing devices to adjust virulence gene expression . Here , we identify the Yersinia virulence regulator RovA as a protein thermometer . Thermal shifts encountered upon host entry lead to a reversible conformational change of the autoactivator , which reduces its DNA-binding functions and renders it more susceptible for proteolysis . Cooperative binding of RovA to its target promoters is significantly reduced at 37°C , indicating that temperature control of rovA transcription is primarily based on the autoregulatory loop . Thermally induced reduction of DNA-binding is accompanied by an enhanced degradation of RovA , primarily by the Lon protease . This process is also subject to growth phase control . Studies with modified/chimeric RovA proteins indicate that amino acid residues in the vicinity of the central DNA-binding domain are important for proteolytic susceptibility . Our results establish RovA as an intrinsic temperature-sensing protein in which thermally induced conformational changes interfere with DNA-binding capacity , and secondarily render RovA susceptible to proteolytic degradation .
Most microbial pathogens occupy different ecological niches inside and outside their mammalian hosts . The temperature shift during entry from the surrounding biosphere or vector reservoirs , where temperature is generally lower than 30°C , into a thermally controlled host environment of 37°C is an important signal informing microbial pathogens to adjust their virulence programs [1] . Most pathogenic bacteria , including the yersiniae , have evolved sophisticated systems to sense the temperature of their environment [2] , [3] . Yersinia pestis is the etiological agent of plague and leads a sheltered lifestyle , alternately growing in fleas at moderate temperatures or in mammalian hosts at 37°C [4] . Yersinia pseudotuberculosis and Yersinia enterocolitica are fecal-oral pathogens that survive in moist environments . They are transmitted via ingestion , and cause gut-associated diseases , including enteritis , mesenteric lymphadenitis and diarrhoea [5] , [6] . Past studies addressing thermo-dependent changes in pathogenic yersiniae revealed that shifts between moderate temperatures and 37°C result in a global transition of gene expression , including multiple metabolic and stress adaptation genes , and most Yersinia virulence factors [7] , [8] . Among the virulence properties that are strongly expressed at environmental temperatures but weakly at 37°C are the heat-stable Y . enterocolitica specific enterotoxin Yst , iron-scavenging systems , smooth lipopolysaccharides , and the production of the primary internalization factor invasin in the enteropathogenic Yersinia species [3] , [7] . These characteristics seem to support initial colonization , penetration and survival in host tissues that are encountered during the very early stages of infection . While these properties are repressed at 37°C , expression of the virulence plasmid-encoded type III secretion system , the antiphagocytic Yop proteins and the adhesin YadA are induced [9] , [10] . This ensures that the bacteria remain cell adherent , become serum-resistant and are prepared for contact with phagocytic cells of the host immune system during ongoing infections . Although most Yersinia pathogenicity factors are thermally regulated , the devices for sensing temperature alterations and the molecular mechanisms how this sensory event is transmitted to globally regulate pathogenicity-associated pathways are not fully understood . Hitherto , it has only been shown that the virulence modulator YmoA of Y . pestis , is rapidly degraded by the Lon and ClpP proteases at 37°C , but not at 25°C . Furthermore , a thermo-responsive hairpin structure ( “thermoswitch” ) has been postulated within the transcript of the transcriptional regulator VirF . Both , YmoA and VirF , control the expression of the type III secretion machinery , translocating the antiphagocytic Yop effector proteins [11] , [12] . In this report , we analyze the thermoregulation of the global Yersinia virulence regulator RovA . The RovA protein belongs to the SlyA/Hor/Rap family of dimeric winged-helix DNA-binding proteins , which control a wide range of physiological processes implicated in environmental adaptation , survival , and pathogenesis in humans , animals and plants [13] , [14] . In pathogenic yersiniae , RovA coordinates the expression of multiple metabolic , stress and virulence genes , which contribute to colonization , host-associated stress adaptation and persistence [15] . In both enteropathogenic Yersinia species , RovA activates the transcription of the internalization factor invasin and other adhesins , allows a more efficient colonization of the Peyer's patches , and leads to a faster progression of the infection [16]–[18] . Moreover , dissemination of a Y . pestis rovA mutant to liver and lungs was reported to be drastically reduced , and its LD50 was shown to be about 80-fold higher than that of the wild-type , demonstrating that the RovA regulator protein is also important for the development of the bubonic plague [19] . The rovA gene is transcribed by two promoters in Y . pseudotuberculosis and Y . pestis , and by three promoters in Y . enterocolitica , at moderate temperatures ( 20–25°C ) , but is not expressed at 37°C [20] , [21] . Transcription of rovA is also modulated by growth phase and composition of the culture medium , and is positively autoregulated . Multiple RovA molecules bind to an extended AT-rich sequence far upstream in the rovA regulatory region and this interaction is required for full activation of rovA transcription [20] . Recent studies showed that rovA expression is subject to silencing by the nucleoid-associated protein H-NS [20] , [22] , [23] . Transcription of the rovA gene is also repressed by the LysR-type regulator RovM , which itself is controlled by the carbon storage regulator system of Yersinia , implicating small regulatory RNAs [18] , [24] . However , none of these established mechanisms can explain the thermal regulation of RovA synthesis . Here we report , that RovA represents a novel proteinaceous thermometer that senses temperature shifts directly through alterations in protein conformation thereby modulating its DNA-binding capacity . We further show that RovA is also subject to temperature and growth phase-dependent degradation by the self-compartmentalized proteases Lon and ClpP .
Although multiple regulatory factors were shown to be implicated in the environmental control of rovA transcription [18] , [20] , [24] , expression studies , in which rovA was expressed from the tetracycline promoter ( Ptet ) in E . coli , indicated that temperature and growth phase control of rovA expression occurs predominantly on the post-transcriptional level ( Figure 1A , B ) . RovA-dependent inv-phoA and rovA-lacZ fusions were still repressed under rovA non-inducing conditions ( 37°C , exponential phase ) , even when rovA was transcribed from Ptet . However , Ptet driven expression of the phoA gene showed that Ptet per se was not induced under these environmental conditions . We also performed Northern blot analysis and found identical amounts of the rovA mRNA expressed from Ptet at 25°C or 37°C during both growth phases ( Figure 1C ) . This argued that the amount of RovA in the bacterial cell is a function of temperature and growth phase , which does not depend on regulation of transcription . Genetic screens to identify components that up- or down-regulate rovA expression in response to temperature or attempts to show temperature-dependent modification of RovA , e . g . by 2D gel electrophoresis and phosphospecific fluorescent dyes failed . We therefore hypothesized that RovA might act as a protein thermometer , which is able to sense temperature of its environment directly through alteration in protein conformation without involvement of other cellular components . To test this , we first performed RovA DNA-binding experiments at 25°C and 37°C with different inv or rovA promoter fragments , harbouring one or more RovA binding sites . To demonstrate specificity of RovA binding , a control fragment encoding the csiD gene of E . coli was also introduced into the assay . The DNA-binding behaviour of RovA is significantly different at both temperatures , in such that the RovA protein binds DNA in vitro more avidly at low temperature . As illustrated in Figure 2A , significantly more of the same protein sample was required to cause the disappearance of free inv promoter fragments and the appearance of higher molecular RovA-DNA complexes at 37°C . To quantify these effects , we carried out band shift assays with a wide range of RovA concentrations , which allowed us to determine the apparent dissociation constant ( Kd ) . We found that at 25°C the RovA protein interacts with a Kd of about 32±5 nM and 45±4 nM with the binding sites I and II of the inv promoter , but it exhibits a considerably lower affinity ( Kd of about 183±24 nM and 190±23 nM ) to the same binding sites at 37°C ( Figure 2B ) . A similar thermo-induced reduction of RovA DNA binding was seen with rovA promoter fragments , harbouring all or only the RovA binding region I upstream of promoter P2 ( Figure 3A ) . RovA affinity to binding site I was about 4-fold reduced upon a temperature shift and occurred with a Kd of 46±3 nM at 25°C and a Kd of 178±5 nM at 37°C ( Figure 3B ) . Interestingly , the defect in RovA binding was especially apparent , when multiple RovA binding sites were present . No or only a very small amount of the highest molecular weight RovA-DNA complexes could be detected with the inv and rovA fragments , containing two or more RovA binding sites , implying that particularly cooperative binding of RovA on DNA is strongly reduced at 37°C . To further demonstrate that this temperature-dependent difference is a RovA specific property , we investigated DNA-binding properties of the RovM protein , which also interacts with the rovA promoter region . In contrast to RovA , no significant difference in the DNA-binding activity to the rovA promoter fragment was observed ( Figure 3C ) . At all tested RovM concentrations , RovM DNA-binding at 25°C ( Kd of 36±6 nM ) was similar or only slightly lower compared with 37°C ( Kd of 41±3 nM ) ( Figure 3D ) . A pH shift from 7 . 3 to 7 . 8 , covering the small temperature-induced increase in the band shift buffer system also had no effect on RovA DNA-binding at 25°C ( data not shown ) . This strongly indicated that observed thermo-dependent changes of RovA DNA-binding are not simply the result of thermo-dynamic effects or thermo-related changes in the pH . We also analyzed the DNA-binding capability of RovA after it was incubated at 37°C for more than 1 h , and allowed to cool down to 25°C . RovA was not affected by this treatment and exhibited similar DNA-binding properties as the protein which was only incubated at 25°C ( data not shown ) . Hence , incubation at higher temperatures does not irreversibly damage the function of RovA . Furthermore , we found , that loss and gain of function upon temperature up- and downshifts occurred very rapidly ( data not shown ) , suggesting that variations of the temperature between 25°C and 37°C induce conformational changes in RovA , that affect its DNA-binding capacity . Structural modelling of RovA based on mutant analysis and homologies with other crystallized MarR-type proteins predicted a highly α-helical dimeric protein , in which the first α-helix of one RovA subunit inserts between the last two α-helices of the other [23] ( Figure S1 ) . Interacting α-helical coiled-coil domains are known to be sensitive to temperature changes [25] . Consequently , loss of RovA DNA-binding functions at 37°C could be the result of thermo-induced conformational changes within the RovA protein . To test this possibility , the temperature effect on the overall conformation of RovA was investigated by using Circular Dicroism ( CD ) spectroscopy . CD spectroscopy was performed with highly purified RovA between 200nm and 250nm at 25°C and after heating the samples to 37°C ( Figure 4A ) . The CD spectrum confirmed RovA being a highly α-helical protein ( Table 1 ) . It also showed that a temperature increase to 37°C led to a profound change of the profile , indicating that α-helical structures were lost , whereas the content of β-sheets increased . The CD spectrum stabilized within 10 minutes following the shift to 37°C , indicating that major changes to the structure were complete by this time ( data not shown ) . In contrast , CD spectroscopy of control proteins ( RovM Figure 4B , lysozyme Figure S2 ) did not lead to detectable conformational changes under identical conditions . Furthermore , the overall shape of the CD spectrum of RovA did not change radically , indicating that the overall structure was only partially altered . In fact , reversibility of the thermal unfolding transitions could be demonstrated by the recovery of the initial secondary structure profile . After recooling to 25°C , the spectrum was similar to the original spectrum recorded at 25°C ( Table 1 , Figure 4A ) . Temperature stability of RovA was further investigated using a temperature scan from 20°C to 60°C , and the melting point ( TM ) of RovA was determined to be 44 . 5°C ( Figure 4C ) . Stability curves , performed with control proteins , usually proceed without a strong slope to the melting point of the particular protein ( Figure 4D , Figure S2 ) . However , a significant gradient between 20°C and 40°C was observed for the RovA regulator protein , also implying conformational changes within this temperature range ( Figure 4C ) . Taken together , temperature-dependent DNA-binding property seems a consequence of reversible conformational alterations of RovA in the range of 25°C–37°C . Conformational transitions with respect to the surrounding temperature outside and inside hosts could represent a reversible conversion between an active and an inactive form . As the purified RovA protein itself is stable even after several rounds of temperature-induced conformational transitions ( Figure 4 ) , it was expected that similar levels of the RovA protein are produced at 25°C and 37°C when the rovA gene is expressed from a foreign promoter to exclude autoactivation . However , lower levels of the RovA protein were detected at 37°C than at 25°C in Y . pseudotuberculosis ( Figure 5A ) and E . coli ( Figure 5B ) during exponential and stationary phase , when the rovA gene was transcribed from Ptet . In contrast , no alteration of RovM levels was observed , when the rovM gene was transcribed from Ptet under the identical growth conditions ( Figure 5C ) . It is possible that conformational changes upon a temperature shift from 25°C to 37°C lead to an altered stability of the RovA protein in vivo . For this reason , we first investigated the influence of different growth temperatures on the stability of RovA in Y . pseudotuberculosis YPIII after blockage of the protein biosynthesis . As shown in Figure 6A , RovA remained stable at 25°C during stationary phase for at least 90 min . At 37°C RovA levels were somewhat reduced , but considerable amounts of RovA were still detectable . During exponential growth , endogenous RovA levels were reduced , as described previously [17] . Furthermore , RovA was rapidly degraded when the culture was shifted to 37°C , and no or only very low amounts of the RovA protein were detectable 30 min after cessation of protein synthesis ( Figure 6B ) . RovA was significantly more stable at 25°C during log phase as considerable amounts of the regulatory protein were still visible 90 min after protein biosynthesis was stopped . A similar temperature- and growth phase-dependent RovA degradation pattern was also observed with E . coli DH5αZ1 pHT123 , in which rovA was expressed from Ptet to allow a stronger synthesis of the RovA protein during exponential growth ( Figure 6C–F ) . In contrast , no difference in RovM levels was detected after blockage of protein synthesis in Y . pseudotuberculosis and after overexpression of rovM from Ptet in E . coli under the same conditions ( Figure S3 ) . These data suggested that RovA is not only inactivated , but is also rapidly degraded under non-inducing conditions , and indicated that a conserved mechanism is responsible for this post-translational control process . Pathways implicated in the proteolysis of regulatory components often involve ClpP or Lon proteases . To determine whether these proteases are responsible for RovA degradation , we constructed Y . pseudotuberculosis clpP , lon and clpP/lon deletion mutants and compared expression of a rovA-lacZ fusion and steady-state levels of the RovA protein between wild-type and the protease mutants at 25°C and 37°C ( Figure 7A ) . The lon and the clpP/lon mutant contained significantly more RovA than the wild-type at 25°C . In contrast , only a slight increase of RovA levels was observed in the clpP mutant , suggesting that RovA is mainly degraded by the Lon protease . Interestingly , no or only very low amounts of the RovA protein were found in cultures grown at 37°C , even in the absence of the ClpP and Lon proteases . A similar expression pattern was also observed with a plasmid-based rovA-lacZ reporter system . This observation supports our previous results and can be explained by the fact , that the DNA-binding capacity of RovA to its own promoter sequence is strongly reduced at 37°C ( Figure 3A ) . Thus , autoactivation of rovA is abolished at 37°C even in the absence of RovA-degrading proteases . To confirm our assumption , we repeated this experiment with a construct in which rovA was now expressed under the control of Ptet to avoid positive autoregulation ( Figure 7B ) . In fact , similar levels of the RovA protein were now synthesized at both temperatures , with significantly higher levels in the lon and lon/clpP mutant , and higher expression of the RovA-dependent rovA-lacZ fusion was detected at 25°C in the protease mutants . However , no or very little activation of rovA transcription was evident at 37°C , even when RovA is abundant . This implies that not proteolysis , but the effect of thermo-induced conformational changes on DNA-binding is most critical for RovA-dependent regulation of rovA expression . To test the impact of the ClpP and Lon proteases on RovA stability in more detail , we expressed RovA from Ptet to exclude autoregulation and guarantee efficient RovA production during log phase . RovA stability was first studied in the wild-type and the protease mutants , when incubated at 25°C or 37°C during exponential phase . As shown in Figures 8 and S4A , RovA concentrations in the wild-type decreased rapidly at 37°C ( t1/2 = 30 min ) , and no or only very low amounts were detectable 1 h after cessation of protein synthesis . Fast degradation of RovA was also observed in the clpP mutant at 37°C , although slightly higher RovA levels were detected than in the wild-type background . In contrast , RovA levels were only slightly reduced at 37°C in the lon mutant ( t1/2>3 h ) , and no degradation of the RovA regulatory protein was detectable in the clpP/lon double mutant ( Figure 8 , Figure S4A ) . In agreement to previous experiments , no or significantly less degradation was observed in the wild-type and protease mutant cultures which remained at 25°C . This degradation pattern was also observed in bacteria grown to stationary phase , yet at much later time points after block of protein synthesis ( Figure S5 ) . In summary , albeit ClpP proteases contributed to RovA degradation , proteolysis of RovA was predominantly mediated by the Lon protease . It has previously been shown that the N- and C-termini of individual proteins can determine the sensitivity to proteolysis . For instance , non-polar amino acids or hydrophobic extensions at the C-termini , e . g . added by the SsrA-tagging system , were shown to destabilize polypeptides [26] , [27] . The importance of the N-termini for degradation is represented by the N-end rule , which relates the in vivo half-life of a protein to the identity of its N-terminal residues , and was also shown by the instability conferred by an N-terminal fragment of the UmuD protein of E . coli or the HemA protein of Salmonella typhimurium [28]–[30] . In order to identify amino acid motifs of RovA important for protease susceptibility , we analyzed the stability of terminally His6-tagged RovA variants , and tested degradation of different RovA deletion mutants missing the last 5 , 23 and 36 C-terminal amino acids [23] . We found that their susceptibility to proteolysis is similar to native RovA ( data not shown ) . We further determined the stability of different RovA-LacZ hybrid proteins . Degradation of the RovA1–127-LacZ chimer , harbouring the first 127 amino acids of RovA , was identical to native RovA , and its turnover was blocked in a lon mutant ( Figures 9A , B and S4B ) . Continuous deletions of the C-terminal portion of RovA revealed that RovA-LacZ hybrid proteins constituted of the N-terminal 96 or more amino acids were rapidly degraded at 37°C , but not in a lon mutant background . A RovA-LacZ fusion protein , including only the first 74 amino acids of RovA , remained more stable at 37°C and was present in considerable amounts even 90 min after blockage of protein synthesis , whereas chimera with the first 42 or less amino acids of RovA were completely stable under the same conditions ( Figures 9C–E , S4B data not shown ) . This demonstrated that addition of the N-terminal 96 residues of RovA to the otherwise stable β-galactosidase is sufficient to signal the chimera for degradation , whereas the first 42 amino acids of the regulatory protein are not . Taken together , this indicated that residues within the center of RovA are important for regulated proteolysis of this protein . The internal portion of the RovA protein forms the winged-helix DNA-binding domain [23] ( Figure S1 ) . Previous work in this study showed that DNA-binding is significantly reduced at 37°C , i . e . conditions under which RovA proteolysis is considerably enhanced . This suggested that binding of RovA to DNA might protect the protein from degradation . Accordingly , we determined the effect of multiple copies of the RovA-binding sequences on the stability of native RovA in vivo . To do so , we transformed plasmid pKH24 harbouring the sequences including all RovA binding sites of the rovA promoter region and the empty vector pUC19 into Y . psudotuberculosis YPIII , and determined the stability of the native RovA protein in the exponential phase at 37°C . Presence of multiple copies of the RovA binding sites led to somewhat higher RovA levels ( Figure S6A ) . Although RovA was still degraded , low amounts of the regulatory protein were still visible 90 min after blockage of protein synthesis . Furthermore , we found that introduction of an amino acid substitution in the winged helix domain of RovA ( RovAE71K ) , shown to diminish DNA-binding [23] , reduced the in vivo half-life of the RovA protein during exponential growth at 37°C ( Figure S6A ) . This suggested , that RovA is less accessible for the proteases when bound to DNA . The analysis of RovA stability in Y . pseudotuberculosis revealed that Lon is the protease primarily responsible for in vivo degradation of RovA ( Figures 7–9 ) . To establish a direct relationship between RovA degradation and the Lon protease , we tested whether Lon is capable of digesting RovA in an in vitro system , containing only purified RovA and Lon of Y . pseudotuberculosis , ATP and a system for ATP regeneration . To ensure that purified Lon protease is active , we used purified α-casein as a control substrate , which was known to be also prone to Lon-mediated proteolysis [31] . Figure 10 shows that α-casein was rapidly degraded by the Lon protease at 25°C and 37°C , demonstrating that the purified Y . pseudotuberculosis Lon protein is active and able to degrade substrates with similar activities at both temperatures . This indicated that temperature-dependent proteolysis of RovA is not primarily attributed to an increased activity of the Lon protease at 37°C . Quantification of native RovA after in vitro degradation further showed that RovA is only very slowly degraded by the Lon protease at 37°C , and remained completely stable at 25°C ( Figures 10 , S4C ) . Similar results were also obtained when purified N- or C-terminally His-tagged variants of the Yersinia or E . coli Lon protein were included in the in vitro assay ( data not shown ) . Slow degradation of RovA in the in vitro system suggested that efficient proteolysis in vivo requires accessory factors or chaperones , which may affect recognition and enzymatic activity of the Lon protease and/or assist in unfolding of the substrate . To explore whether additional components are required for Lon-mediated proteolysis of RovA , we tested degradation of RovA in a crude extract of exponentially grown E . coli strain BL21 ( lon− ) at 37°C after addition of the purified Lon protein of Y . pseudotuberculosis in the presence and absence of the ATP regeneration system . Efficient RovA degradation can be observed in the presence of the Lon protease and ATP ( Figures 10F , S4C ) , but not in the absence of ATP or Lon , or with extract from cells grown under conditions that do not result in RovA degradation ( Figures 10E , S4C and S7 ) . This strongly suggests that one or more conserved accessory factors are essential for RovA proteolysis . The lon gene in E . coli is part of the heat shock regulon and increases after exposure to high temperature ( 42°C ) in a manner that affects overall rates of protein degradation [32] , [33] . To test whether also synthesis of the Lon protease is increased in Y . pseudotuberculosis under conditions in which RovA is preferentially degraded , we analyzed lon expression of Y . pseudotuberculosis YPIII at 25°C and 37°C during exponential and stationary phase . As shown in Figure S8 , expression of a lon-lacZ fusion was significantly elevated at 37°C . Moreover , low amounts of the Lon protein were detectable in cell extracts of YPIII at 25°C , whereas significantly higher levels of the protease were found at 37°C . Although analysis of RovA degradation after blockage of protein synthesis clearly demonstrated that thermo- and growth phase-dependent proteolysis of RovA occurs mainly through a superimposed post-translational mechanism ( Figures 8–10 ) , it is very likely that increased synthesis of the Lon protease at 37°C also contributes to thermo-regulated proteolysis of RovA .
In this study we demonstrate that the global MarR-type virulence regulator RovA of Yersinia acts as an intrinsic protein thermometer , that controls its DNA binding activity and regulates its degradation by the ATP-dependent protease Lon , a process which is also subject to growth phase control . RovA activity is shown to be strongly dependent on temperature , both in vivo and in vitro , in such that it binds with higher affinity and enhanced cooperativity to DNA at lower temperatures . According to our model ( Figure 11 ) , the sensor-regulatory activity is based on a conformational adaptation of RovA in response to temperature for regulation of its DNA-binding function . The loss of structured elements upon a temperature shift from 25°C to 37°C strongly supports partial defolding of RovA in this temperature range . This makes RovA less capable of binding DNA in a cooperative manner , and reduces its ability to stimulate the inv and rovA promoters . Thermo-induced conformational changes are reversible , as α-helicity and DNA-binding capacity of RovA are regained upon cooling to 25°C , and this could be important for the regulatory function of RovA . To the best of our knowledge only two other bacterial regulators , which belong to a different regulator class are capable of responding directly to temperature: the transcriptional regulator TlpA of Salmonella enterica serovar Typhimurium which exhibits homology to KfrA and the SMC proteins involved in plasmid partition or chromosome segregation and the heat shock gene repressor protein RheA of Streptomyces albus [34]–[36] . The dimeric TlpA protein forms a long left-handed supercoiled coiled-coil domain in which two subunit α-helices wind around each other and pack their side chains in a “knobs-into-holes” manner [35] . The coiled-coil structure is a versatile and rather flexible motif in mediating protein-protein interactions , and it has been shown that TlpA undergoes a reversible conformational switch in response to temperature changes , leading to alterations between the unfolded monomeric form and the folded DNA-binding coiled-coil oligomeric structure . Also the RheA protein acts as a protein thermometer with shorter coiled-coil domains , and a thermo-induced change in the repressor leads either to an active or inactive form [36] . The structure of the MarR-type regulator RovA is significantly different from TlpA and RheA . Its internal region contains the DNA-binding domain that is predicted to adopt a winged-helix fold . The first N- and the last two C-terminal α-helices appear to form an extensive and well-packed dimer interface , similar to that seen in the structure of other MarR-type regulators . These help to stabilize the formation of dimers with properly positioned DNA-binding segments [23] . This type of dimer formation makes it very unlikely that thermo-sensing of RovA is based on a monomer-to-dimer conversion . In fact , we found no evidence to suggest dissociation of the dimers into monomers at 37°C; in contrast , RovA dimers are still detectable after heating to 95°C in the presence of SDS ( H . Tran-Winkler , unpublished results ) . However , our previous structural-functional analysis showed that even small changes , i . e . single amino acid substitutions in the N-terminal region ( L12A , W16A ) , which have no detectable effect on RovA dimer formation , cause a severe defect in DNA-binding [23] . This suggests that temperature-dependent structural changes within thermo-sensitive elements of the RovA dimer might effect the proper positioning of the DNA-binding segments . Our studies further demonstrate that the level of RovA in the bacterial cell is not only determined by transcription , but is also subject to growth phase- and temperature-regulated proteolysis . The RovA protein is stable during stationary phase and/or at moderate temperatures ( 20°C–25°C ) , but becomes highly unstable during exponential growth at 37°C . Degradation of RovA under these conditions is primarily mediated through the Lon protease , albeit the Clp proteases also appear to participate to a very small extent in the degradation process . Both types of proteases are ATP-dependent and assigned to the AAA+ superfamily of ATPases . Lon and ClpP proteases share two common features: ( i ) access to the proteolytic chamber of the enzyme is usually prohibited to globular proteins , most likely to prevent unrestrained protein degradation , and ( ii ) they require ATP hydrolysis to unfold and translocate the substrates into the protease chamber . Degradation within the chamber is processive , including sequential rounds of substrate binding , release and rebinding to the proteolytic site , and generates 10–15 amino acid peptides without generation of partially digested protein intermediates [37]–[40] . The ClpP and Lon proteases usually degrade improperly folded or damaged proteins . However , undamaged proteins , in particular short-lived regulatory factors which are implicated in developmental processes , stress resistance and bacterial fitness can also serve as substrates for proteolysis [37] . According to our results , RovA degradation by the Lon protease is clearly temperature-dependent , but alterations of the DNA-binding activity due to thermo-induced conformational changes seems much more critical for thermo-dependent rovA expression than the change of RovA stability . For instance , little autoactivation of rovA transcription is evident at 37°C , even when RovA is abundant in the absence of the Lon and Clp proteases ( Figure 7 ) . This raises the question of why is RovA also rapidly degraded at 37°C . Thermo-induced structural alterations led to a strong reduction , but did not cause a complete inactivation of the DNA-binding function of RovA ( Figures 2 , 3 ) . Accordingly , degradation of the global regulator might be required to prevent binding to higher affinity sites of RovA within the Yersinia genome . Homologous MarR-type regulators were also shown to interact with small metabolites or signalling molecules [41] , [42] . In this case , rapid proteolysis of RovA might be essential to prevent sequestration and inactivation of important regulatory components through complex formation with RovA . As RovA proteolysis is also responsible for growth phase regulation of rovA and RovA-dependent genes , thermo-dependent alterations of the degradation process might be important to link and coordinate thermoregulation and growth phase control ( Figure 11 ) . To date very little is known as to what feature identifies transcriptional virulence regulators as substrates of Lon or Clp proteases . Certain peptide motifs of an exposed or unstructured region of a protein , including a few non-polar , aromatic amino acid residues can serve as recognition signal for proteases [31] , [43] , [44] . Although the major determinants for Lon-mediated proteolysis are frequently found at the N- and C-termini of target proteins [26] , [28] , [30] , [45] , it has recently been reported that Lon can obviously also recognize internal tags [44] . In MarR-type regulator proteins , such as RovA , both termini form α-helical structures , which contribute to the formation of the dimer interface [23] , [41] . Here we report that transplantation of the N-terminal 96 amino acids of RovA to a normally stable protein confers instability to Lon , but a transfer of the first 42 amino acids does not . Also introduction of a mutation abolishing the DNA-binding function of RovA rendered the protein more susceptible to Lon , whereas presence of a multi-copy plasmid harbouring RovA-binding sites reduced degradation of the protein . This suggests that amino acid residues in the vicinity of the central winged-helix DNA-binding domain comprise the information necessary for Lon binding and/or provides the foundation for degradation by Lon . This finding raises the intriguing possibility that Lon recognition and degradation of RovA is in direct competition with the DNA-binding function of the regulator . In fact , two other substrates , SoxS and the N protein of bacteriophage λ are known to be protected from Lon-mediated degradation when bound to DNA , although their instability is an intrinsic property and does not require an external signal to trigger degradation [46] , [47] . From results in this study we know that temperature strongly affects the susceptibility of the RovA protein to Lon-mediated proteolysis . In this regard , it is very likely that the protein degradation signals are normally buried in active RovA and less accessible in RovA-DNA complexes , but become more accessible to the AAA+ protease in the non-bound state as a consequence of thermo-induced defolding events ( Figure 11 ) . Lon-mediated turnover of RovA in vivo is likely modulated or regulated by accessory factors that affect either the enzymatic activity of the protease or the conformational state of the target protein . Experiments with an in vitro degradation system demonstrated rapid degradation of α-casein by the Yersinia Lon protease , but RovA proteolysis was significantly less efficient than proteolysis in vivo . In contrast , rapid degradation of RovA could be observed when crude extract of an exponentially grown lon mutant strain was added to the in vitro system , indicating that an additional component is required for RovA-mediated proteolysis . As RovA degradation is significantly reduced during stationary phase , it seems likely that the postulated accessory component is only present during exponential growth ( Figure 11 ) . Lon-mediated proteolysis might be influenced by interactions with partner proteins , such as adaptors that tether substrates , or chaperones and proteases , which might create or help to expose the degradation signals of RovA . For example , the DnaJ/DnaK/GrpE molecular chaperone system is required to promote formation of certain AAA+ protease-substrate complexes [48] , and only endoproteolytic cleavage and interaction with the SspB adaptor protein render the transmembrane RseA protein susceptible for ClpXP-mediated proteolysis [49] . Activity of AAA+ proteases was further shown to be subject to modulation by cellular compounds , ions and metabolites [50]–[52] . Since MarR-type regulators are frequently modulated by small effector molecules [41] , [42] , it will also be interesting whether RovA also undergoes effector-induced conformational changes , which could be crucial for growth phase-dependent RovA degradation .
Strains used in this study are listed in Table S1 . If not indicated otherwise , bacteria were grown at 25°C or 37°C to exponential phase ( OD600 = 0 . 4–0 . 6 ) or stationary phase in LB broth supplemented with antibiotics as follows: ampicillin 100 µg ml−1 , chloramphenicol 30 µg ml−1 , tetracyclin 5 µg ml−1 , and kanamycin 50 µg ml−1 . Plasmids and primers used in this study are listed in Tables S1 , S2 and S3 . The His6-RovA overexpression plasmid pHT95 was cloned by inserting a PCR fragment amplified with primers 1 and 2 into the BamHI/PstI sites of pQE30 . pHT105 was constructed by inserting the AatII/AvrII fragment of pZE21 , harbouring the PtetO-1 promoter , into pZS*24 . Plasmid pHT123 was constructed by inserting a rovA+ fragment amplified with primers 3 and 4 into the KpnI and ClaI sites of pHT105 . Plasmid pHT125 was obtained by inserting a PCR derived fragment amplified with primers 5 and 6 from pTAC3575 into plasmid pGP704 cut with SacI and SmaI . For the construction of pKH01 harbouring the lon-lacZ fusion , a PCR-derived fragment amplified with primers 7 and 8 from YPIII chromosomal DNA was inserted into pGP20 cut with HindIII and EcoRI . pKH04 was obtained by inserting the KpnI-HindIII phoA+ fragment of pHT125 into pHT105 . Plasmids pKH08 and pKH26 were obtained by QuikChange mutagenesis ( Stratagene ) using primer pairs 9/10 and 11/12 . The rovA-lacZ fusion plasmids pKH15–23 harbouring different portions of the 5′-end of the rovA gene were obtained by the insertion of PCR-derived fragments amplified with the upstream primer 13 and the downstream primers 14–21 . Plasmid pKH31 was constructed by insertion of a PCR fragment amplified with primers 22 and 23 , cloned into the KpnI and ClaI sites of pHT105 . The rovA promoter sequence harbouring the RovA binding sites were amplified by PCR with primers 24 and 25 and cloned into pUC19 , generating pKH24 . Construction of pKHTS3 was performed by ligation of a PCR fragment amplified with primers 26 and 27 into the SacI and PstI sites of pBAD-HisA . For the construction of pMB113 , a lon+ PCR fragment , amplified with primers 28 and 29 , was inserted into the NcoI and BglII sites of pQE60 . For construction of pMB114 the tetR gene was amplified from pACYC184 using primers 30 and 31 , and inserted into the XhoI and SacI sites of pHT123 . All deletion mutants are derivatives from wild-type strain YPIII and were generated using the RED recombinase system ( Derbise et al . , 2003 ) as described in detail in our previous study [24] . For the construction of E . coli strain KB2 , a kanamycin cassette was integrated into the hns locus of E . coli and removed as described [53] . The kanamycin or ampicillin resistance gene was amplified with primer pairs 32/33 or 34/35 ( Table S2 ) . Primers used to amplify 500-bp regions flanking the target genes of Y . pseudotuberculosis and primers used for E . coli strain KB2 construction are given in Table S3 . All His-tagged proteins were overexpressed in E . coli BL21λDE3 , native RovA was expressed in E . coli strain KB2 . Overnight cultures of E . coli strains , harbouring the overexpression plasmids pLW1 ( rovA+ ) , pLW2 ( rovA-his6 ) , pHT95 ( his6-rovA ) , pAKH43 ( rovM+ ) , pKHTS3 ( his6-lon ) or pMB113 ( lon-his6 ) were diluted 1∶100 and grown at 37°C in M9 medium for 2 h . Subsequently , protein synthesis was induced with 100 µM IPTG or 0 . 02% arabinose ( pKHTS3 ) and grown to an A600 of 0 . 6 . Cells overexpressing His-tagged proteins were purified as described [24] . Cells expressing the native RovA protein were resuspended in lysis buffer ( 10 mM Tris-HCl pH 8 . 0 , 1 mM EDTA , 5 mM DTT , 5% glycerol ) and purified on a dsDNA cellulose column ( Amersham ) using elution buffer ( 10 mM Tris-HCl pH 8 . 0 , 1 mM EDTA , 5 mM DTT , 5% glycerol , 300 mM NaCl ) . For the in vitro degradation assay , proteins were dialyzed against 50 mM Tris-HCl plus 10 mM MgCl2 . The purity of the proteins was estimated to be >95% . The DNA retardation assays were performed as described [18] . The fragments of the inv promoter regions including binding site I , II or both were amplified with the primer pairs 36/37 , 38/39 and 40/41 . The rovA promoter fragments including the RovA or RovM binding sites were amplified with primers 42/43 , 42/44 and 45/46 , respectively . As a negative control , a 350 bp long fragment from the E . coli csiD gene was amplified using primers 47 and 48 . Alkaline phosphatase and β-galactosidase activity were determined as described [24] . Preparation and separation of cell extracts as well as Western blotting with polyclonal antibodies directed against RovA , Lon , His-tag or LacZ were performed as described [17] , [20] . 0 . 16 mg/ml purified RovA , 0 . 4 mg/ml purified RovM and 0 . 16 mg/ml lysozyme in CD buffer ( 10 mM NaH2PO4 pH 8 . 0 , 10 mM NaCl , 5 mM DTT , 1 mM MgCl2 ) were used for CD spectroscopy and CD spectra were recorded on a Jascow J-810 spectrometer using a thermo-stated cell holder . Each spectrum was the result of five successive spectra , each normalized against the CD buffer . Spectra were recorded starting at 25°C . A 10 min equilibrium delay was allowed , after raising the temperature to 37°C . After re-shifting the temperature to 25°C , 2 hrs of equilibration time were allowed . Secondary structure estimation was performed using the analysis software provided by the Jascow J-810 spectrometer , which makes use of the Yang algorithm . Temperature stability was investigated by using a temperature scan from 20°C to 60°C with a temperature slope of 2°C/min at a fixed wavelength of 222 nm . Concentrations of 0 . 5 mg/ml RovA , 0 . 2 mg/ml RovM , and 0 . 5 mg/ml lysozyme were used . Protein biosynthesis of bacterial cultures in exponential or stationary phase was stopped by adding 200 µg/ml chloramphenicol or 50 µg/ml tetracycline . Subsequently , cultures were further incubated at either 25°C or 37°C and samples were taken at indicated time points . Degradation was visualised by Western blotting using a polyclonal anti-RovA or an anti-LacZ antibody as described [20] . Purified Lon protease and RovA were dialysed against the reaction buffer ( 50 mM Tris-HCl pH 8 . 0; 4 mM DTT , 10 mM MgCl2 ) . A concentration of 3 µM Lon protease was mixed with 12 . 5 µM RovA or α-casein , respectively . ATP was supplied at a concentration of 4 mM and for ATP-regeneration , 80 µg/ml ( 20 U ) creatine kinase and 20 mM creatine phosphate were added . Reaction mixtures were separated and incubated at either 25°C or 37°C . After specified time points the reaction was stopped by adding SDS-sample buffer and heat denaturation at 95°C for 5 min . Degradation of RovA and α-casein was visualised by SDS-PAGE and coomassie staining . To test whether an additional component is required for in vitro proteolysis , BL21λDE3 ( lon− ) pLW1 was grown to A600 of 0 . 6 at 25°C and 37°C , resuspended in reaction buffer and lysed using a French Press ( 120 000 psi ) . A concentration of 3 µM purified Lon protease was given to the extracts , with or without addition of ATP and the ATP-regeneration system . Degradation of the RovA protein in the sample was analyzed as described above .
|
Temperature is one of the most crucial environmental signals sensed by pathogens to adjust expression of their virulence factors and host survival programs after entry from a cold external environment into a warm-blooded host . Thermo-induced structural changes in bent or supercoiled DNA or mRNA secondary structures are frequently used to modulate virulence gene transcription or translation . Here we introduce a unique alternative mechanism , in which a central regulator of Yersinia virulence ( RovA ) uses an in-built thermosensor to control its activity in order to modulate virulence gene expression . According to our results , small thermo-induced structural alterations reduce the DNA-binding capacity of the virulence regulator and render the protein more susceptible to proteolytic degradation by ATP-dependent proteases . Amino acids in the vicinity of the DNA-binding region appear to comprise the information required for proteolysis . We therefore postulate a model in which proteolytic degradation is in direct competition with the thermo-sensitive DNA-binding function of the regulator . This regulatory concept constitutes a new example of how microbial pathogens are able to rapidly adjust virulence-associated processes in the course of an infection .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"molecular",
"biology/post-translational",
"regulation",
"of",
"gene",
"expression",
"microbiology/cellular",
"microbiology",
"and",
"pathogenesis"
] |
2009
|
Intrinsic Thermal Sensing Controls Proteolysis of Yersinia Virulence Regulator RovA
|
Many bacteria are able to efficiently bind and take up double-stranded DNA fragments , and the resulting natural transformation shapes bacterial genomes , transmits antibiotic resistance , and allows escape from immune surveillance . The genomes of many competent pathogens show evidence of extensive historical recombination between lineages , but the actual recombination events have not been well characterized . We used DNA from a clinical isolate of Haemophilus influenzae to transform competent cells of a laboratory strain . To identify which of the ∼40 , 000 polymorphic differences had recombined into the genomes of four transformed clones , their genomes and their donor and recipient parents were deep sequenced to high coverage . Each clone was found to contain ∼1000 donor polymorphisms in 3–6 contiguous runs ( 8 . 1±4 . 5 kb in length ) that collectively comprised ∼1–3% of each transformed chromosome . Seven donor-specific insertions and deletions were also acquired as parts of larger donor segments , but the presence of other structural variation flanking 12 of 32 recombination breakpoints suggested that these often disrupt the progress of recombination events . This is the first genome-wide analysis of chromosomes directly transformed with DNA from a divergent genotype , connecting experimental studies of transformation with the high levels of natural genetic variation found in isolates of the same species .
For many bacteria , natural transformation is the dominant mode of genetic transfer between close relatives . These naturally competent bacterial species can actively take up DNA fragments from their surroundings and incorporate it into their chromosomes by homologous recombination [1]–[3] . Like sexual reproduction in eukaryotes , natural transformation moves alleles and loci between related bacterial lineages , allowing pathogens to share antibiotic resistances , antigenic determinants , and other virulence factors [4]–[10] . Population genetic studies have found evidence of pervasive recombination between lineages of human pathogenic bacteria , especially in taxa known to be naturally competent [11]–[12] . However such estimates of recombination are confounded by the other evolutionary forces of mutation and selection , and by the poorly understood demographic histories of the sampled isolates [13]–[15] . Naturally competent bacterial cells bind double-stranded DNA fragments at the cell surface but transport only single strands into the cytoplasm ( Figure 1 ) [1]–[2] . Although several details of DNA uptake differ between Gram-positive and Gram-negative bacteria , in all bacteria the ensuing recombination between donor molecule and recipient chromosome is mediated by RecA homologs and other cytoplasmic proteins that limit DNA degradation and/or facilitate RecA-mediated strand exchange [1] , [16]–[17] . In the laboratory competent cells can take up multiple long DNA fragments , although typically only a fraction of cells in a culture becomes competent [18] . As a consequence selection for transformation at one marker increases the fraction of cells found to be transformed by markers on independent DNA fragments . During natural transformation , the extent to which incoming donor DNAs replace segments of recipient chromosomes is limited by the extent and type of sequence differences between the two , as is the case with other pathways that depend on homologous recombination [19]–[20] . Higher sequence identity between donor DNA and recipient chromosome increases transformation efficiency , while transformation by insertions and deletions is less efficient and requires flanking sequence homology [21]–[23] . The heteroduplex DNA created by strand exchange may be subsequently corrected by mismatch repair ( to either a donor or recipient allele ) , or the uncorrected strands may segregate into daughter cells after DNA replication ( Figure 1 ) [24]–[26] . The genomes of independent isolates of many bacterial species differ in two ways [27]–[30] . First , the 80–95% of two isolates' genome sequences that can be readily aligned differ at about 1–5% of bases . Second , the remaining 5–20% of unalignable DNA consists of structural variation resulting from past insertions , deletions , and more complex events . The finding that species show such high variation in gene content has led to a ‘supragenome hypothesis’ , under which non-essential loci are frequently exchanged between lineages by transformation , potentially enabling rapid adaptation to varying conditions [29]–[31] . Natural genetic variation between bacterial strains has previously been used to characterize transformation at specific selected loci [18] , [32]–[34] , but no transformant has been genotyped across the entire chromosome . To investigate the factors that either promote or constrain the movement of genetic variation between otherwise clonal lineages of bacterial pathogens , we are using the well-characterized natural transformation system of Haemophilus influenzae , combining inexpensive sequencing technology with the availability of complete genome sequences of divergent strains . Here we report a high coverage sequence analysis of four H . influenzae genomes derived by natural transformation of a recipient strain with donor DNA of another strain differing at ∼40 , 000 genetic markers ( ∼2 . 5% of aligned positions and ∼300 indels and other rearrangements ) .
Competent cultures of H . influenzae are reported to contain many non-competent cells [18] . To avoid wastefully sequencing clones derived from non-competent cells , we chose for sequencing clones that had acquired a phenotypic marker in a standard transformation experiment . Before doing this , we confirmed that selection for transformation at one locus does not compromise transformation at distant loci , using donor DNA purified from the multiply marked Rd strain MAP7 [35] to transform our standard laboratory Rd strain Rd-RR ( strains used are listed in Table 1 ) . Figures 2A and B show that the relative transformation frequencies at five loci were not altered by selection for a distant marker , and Figure 2C shows that the transformation frequency of a nalidixic acid resistance marker ( NalR ) did not vary when any of 4 distant markers is used for selection of transformed clones . Accurate mapping of recombination events depends on the density and distribution of sequence differences between the donor and recipient genomes . We chose the clinical isolate 86-028NP as the source of donor DNA because , like Rd , it has been completely sequenced and annotated , and because its genome differs from Rd at about 2 . 4% of their alignable bases , typical for a pair of H . influenzae isolates [30] , [36] . To provide phenotypic markers in 86-028NP for transformant selection , we introduced NovR ( novobiocin resistance ) and NalR alleles from MAP7 by transformation with short PCR fragments . We used this doubly marked strain ( NP-NN ) to investigate how strongly the sequence divergence between NP-NN and Rd-RR limits transformation . Figure 2D shows that the sequence divergence of the NP-NN donor DNA at or near the NovR and NalR alleles reduced transformation efficiency into recipient chromosomes by ∼3-fold , compared to Rd-derived MAP7 donor DNA . To identify recombination events in clones transformed with chromosomal DNA from a divergent strain , we selected four Rd-RR clones that had been transformed with NP-NN chromosomal DNA to either a NovR ( transformants Nov1 and Nov2 ) or a NalR ( transformants Nal1 and Nal2 ) phenotype . Acquisition and processing of sequence data for these clones is described in detail in the Materials and Methods . Briefly , the Illumina GA2 sequencer [37] was used to obtain a high yield of short paired-end sequence reads from genomic DNA of these four transformants ( two individually and all four as a pool ) , and also to individually resequence the genomes of the Rd-RR recipient and NP-NN donor strains as controls ( Table 1 and Table S1 ) . Each set of paired-end reads was separately aligned to each of the two reference genomes ( Rd and 86-028NP [38]-[39] ) using the alignment software BWA [40] , and pileups and consensus base calls at each reference position were generated using the SamTools software package [41] . Since these genomes are <2 Mb , genome coverage per set of reads was high , with median read depths of ∼400 per mapped reference position ( Table S2 and S3 and Figure S1 ) . For each set of reads , the base corresponding to each position in each of the two reference genomes was classified as: ( a ) the same as the reference , ( b ) different from the reference , ( c ) ambiguous , or ( d ) unmapped ( summarized in Tables 2 and 3 ) . Ambiguous positions were those where the specific base at a position could not be confidently identified , likely due to sequencing or read-mapping artifacts ( see Materials and Methods ) . Positions were classed as unmapped if none of the reads aligned , either because the positions were absent from that DNA or unmapped for other reasons . Before transformant data sets were analyzed to identify recombination events , the control sequence reads were used to ( 1 ) identify differences between the published reference genome sequences and the genomes of the donor and recipient strains we used; ( 2 ) confirm the reliability of single-nucleotide variants ( SNVs ) for distinguishing between donor and recipient sequences; and ( 3 ) identify positions that were systematically error-prone , ambiguous , or unmapped in the alignment of reads to references . The three steps ( A , B and C ) of these control analyses are illustrated in Figure 3 . To determine the locations of donor alleles in transformant clones Nov1 and Nal1 , sequence reads were aligned to both reference genomes and each cross-validated SNV position was classified as donor-specific , recipient-specific , or ambiguous . The two clones contained 1 , 133 and 1 , 213 donor-specific SNVs , while nearly all of the remaining cross-validated positions contained recipient-specific SNVs ( Figure 4A and Table 6 ) . As expected for the products of homologous recombination , donor-specific alleles in the transformed clones were found in contiguous runs , which we term donor segments ( Figure 4A , Figure 5 , and Figure S7 ) . The lengths of the 10 donor segments in the Nov1 and Nal1 transformants ranged from 1 . 2 kb to 16 . 6 kb ( Figure S5 and Figure S6 ) . As expected , transformant Nov1 had a donor segment spanning the NovR allele at gyrB ( Segment F ) , and transformant Nal1 had a donor segment spanning the NalR allele at gyrA ( Segment L ) . Nov1 contained five additional donor segments , two separated by only 4 . 5 kb and the other three adjacent to the selected Segment F . Nal1 contained three other widely spaced segments in addition to the selected Segment L; one of these overlapped one of those in the Nov1 transformant by 10 . 9 kb ( Segments A and G , shown expanded in Figure 4C and Figure 5 ) . Additional analysis of these segments is presented below . The ∼400-fold coverage obtained per sequencing lane was much higher than needed , so we tested whether sequencing a pool of genomic DNA from four transformants could accurately identify donor segments without compromising the resolution of recombination breakpoints . Equal amounts of four genomic DNAs were pooled and sequenced ( two transformant clones , Nov2 and Nal2 , and as internal controls , the individually sequenced clones Nov1 and Nal1 ) . The reads from this pool were aligned to the donor and recipient references , and the allele frequencies at each position for each reference were calculated ( Tables 2 and 3 ) . In this analysis donor alleles acquired by transformation of one clone will be given ‘ambiguous’ base assignments , with donor alleles at 25% . These are seen in the allele-frequency histogram in Figure 4D as the large peak centered on 25%; the smaller peak centered on 50% reflects the donor alleles present in two clones . When plotted against chromosome coordinate , the recombination breakpoints of the donor segments in the pool are evident as abrupt transitions of donor-allele frequency ( Figure 4 and Figure S7 ) . For this pool of 4 genomes , donor segments are seen as intervals of contiguous ∼25% or 50% donor allele frequency . As expected , overlapping segments were seen at the selected NovR and NalR alleles , with the NovR allele in the Nov1 and Nov2 transformants and the NalR allele in the Nal1 and Nal2 transformants ( purple diamonds in Figures 4 and Figure S7 ) . The previously identified overlapping segments A and G were also detected ( in Nov1 and Nal1 , respectively ) . The pool contained three more unselected donor segments specific to either Nov2 or Nal2 . Allele-specific PCR was performed to determine which of the two clones these three donor segments were found; Segment M was in Nov2 , while Segments N and O were in Nal2 . While the pooling approach was successful at precisely identifying recombination breakpoints and overlaps between donor segments in the four different clones , the assignment of endpoints to overlapping donor segments and to particular recombinant clones required additional information . The increasing availability and decreasing cost of multiplexed sequencing methods will partially circumvent this problem in the future . In total , we identified 16 donor segments across the four transformants , spanning a total of ∼130 kb and containing 3 , 183 donor-specific SNVs ( Table 6 and Tables S9 and S10 ) . This is 7 . 1% of the Rd genome , or 6 . 0% if overlaps are only counted once . The 16 donor segments had a mean length of 8 . 1±4 . 5 kb , suggesting that transformation of very short DNA fragments is rare ( at least with the high molecular weight donor DNA prep we used ) . The average amount of sequence introduced into each transformed clone ( ∼1 . 8% ) was consistent with the transformation frequencies of individual selectable markers shown in Figure 2 . Although transformation might be expected to preferentially occur at regions with low sequence divergence , the regions participating in recombination had divergences typical of the whole genome ( 2 . 4±0 . 9% vs 2 . 3% ) . A more detailed analysis of sequence divergence in these regions is shown in Figure 6 Notably , the extent of sequence divergence is locally highly variable , ranging from less than 1% to more than 15% within only a few kilobases . However this variation did not appear to affect recombination , since all donor segments contained regions of both high and low divergence , and there were no obvious correlations between recombination breakpoints and extremes of divergence . Notably , recombination was not interrupted even when divergence was as high as 20% ( light green line , 250 bp sliding windows ) . The adjacent locations of many donor segments ( Figure 5 ) likely resulted from disruption of longer transforming DNA fragments rather than independent events . For example the 6 donor segments in Nov1 were found in 2 clusters of 22 and 24 kb . Across all four clones , there were 6 instances of apparent disruptions within longer “recombination tracts” , where adjacent donor segments were separated by relatively short intervals ( <10 kb ) of recipient-specific alleles . The longest is the 5 . 3 kb interval separating segments A and B in transformant Nov1 , and the shortest is the single recipient SNV dividing Segments N and O in transformant Nal2 ( Tables S9 and S10 ) . When the 16 donor segments were treated as 10 clusters , the mean recombination tract length was 14 . 2±8 . 8 kb . Recombination not only brought thousands of donor-specific SNVs into the transformant genomes but introduced several donor-specific insertions and deletions ( Table S11 ) resulting in some donor segments being different lengths than the recipient segments they replaced ( Figure 5 , Tables S9 and S10 ) . In particular , strain Nov1 received two large donor-specific insertions ( 1 . 2 and 2 . 7 kb ) as parts of Segments C and E ( Figure 5 asterisks and Figure 7A ) . These were confirmed by read depth analysis along the two reference sequences ( Figure S8 ) . On the other hand , indels and other structural variants between the donor and recipient chromosomes appear to have blocked progression of strand exchange in several instances ( Figure 5 and Table S12 ) . Of the 32 donor segment breakpoints , 12 are within 5 kb of indel or other structural variation; 6 of these are within 3 of the 6 apparent disruptions described above and thus are likely sites of restoration repair . Indeed , one structural variant gave different outcomes in different recombinants: the 2 . 7 kb donor insertion allele that was acquired by strain Nov1 was not inserted into strain Nov2 , but instead a short segment of recipient sequence interrupts donor segments K and L . Figure 7B illustrates another example of putative restoration repair at an insertional deletion difference between the donor and recipient , as indicated by the interruption between Segments D and E by the recipient insertion allele along with 26 flanking recipient SNVs ( Figure S8 ) .
The plummeting cost of deep sequencing allowed us to characterize the genome-wide consequences of natural transformation , but the ability of this analysis to account for artefacts depended on our high-coverage control sequencing of the donor and recipient genomes . Aligning these control reads to the two reference genome sequences revealed many positions prone to ambiguity or false-positive SNV calls . In the absence of these controls , such artifacts would have mistakenly been interpreted as recombination-induced mutations , since mapping reads to divergent references generated these erroneous variants , while mapping reads to highly similar references did not . The frequency of these artifacts depends not only on nucleotide divergence , but also on the spectrum of structural variation and the complexity of the genome . Analysis of such high-coverage control datasets will be essential for reference-guided assembly approaches that use data with lower coverage , such as that obtained using inexpensive multiplexing methods . H . influenzae's normal environment is the mucosal layer of the human respiratory tract , which contains abundant DNA , much of it high molecular weight like that we used [43] . The broad spectrum of sequence differences between donor and recipient used in these experiments is typical of the natural genetic variation between H . influenzae strains and present in the human host [44]-[45] . However most of the DNA in respiratory mucosa is from human cells and , although bacterial DNA is known to be abundant in biofilms , its fragment sizes and composition in mucus are not known . The short DNA fragments also present in mucus may be taken up more efficiently than long fragments , since H . influenzae cells take up more fragments when fragments are short ( 40 fragments of 120 bp ) [46]–[47] , but the implications for transformation are not clear . Competent cells incubated with short donor DNAs might acquire more donor segments , but short fragments will also be more severely affected by the exonucleolytic degradation that accompanies translocation into the cytoplasm . H . influenzae's preference for DNA containing uptake sequences ( see below ) will also affect both the sources and sizes of the fragments cells take up . Analysis of recombination tracts showed that the four transformants had replaced ∼1–3% of their genomes with 3–6 segments of donor DNA ranging in length from 1 . 2 to 16 . 6 kb . The number of donor segments per transformant agrees well with the ∼3 . 3 fragments found to be taken up per cell in laboratory experiments using long 14 . 4 kb fragments [47] . The lengths of the donor segments we found in H . influenzae are similar to those reported from analysis of naturally occurring recombination tracts observed in Neisseria meningitidis at specific loci[48] , but contrast with the shorter tracts seen for Helicobacter pylori ( ∼0 . 5–3 . 5 kb ) in experiments using DNAs of similar divergence to ours [33]–[34] , [49] . The difference suggests that population genetic models for measuring recombination in nature will require incorporating species-specific estimates of the distribution of recombination tract lengths [15] . The lengths of donor segment found in recombinant chromosomes may underestimate the original lengths of DNA fragments participating in uptake and recombination , because clustering of donor segments suggests that longer incoming DNA fragments are often disrupted before transformation is complete . Similar clustering of donor segments was seen when recombination at a single locus was examined in Helicobacter pylori [33]–[34] . The clustering of H . influenzae donor segments is unlikely to be due to chance , because of the small number of donor segments in each transformant . More probable explanations are that ( 1 ) cytosolic or translocation endonucleases degrade incoming DNAs prior to strand exchange , or ( 2 ) sequence heterology blocks progression of strand exchange , with the heterologous sequences trimmed away by nucleases . Intracellular cleavage of incoming DNA by restriction enzymes has been proposed for competent Helicobacter pylori [49] , but this is problematic because , in both H . pylori and H . influenzae , only single DNA strands are thought to enter the cytoplasm [50]–[51] . Although McKane and Milkman have shown that restriction can create clustered recombination tracts in E . coli transduction experiments [52] , the single strands brought into the cytoplasm by transformation are not normally substrates for restriction enzymes , and donor strands recombined into the chromosome will be protected from restriction by the methylation of the base-paired recipient strands . The effect of restriction in H . pylori may instead be due to the accumulation of extracellular restriction enzymes during the long transformation protocol [34] . Similar accumulation might be a transformation-limiting factor for many species that normally live in mixed-species biofilms , whenever environmental DNA encounters restriction enzymes derived from other strains or species . We found no evidence that recombination preferentially occurred in regions of lower nucleotide divergence than the genome-wide average . Instead , sequence divergence varied on a scale much shorter than the donor segments , with most segments spanning local regions of both high and low divergence ( Figure 6 ) . Although strand exchange of short fragments is known to be dramatically inhibited by sequence divergence of >10% [19]–[20] , most donor segments contained one or more regions with >15% divergence . This suggests that , although strand exchange may initiate between regions of high sequence identity , it readily extends into and through regions with many mismatches . Measuring the effect of divergence on recombination break points and interruptions will require sequencing many more recombinants . Effects of structural variation on recombination were evident even with this small sample size , as heterologous sequences were much more common at donor-segment breakpoints than expected from their abundance in the recombining genomes , e . g . between the clustered segments C and D , D and E , and K and L ( Figure 5 and 6B ) . This is consistent with previous genetic experiments showing that insertions and deletions transform at much lower rates than do substitutions [22] and may be due to inhibition of strand exchange or to subsequent excision of heteroduplex from recombination intermediates by a mismatch correction mechanism . However , at other sites the donor versions of structural variation were acquired as parts of longer donor segments , showing that such accessory loci can indeed readily move by natural transformation . Other factors could have influenced the transformation events we observed: ( 1 ) H . influenzae's strong preference for DNA fragments containing uptake signal sequences ( USS ) biases transformation to USS-containing fragments [53]–[54] . However USSs are unlikely to have been a factor in these experiments , since they occur at a high density in the genome ( ∼1/kb ) and the donor DNA fragments used typically contained dozens of USSs . ( 2 ) Segregation of uncorrected heteroduplex at the first post-transformation cell division could cause the extent of strand replacement in individual competent cells to be underestimated by up to 2-fold ( Figure 1 ) . We do not know the extent of heteroduplex correction at these or other independently transforming sites , nor how recombination tracts are distributed between the two strands of the originally transformed chromosome . Although the relatively short putative restoration repair events observed in this study might suggest that heteroduplex correction only act on parts of larger heteroduplex recombination products , other repair events might have completely removed shorter segments of donor DNA . Because clones chosen for sequencing had acquired one of two antibiotic resistance alleles from the donor , we were able to examine overlapping recombination events at each of these loci , detecting striking differences at the NovR locus . The selection for NovR and NalR also showed that unselected events are common , as 58% of donor alleles were found in segments distant from the selected loci . On the other hand , the 11 kb overlap between the unselected donor segments A and G was unexpected given the transformation frequencies of single markers ( Figures 4C and 5 ) , and a sufficiently large dataset might identify a transformation hotspot , as has recently been found in Neisseria meningitidis [55] . The overlapping sequences do not have any obvious distinguishing features: divergence between donor and recipient is typical , no virulence genes have been annotated , and density of USSs is slightly lower than the genome average . In addition to the selected antibiotic resistance alleles , the recombination events characterized here had the potential to significantly change the cell's biology , both by introducing new genes and by creating new genetic combinations by homologous recombination both between and within genes alleles . In particular , the Segment E insertion contains four donor-specific ORFs , one encoding a predicted transposase , and the Segment C insertion contains the LPS biosynthesis gene lic2C ( between infA and ksgA ) . Each recombinant clone also acquired donor-specific versions of 20–50 shared genes , and these may have altered phenotype both directly and because of new interactions with recipient alleles at unrecombined loci . Recombination breakpoints that were not at structural variation usually fell within genes ( Figure 5 ) and , because of the high level of sequence variation , these are likely to have created novel recombinant alleles potentially with substantial changes to function . The results presented above considered only four recombinant clones , but continuing advances in DNA sequencing technology and bioinformatics methods will allow characterization of many more recombinants under a variety of experimental conditions and using different donor DNAs . This will help bridge experimental studies of transformation with the population genomic approaches used to detect recombination between bacterial lineages in nature . The comprehensive identification of donor segments in a large set of experimentally transformed clones will also provide a novel resource for the genetic mapping of phenotypes that differ between the donor and recipient strains , such as their dramatic natural variation in transformability [56] , as well as natural variation in pathogenesis-related traits like serum-resistance [57]–[58] .
Standard protocols were used for growth and manipulation of H . influenzae , preparation and storage of competent cultures , and purification of high molecular weight chromosomal DNA from overnight cultures [35] , [59] . Briefly , cells were grown in the rich medium sBHI and made competent by transfer of log-phase cultures to the starvation medium M-IV for 100 minutes before transformation experiments or storage in 15% glycerol at −80°C . The H . influenzae recipient strain Rd-RR ( RR722 ) was obtained from H . O . Smith in 1988 , and is separated by ∼10 passages ( ∼500 generations ) from the KW20 Rd strain sequenced in 1995 [38] ( NCBI Taxonomy ID: 71421 ) . The donor strain NP-NN ( RR3131; resistant to novobiocin and nalidixic acid ( NovR and NalR ) ) was derived from the clinical isolate 86-028NP [39] ( RR1350 , gift of Richard Moxon in 2006 , NCBI Taxonomy ID: 281310 ) ; it is separated from the sequenced 86-028NP strain by ∼5 passages ( ∼250 generations ) . NP-NN was constructed by PCR-mediated transformation of 86-028NP with NovR and NalR amplicons of gyrB and gyrA , respectively , ( both caused by point mutations ) . For the NovR allele of gyrB , a 2 . 6 kb fragment ( Rd coordinates 585 , 533 to 588 , 096 bp ) was amplified from MAP7 ( RR666 ) chromosomal DNA . For the NalR allele of gyrA , a 2 . 8 kb fragment ( Rd coordinates 1 , 341 , 635 to 1 , 344 , 397 ) was amplified . Transformation experiments used 2 µg of chromosomal DNA per 1 ml of M-IV competent culture ( ∼109 cells ) for a final DNA concentration of ∼1 genome equivalent per cell . Cells were incubated with DNA at 37°C for 20 min , diluted 1∶5 into sBHI , and incubated at 37°C for 80 min to allow expression of donor resistance alleles before dilution and plating to sBHI agar ± antibiotics [35] . Experiments were performed in triplicate from frozen aliquots of competent cultures prepared on three separate occasions . No DNA controls were performed in parallel , and antibiotic resistant colonies were not observed ( limit of detection typically ∼10−9 resistant colonies/CFU ) . Cells from defrosted aliquots were pelleted and resuspended in fresh MIV before transformation . Two NovR and two NalR transformant colonies ( Nov1 , Nov2 , Nal1 , and Nal2; Table 1 ) were randomly selected for sequencing from a single experiment that used Rd-RR competent cells and NP-NN chromosomal DNA fragments ( size range ∼20–100 kb ) . The reference sequences for Rd and 86-028NP were compared using the Mauve whole-genome alignment software [42] . The complete genome sequences were aligned twice , once with Rd as the query and once with 86-028NP as the query . SNVs were then extracted using Mauve's “Export SNPs…” function . The few identified SNVs that were inconsistent between the two independent whole-genome alignments were excluded . The two resulting files provided positions of each SNV in each genome , ordered against one or the other reference . Chromosomal DNA was sheared by nebulization , and converted into paired-end sequencing libraries with an insert size of ∼100–300 bp , as previously described [37] . About 10 million paired-end sequences of 42 bases were obtained from each library on individual lanes of an Illumina GA2 flow cell ( Table S1 ) . Raw data was processed using Illumina Pipeline Version 1 . 4 , and all paired-end reads that passed standard Illumina quality control filters were used for analysis ( i . e . those in the “ . sequence . txt” file ) . The raw sequence reads for each DNA sample ( Table S1 ) were deposited at the NCBI short-read archive under project accession SRP003474 . The Rd ( KW20 ) and 86-028NP complete genome sequences ( NCBI genome accessions NC_000907 and NC_007416 ) were each used as references for read alignment , with the BWA algorithm ( version 0 . 5 . 5 ) [40] set to highly sensitive alignment parameters ( bwa aln -n 8 -o 3 -e 3 -l 20 -R 100000; bwa sampe -a 400 -o 1000000 ) . While this generates some spurious mapping artifacts , it ensures that reads will map to both references when possible , even where there is high divergence . A combination of two criteria was used to identify differences between sequence reads and their references and to flag positions with ambiguous base identity . The first method used the SamTools ( version 1 . 12a ) [41] consensus caller , which either assigns positions a standard A , C , G or T base or tags them as ambiguous . Reference positions missing from the SamTools consensus were treated as unmapped positions presumably within or near deletions . The second method used direct calculation of the frequency of each base at each reference position . This used a Perl script obtained from Galaxy [60] ( pileup_parser . pl , parameter settings: 3 9 10 8 40 20 “No” “No” 2 ) to parse the pileup output from SamTools , and provided the count of each non-reference base call at each position . Parsed pileup files were subsequently analyzed using custom scripts written in the R statistical programming language [61] . Plots including gene maps were made with the assistance of the ‘genoPlotR’ package [62] . Transformant sequence reads were analyzed as above . Recombination events were identified in the transformed clones by classifying the positions of cross-validated SNVs as donor , recipient , or ambiguous . Donor segments were defined as contiguous runs of donor-specific SNVs , uninterrupted by recipient-specific SNVs ( ambiguous cross-validated SNV positions were ignored ) . Individual donor segments breakpoints were defined by the positions of their outermost donor-specific alleles . Donor segments were then manually inspected using the Integrated Genomics Viewer [63] to validate the donor segment breakpoint locations . For the pooled sample of four transformed clones ( RR3135-RR3138 ) , donor-specific allele frequencies were determined at each cross-validated SNV position . Non-overlapping donor segments were unambiguously identified as contiguous runs of SNV positions with ∼25% donor-specific alleles . Overlapping donor segments ( contiguous SNV positions with ∼50% donor-specific alleles ) were disambiguated by comparison with the segments identified in RR3135 and RR3137 . Segments unique to either RR3137 or RR3138 were disambiguated using allele-specific PCR; two primer pairs were designed that contained several SNVs that distinguished Rd and NP alleles ( Table S13 ) . Positions that were unmapped by reads in the reciprocal alignments ( but mapped in self-alignments ) were used as markers of indel differences and other structural variation between donor and recipient , and the donor segment intervals were examined for read coverage at positions unmapped by either reciprocal alignment . Indel differences flanking the observed donor segments were also tabulated . Manual inspection of read alignments to both references used the Integrative Genome Viewer , and the “ . rdiff” and “ . qdiff” output from the dnadiff utility of Mummer [64] was used to cross-validate . GenomeMatcher [65] was used to view annotated sequence alignments at transforming and flanking structural variation to identify affected loci .
|
The ability of bacteria to acquire genetic information from their relatives—called natural competence—poses a major health risk , since recombination between pathogenic bacterial lineages can help bacteria develop resistance to antibiotics and adapt to host defenses . In this study we transformed competent cells of the human pathogen Haemophilus influenzae with genomic DNA from a divergent clinical isolate and used deep sequencing to identify the recombination events in four transformed chromosomes . The results show that transformation of single competent cells is more extensive than expected , and suggests that transformation can be used as a tool to map traits that vary between clinical isolates .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"genetics",
"biology",
"genomics",
"microbiology",
"genetics",
"and",
"genomics"
] |
2011
|
Transformation of Natural Genetic Variation into Haemophilus Influenzae Genomes
|
As an important vector of dengue and Zika , Aedes albopictus has been the fastest spreading invasive mosquitoes in the world over the last 3–4 decades . Cold tolerance is important for survival and expansion of insects . Ae . albopictus adults are generally considered to be cold-intolerant that cannot survive at subzero temperature . However , we found that Ae . albopictus could survive for several hours’ exposure to -9 to -19 oC so long as it was exposed with water . Median lethal time ( LT50 ) of Ae . albopictus exposed to -15 and -19 oC with water increased by more than 100 times compared to those exposed to the same subzero temperature without water . This phenomenon also existed in adult Aedes aegypti and Culex quinquefasciatus . Ae . albopictus female adults which exposed to low subzero temperature at -9 oC with water had similar longevity and reproductive capacity to those of females without cold exposure . Cold exposure after a blood meal also have no detrimental impact on survival capacity of female adult Ae . albopictus compared with those cold exposed without a blood meal . Moreover , our results showed that rapid cold hardening ( RCH ) was induced in Ae . albopictus during exposing to low subzero temperature with water . Both the RCH and the relative high subzero temperature of water immediate after cold exposure might provide this strong protection against low subzero temperature . The molecular basis of water-induced protection for Ae . albopictus might refer to the increased glycerol during cold exposure , as well as the increased glucose and hsp70 during recovery from cold exposure . Our results suggested that the water-induced strong protection against acute decrease of air temperature for adult mosquitoes might be important for the survival and rapid expansion of Ae . albopictus .
Cold tolerance or cold hardiness , the ability of an insect to survive at low temperature , is important in defining the distribution and survival of insects . There are two different cold hardening in insects at present . One is accomplished by long term ( weeks or months ) cold acclimatization to overwinter that occurs in an inactive or diapausing stage; the other called rapid cold hardening ( RCH ) which is accomplished by a brief exposure ( minutes or hours ) to low temperature that occurs even in feeding and reproductive stages [1–4] . As an efficient ability utilized by insects to survival in environment with rapid and unexpected changes in temperature , RCH has been found in numerous insect species belonging to different orders including Diptera [2 , 4–7] . Aedes albopictus is an epidemiologically important vector for several arboviruses such as dengue , yellow fever , zika , and chikungunya . During the last 3–4 decades , Ae . albopictus has spread from native Asian area to all continents except Antarctica , becoming the most invasive mosquitoes which imposed extensive public health threat to human beings throughout the world [8–12] . During the spread of this species , a spatial expansion to cooler climate areas has also been reported and the ability to rapidly produce low temperature phenotypes has been considered as an important factor for the successful establishment in these cooler habitats [13] . Therefore , cold hardiness is a key trait for the distribution of this species and strong resistance to cold temperature provides more chances for efficient invasiveness to colder zones . Adult Ae . albopictus are generally considered to be cold-intolerant that can not survive at subzero temperature and the only life stage that can cope with this low temperature is its eggs [13 , 14] . Thus , the studies on the cold hardiness of this mosquito have been focused on eggs [13–19] . Moreover , the RCH has not been reported in adult and eggs of this species yet . In our current study , we found that adult Ae . albopictus could survive for several hours’ exposure under -10 oC when transferred from room temperature to low subzero temperature with water , which was also found in adult Aedes aegypti and Culex quinquefasciatus . Median lethal time ( LT50 ) of adult mosquitoes was increased by nearly 100 times compared to those transferred directly to subzero temperature lower than -10 oC without water . This indicated that RCH was also existed in adult mosquitoes and that water in nature might provide strong protection for adult mosquitoes against sudden drop in air temperature that often happens in early spring and winter , and late autumn [2 , 4 , 6 , 7] . The aim of this study is to compare the cold hardiness of adult Ae . albopictus exposed to low subzero temperature with water with those exposed directly , analyze the possible molecular mechanism for this cold hardiness and determine the impact of exposure to low subzero temperature with water on fitness costs of adult female Ae . albopictus .
Mosquitoes used in this study including Ae . albopictus , Ae . aegypti and Cx . quinquefasciatus were all established for several years in our laboratory with Guangdong origin . All mosquitoes were reared in a climate-controlled room at 28 ± 1°C and 80 ± 5% relative humidity with a 12:12-hour ( light: dark ) photoperiod . Adult mosquitoes were provided with 10% glucose solution . To evaluate the survivorship , 3 groups of 20 adult mosquitoes with 3- to 5-day-old were immobilized by CO2 and then transferred to a disposable 100ml-plastic cup with ( treatment ) or without ( control ) dechlorinated tap water ( 50 mL ) . A plastic lid was used to cover with cup for preventing escape of mosquitoes during experiment . These adults were then allowed to recover from anaesthetization at room temperature for 1 h . Subsequently , the cups of mosquitoes with water were exposed to low subzero temperatures at -9 , -15 and -19 oC for 3 to 8 h , and the cups of mosquitoes without water were exposed to the same temperatures for 3 to 30 min . Twenty adults ( one cup ) were removed from each temperature at a 1 h interval for cups with water and 1- to 5-min interval for cups without water until 100% mortality were attained . The exposed mosquitoes were then transferred to 650ml-plastic cages with a piece of wet filter paper to keep humidity and maintained at normal climate-controlled room . The mosquito survival was recorded 24 h after cold exposure and survival was defined as the ability of righting themselves and flies [20] . To evaluate the impact of cold exposed to low subzero temperature on fitness costs of female adult Ae . albopictus , 3 groups of 30 adult mosquitoes with 3- to 5-day-old were cold exposed to -9 oC with water for 3 to 5 h as described above . Subsequently , the exposed mosquitoes were transferred to cages , maintained under normal climate-controlled room and provided with 10% glucose solution . Control groups of 90 female mosquitoes without cold exposure were also transferred to cages and maintained under normal condition . Three days after recovery from cold exposure dead mosquitoes were discarded and survived mosquitoes that have starved for 24 h were blood fed on mice for 30 min . The engorged mosquitoes were then counted and transferred to new cages , and were aspirated into individual 50 mL Corning tubes 2 days post-blood meal ( PBM ) with bottom lining of moist filter paper supported by water-soaked cotton [21] . Two days after oviposition , mosquitoes were aspirated out and killed by cold , eggs were removed out for maturation for 3 days and then counted . After maturation , eggs on filter paper were immersed in 50 mL water in a 100ml-plastic cup and egg hatch rates were determined by counting the number of hatched second instar larvae [22] . Adult lifespan of female Ae . albopictus after cold exposure to -9 oC with water for 3 h were also monitored and compared with control group without cold exposure . Cold treatments were conducted as above and dead mosquitoes were removed at 24 h after recovery . Survived mosquitoes were reared under normal condition and dead mosquitoes were counted and removed daily for a month . To evaluate the impact of cold exposure on blood-fed Ae . albopictus , engorged mosquitoes were collected and transferred to a new cage . Three groups of 30 adult mosquitoes were selected at 24 and 48 h PBM and exposed to -9 oC with water for 3 h . Mosquitoes without a blood meal were also selected and cold exposed at the meanwhile . Cold exposed mosquitoes were maintained in cages under normal condition and survivals were recorded 24 h after recovery . Two days after blood meal , mosquitoes of cold exposure at 24 h PBM were reared individually , and mosquitoes of cold exposure at 48 h PBM were reared as a pool , egg numbers and egg hatch rates were determined as above . After transferring water from room temperature into low subzero temperature , the dynamics of water temperature was monitored by a mini-thermometer ( Testo 175 H1 , Lenzkirch , Germany ) . The temperature probe was wrapped by plastic film and immersed into 50 mL water in a disposable 100ml-plastic cup and then transferred to different low subzero temperature for 12 h . Temperature was detected and recorded at a 5-min interval . Adult mosquitoes were cold exposed as above and collected at 0 , 1 , 2 and 4 h after recovery from cold treatment . Six adults were pooled in a single replicate and five replicate biological assays were performed . These sampled mosquitoes were homogenated in 300 μL distilled deionized water . Two hundred microlitres of the homogenates were used for RNA extraction and another 100 μL were filtered at room temperature through a spin filter ( Pall , nanosep 10k Omega , NY ) at 12 , 000 rpm for 15 min . For glycerol analysis , filtered homogenates were diluted 1:100 v/v with distilled deionized water . Next , 10 μL of the diluted homogenates were incubated with 100 μL Master Reaction Mix ( Sigma-Aldrich , MAK117 , USA ) for 20 min at room temperature . Absorbance was measured at 570 nm and glycerol contents were calculated from standard curve . Glucose levels were determined by using the Glucose ( GO ) Assay Kit ( Sigma-Aldrich , GAGO-20 , USA ) according to the manufacturer’s protocol with minor modifications . The above filtered homogenates were diluted 1:5 v/v with distilled deionized water and 50 μL of these diluted samples were then incubated with 100 μL Assay Reagent for 30 min at 37 oC . After the incubation , 100 μL of 12N H2SO4 were added to stop the reaction . Then , the absorbance was measured at 540 nm and glucose contents were calculated from standard curve . Total RNA was extracted from samples collected above using TRIzol Reagent ( Invitrogen , Carlsbad , CA ) and the first strand cDNA was synthesized using HiScript II Q SuperMix for qPCR ( +dDNA wiper ) ( Vazyme , Nanjing , China ) following the manufacturer's protocol . Relative expression level of Hsp70 mRNA was performed by quantitative real-time PCR ( qPCR ) on the LightCycler96 Detection System ( Roche , Mannheim , Germany ) using TB Green Premix Ex Taq II ( Tli RNaseH Plus ) ( TaKaRa , Otsu Shiga , Japan ) . The primer sequences for Hsp70 were forward ( 5’-TACCAACGGCGACACTCAC-3’ ) and reverse ( 5’-TTGCGGATGTCCTTACCCT-3’ ) . Each reaction consisted 0 . 5 μL of cDNA , forward and reverse primer ( 10 μM ) , 10 μL of TB Green Premix Ex Taq II ( 2× ) , and 8 . 5 μL of distilled deionized water to a final volume of 20 μL . The qPCR program was 95 oC for 30 sec , then 40 cycles of 95°C for 5 sec and 60°C for 30 sec followed by a melt-curve analysis . Ae . albopictus rpS7 was used as internal control and the relative Hsp70 expression of cold exposed samples were calibrated by samples collected at room temperature that without cold treatment . The relative expression levels of Hsp70 were determined by using the 2-△△CT calculation method [23] . All data were analyzed by using SPSS statistics 19 . Comparison of LT50 , egg numbers and egg hatch rates per female ( except egg hatch rate of mosquitoes cold exposed at 48h PBM ) between different groups were conducted using Student’s t-test . Comparison of the percent of mosquitoes imbided a blood meal between cold exposed and non-exposed groups , of the survival of mosquitoes cold exposed at 24 h PBM with those without blood meal and of the egg hatch rate of mosquitoes cold exposed at 48 h PBM with those maintained at room temperature were performed using chi-square test . Comparisons of glycerol , glucose and hsp70 mRNA levels between groups were assessed using ANOVA followed by Tukey’s multiple comparison . The GenBank accession number of hsp70 mentioned in the text is JN132155 . 1 .
We found that adult Ae . albopictus could survive for several hours’ exposure to subzero temperature even below -10 oC when it was transferred from room temperature to the low temperature with water . Subsequently , we analyzed the survival of adult female Ae . albopictus that transferred to different low subzero temperatures from -9 oC to nearly -20 oC with water and compared to the mosquitoes that directly transferred to the same temperatures without water . The results showed that when exposed adult mosquitoes to low subzero temperature with water the cold tolerance of these mosquitoes were strongly increased compared to those directly exposed without water ( Fig 1 ) . About 70 to 45% of these mosquitoes survived a 5 to 2 hours’ exposure to -9 to -19 oC when exposed with water . However , mosquitoes that directly transferred to the same low temperature without water were found to 100% mortality within 3 to 30 min . The LT50 of these adult Ae . albopictus were about 292 , 184 and 106 min for those exposed to -9 , -15 and -19 oC with water , respectively , and these were 13 . 6 , 145 . 1 and 108 . 7 times longer than those exposed to the same low temperature without water ( Table 1 ) . Moreover , we did the same analysis on adult male Ae . albopictus , female Ae . aegypti and Cx . quinquefasciatus to see if the phenomena also existed in male and other mosquitoes . The results suggested that when transferred these adult mosquitoes from room temperature to low subzero temperature ( -15 oC ) with water the cold tolerance were also significantly increased compared to those transferred to the same low temperature without water ( Fig 2 and Table 2 ) . When transferring from room temperature to low subzero temperature with water adult mosquitoes will fall on the surface of water after anaesthetized by cold , we detected the change in temperature of water during this process ( Fig 3 ) . We found that when transferring from room temperature to -9 , -15 and -19 oC , the water temperature would reach subzero and iced within 55 , 45 and 25 min , respectively . After that the temperature of ice kept at a relative high level of -2 , -3 and -5°C for 6 , 3 and 2 h , respectively , and then rapidly decreased to the level equal to the low ambient air temperature . To evaluate whether it is the relative high temperature of ice just after dropping under subzero that protected adult mosquitoes from low temperature , we analyzed the survival of adult female Ae . albopictus when exposed to these relative high subzero temperature without water ( Fig 4 and Table 1 in parentheses ) . The results indicated that when exposed to -2 to -5°C , 100% mortality of these adult mosquitoes were reached within 60 to 100 min and that the LT50 were about 85 , 57 and 38 min for those exposed to -2 , -3 and -5 oC , respectively . Even subtracting the time needed to cool water from room temperature to the relative high subzero temperature , the LT50 of adult female Ae . albopictus transferred to -9 to -19 oC with water still were 2 . 7 to 2 . 0 times longer than those directly exposed to -2 to -5°C without water ( Table 1 ) . These results indicated that when exposed adult mosquitoes to low subzero temperature with water a RCH response was induced during this process . Since adult Ae . albopictus could survive several hours’ exposure to low subzero temperature with water , it is important to know the effects of cold exposure on the bite behavior and reproductive capacity of female adults . Our results showed that after exposure to -9°C for 3 and 5 h with water , there still were 70 . 2% and 56 . 7% of mosquitoes that successfully had a blood meal on mice , respectively , and there was no significant difference between cold exposed and non-exposed mosquitoes ( Fig 5A ) . The fecundity ( egg numbers per female ) and egg hatch rates between cold exposed and non-exposed mosquitoes also had no significant difference except for the fecundity of mosquitoes exposed to -9°C for 5 h ( Fig 5B and 5C ) . Although the fecundity of these mosquitoes was significantly lower than the mosquitoes without cold exposure , they still could lay about 41 eggs per female after 5 hours’ exposure to -9°C with water . Meanwhile , we found that blood meal had no impact on cold tolerance of adult female mosquitoes compared to mosquitoes without blood meal . There were no significant difference on survival capacity between blood-fed and non blood-fed mosquitoes after 3 hours’ exposure to -9°C with water at both 24 and 48h PBM ( Fig 6A ) , and these mosquitoes could still lay viable eggs ( Fig 6B and 6C ) . Indeed , the egg hatch rate of mosquitoes cold exposed at 48h PBM was significantly higher than mosquitoes without cold exposure after blood meal ( Fig 6C ) . Moreover , the lifespan of adult female Ae . albopictus was also compared between cold exposure and no exposure , and no significant difference was observed ( S1 Fig ) . Since the cold tolerance was significantly increased when exposed adult mosquitoes to low subzero temperature with water , we analyzed the levels of two important cryoprotectants glycerol and glucose in whole body of adult Ae . albopictus that exposed to -15 oC with water for 2 . 7 h and compared to those exposed to -3 and -15 oC directly for 1 . 5 h and 3 min , respectively . The duration of time were chose because at this duration mosquitoes exposed to -3 and -15 oC directly were 100% mortality but mosquitoes exposed to -15 oC with water were less than 30% mortality at the same duration of time as those exposed to -3 oC directly after subtracting the time needed to cool water from room temperature to -3 oC . The results showed that the glycerol level of mosquitoes exposed to low temperature with water was 6 . 2 μmol per mosquito just after recovery from cold exposure and this was significantly higher than those directly exposed without water and those maintained at room temperature ( Fig 7A ) . Another , the glycerol level of mosquitoes exposed to low temperature with water was significantly decreased at 1 h after recovery from cold exposure . However , the glycerol levels of mosquitoes exposed to subzero temperature without water had no difference with time after recovery from cold exposure . The glucose levels of mosquitoes exposed to -15 oC without water were significantly increased with time after recovery from cold exposure ( Fig 7B ) . While those exposed to -3 oC directly and to -15 oC with water had no difference in glucose levels with time after recovery . However , the glucose level of mosquitoes exposed to -15 oC with water was significantly higher and lower than those exposed to -3 and -15 oC directly at 4 h after recovery from cold exposure , respectively . These results implied that glycerol and glucose might play important role in water-induced RCH of adult mosquitoes but at different stage of cold exposure . Because the upregulation of heat shock protein 70 ( Hsp70 ) have been reported in several insect species in response to cold temperature [24–27] , we wondered if this protein was also involved in mosquitoes to cope with cold temperature . We found that Hsp70 expression gradually increased with time after recovery from cold exposure of mosquitoes exposed to -3 oC directly and to -15 oC with water and the expression levels were the highest at 4 h after recovery ( Fig 7C ) . The Hsp70 expression of mosquitoes exposed to -15 oC without water was significantly increased at 1 h after recovery from cold exposure and maintained at these high levels till 4 h after recovery and the expression levels from 1 to 4 h after recovery of these mosquitoes were significantly higher than that exposed to -3 oC directly and to -15 oC with water . The results demonstrated that the degree of cold shock of mosquitoes exposed to -15 oC with water was similar to those exposed to -3 oC directly and both of them were significantly lesser than those exposed to -15 oC without water .
Adult Ae . albopictus has long been recognized as freeze-intolerant that can’t cope with subzero temperature . Most of the studies on cold hardiness of this species was focused mainly on eggs , which were the only life stage that can survive under subzero temperature as know so far [13–19] . However , in our current study , we found that when exposed adult Ae . albopictus to low subzero temperature ( -9 to -19 oC ) with water it could survive several hours . Moreover , LT50 of these adult mosquitoes were increased by 13 . 6 to more than 100 fold changes when compared with the counterpart exposed without water . In consistent , this phenomenon also existed in Ae . aegypti and Cx . quinquefasciatus . Cold tolerance or cold hardiness is important for the distribution and survival of insects . Over the last 3–4 decades , Ae . albopictus has spread from its native Asian area to all continents except Antarctica [8 , 9 , 12] . Such widespread distribution of Ae . albopictus implied the strong cold hardiness of this species and that they might have more chance to experience sharp decrease of air temperature than other local mosquitoes . Thus far , studies about the cold hardiness of Ae . albopictus has been focused on egg stage and previous studies indicated that they overwintered predominantly through diapause eggs [16 , 17 , 28–30] . A previous study showed that diapause eggs from Ae . albopictus could only survive for 1 hour under -12 oC while non-diapause eggs could survive for 4 hours . In addition , neither Ae . albopictus nor Ae . aegypti eggs could be hatched after exposure to -15 oC [19] . However , our results showed that both the adult of Ae . albopictus and Ae . aegypti could survive under -15 oC for more than 3h-exposure ( Figs 1 and 2 ) . This indicated that the cold hardiness of adult mosquitoes was even stronger than eggs when exposed with water . The cold hardiness of Ae . albopictus eggs was also highly correlated with the origin . Eggs from northern were more cold-hardy than those from southern , while eggs from tropical Ae . albopictus are much more susceptible to low temperature than those from temperate counterpart [15 , 16 , 18 , 19] . Moreover , similar comparisons of Ae . albopictus larvae from different regions have been conducted at low temperature above zero [31] . So far as we knew , only adult Culex pipiens could tolerate for several to tens of hours’ exposure to subzero temperature ( -5 oC ) [20] . Interestingly , our results showed that the cold hardiness of adult Ae . albopictus , Ae . aegypti and Cx . quinquefasciatus could be significantly enhanced in the presence of water . When exposed to -15 oC with water these adult mosquitoes can survive for several hours more . This is the first report that adult Aedes mosquitoes cold also cope with low subzero temperature even below -10 oC so long as there are waters when these mosquitoes exposed to this low temperature . Meanwhile , our results suggested that water-induced enhancement of cold hardiness might be a universal phenomenon in adult mosquitoes . This study showed that the relative high subzero temperature of water immediate after transferring to low subzero temperature just provided partial protection for adult Ae . albopictus against low subzero temperature ( Table 1 ) . Previous studies showed that RCH of arthropods could be induced through gradual cooling from 0 . 1 to 1°C/min [4 , 6] . We found that when transferred from room temperature to -9 to -19 oC , the cooling rate of water were about 0 . 42 to 0 . 95 oC/min till reaching the relatively high subzero temperature ( Fig 3 ) . These results demonstrated that RCH might be induced in these adult mosquitoes and provided partial protection during the process of cold exposure . Considering water is common in nature , our results suggested that RCH of adult mosquitoes induced by waters in nature might provide strong protection against acute decrease of air temperature to low subzero that would be lethal , which often happens in early spring and winter , and late autumn [2 , 4 , 7] . Ae . albopictus overwintered predominantly through diapause eggs , nevertheless , adult Ae . albopictus was also found occasionally during winter season [29 , 32 , 33] . This indicates these adults might experience the low subzero temperature sometimes during their lifetime like eggs do . Ae . albopictus has been considered to be the most invasive mosquito species worldwide and to be passively spread over long distance principally through the transportation of eggs by global shipments of used tires and other artificial containers [34–38] . In recent years , however , studies reported that , over a long distance , adult mosquitoes including Aedes species could be transported by aircraft [39–44] , and , at a more regional level , adult Ae . albopictus are frequently transported by ground vehicles like cars [10 , 37 , 38] . These also pose a threat of experiencing low subzero temperature to the adult mosquitoes , especially those be transported to cooler climate areas . In the light of situations mentioned above , it is not known until now how the adult mosquitoes can cope with low subzero temperature in nature . In this study , we found that adult Ae . albopictus could cope with low subzero temperature even below -10°C in the presence of water and that after exposure to low subzero temperature for several hours the adult mosquitoes could still bite , lay eggs , and eggs could hatch to larvae ( Fig 5 ) . In addition , after a blood meal female mosquitoes must find a micro-habitat with water to lay eggs . We found that , after cold exposure to low subzero temperature with water , blood-fed Ae . albopictus could still lay viable eggs ( Fig 6 ) . This implies that water in nature not only provide a micro-habitat for mosquitoes’ egg-laying but also can act as a shelter against acute decrease of air temperature . In a word , our results indicated that the water-conferred strong protection against low subzero temperature might be an important means for adult Ae . albopictus to survive the lethal low temperature and therefore might be important for the expansion of this species to cooler areas . Glycerol is the most commonly produced cryoprotectant for insects to cope with freeze damage [45 , 46] . High accumulation in haemolymph and tissues was important for overwintering survival of insects while lower glycerol content often resulted in higher overwinter mortality [46–49] . Furthermore , the accumulation of glycerol was also highly correlated with survival of some insects in RCH [1 , 3 , 50–52] . In this study , we found that the glycerol levels of the adult Ae . albopictus transferred from room temperature to -15 oC with water was significantly higher than those exposed to -3 and -15 oC without water and those maintained at room temperature at 0h after recovery from cold exposure and then significantly decreased to normal level ( Fig 7A ) . Because freeze damage might happen when adult mosquitoes anaesthetizing on -3 oC ice and underneath -15 oC air for a while , our study indicated that accumulated glycerol in adult Ae . albopictus during being exposed to -15 oC with water may confer strong protection to freeze damage and contribute to RCH induced by water . Sugars are also important cryoprotectants in insects to eliminate or minimize freeze damage [45] . There were studies that the levels of glucose in some insects were increased in response to RCH or cold stress [50 , 53–55] . Our results showed that the glucose level of adult Ae . albopictus exposed to -15 oC with water was significantly higher than those exposed to -3°C and lower than those exposed to -15 oC without water at 4 hours after recovery from cold exposure ( Fig 7B ) . This results indicated that the accumulation of glucose in adult Ae . albopictus was important during recovery from cold exposure but not in the process of cold exposure . Moreover , we found that glucose levels went through significant changes with time-dependent manner during recovery from cold exposure to -15 oC without water . Heat shock protein 70 ( Hsp70 ) also played an important role in cold hardiness of overwintering and cold stress of insects [27 , 56 , 57] . Our results showed that the expression of Hsp70 in adults Ae . albopictus , exposed to -3 or -15 oC with or without water , were all significantly up-regulated during recovery from cold exposure ( Fig 7C ) . This is consistent with previous studies of different insects ( including Culex pipiens ) that Hsp70 expression were up-regulated during recovery from exposure to subzero temperature [20 , 57–59] . Our results and others indicated that the up-regulation of Hsp70 might be required for the repair of cold injury caused by cold exposure [27] . Hsp70 protein went though obvious changes when exposing adult Ae . albopictus to -15 oC without water , which was similar with glucose . It increased dramatically by 6 fold from 1 to 4 h after recovery compared with those maintained under room temperature . This implied that severe acute cold shock might be happened in adult Ae . albopictus exposed to -15 oC without water and caused the serious disorders of glucose metabolism that eventually lead to the death of these adult mosquitoes . In conclusion , adult mosquitoes especially Ae . albopictus and Ae . aegypti , which are the most important vectors for dengue and Zika virus , could survive at low subzero temperature even below -10 oC for several hours’ exposure in the presence of water and this cold exposure have no detriment impact on fitness costs of adult Ae . albopictus . Both the relative high subzero temperature of water immediate after cold exposure and RCH induced by gradual cooling of water provided this strong protection against low subzero temperature . The cold tolerance might be conferred by accumulation of glycerol during cold exposure stage , and contributed by both glucose accumulation and Hsp70 up-regulation during the recovery stage from cold exposure . The RCH of adult mosquitoes induced by waters in nature might provide strong protection against acute decrease of air temperature to low subzero temperature , which often happens in early spring and winter , and late autumn , and this might be important for the survival and rapid expansion of Ae . albopictus to cooler areas . Our subsequent studies would be performed further to identify whether water-induced protection could be eliminated by down-regulation of glycerol or Hsp70 , and cold hardiness of eggs when exposed to low subzero temperature with water .
|
Aedes albopictus is one of two most important vectors for dengue and zika . During the last 3–4 decades , this mosquito has spread from native Asian area to all continents except Antarctica , becoming the most invasive mosquitoes which imposed extensive public health threat to human beings throughout the world . Cold tolerance is important for distribution and survival of insects . During the expansion of Ae . albopictus , especially a spatial expansion to cooler climate areas , it needs to cope with cold temperatures . Moreover , because of such widespread distribution adult Ae . albopictus will certainly often encounter sudden drops in air temperature even below subzero that often happens in early spring and winter , and late autumn . Thus far , adult Ae . albopictus are generally considered to be cold-intolerant that can not survive at subzero temperature . In this study , we found that water can provide strong protection against low subzero temperature even below -10 oC . Cold exposure of adult female Ae . albopictus to low subzero temperature with water either before or after a blood meal have no detrimental impact on fitness costs of these adult mosquitoes . Considering water is common in nature , our results indicated that during the expansion of Ae . albopictus especially when adult mosquitoes encounter a sudden drop in air temperature water could be a good shelter for cope with such cold temperature below subzero .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"methods",
"Results",
"Discussion"
] |
[
"invertebrates",
"medicine",
"and",
"health",
"sciences",
"body",
"fluids",
"chemical",
"compounds",
"monomers",
"carbohydrates",
"animals",
"organic",
"compounds",
"glucose",
"cold",
"hardiness",
"insect",
"vectors",
"polymer",
"chemistry",
"infectious",
"diseases",
"aedes",
"aegypti",
"cryobiology",
"culex",
"quinquefasciatus",
"chemistry",
"disease",
"vectors",
"insects",
"arthropoda",
"mosquitoes",
"eukaryota",
"blood",
"anatomy",
"organic",
"chemistry",
"physiology",
"glycerol",
"monosaccharides",
"biology",
"and",
"life",
"sciences",
"species",
"interactions",
"physical",
"sciences",
"organisms"
] |
2019
|
Water-induced strong protection against acute exposure to low subzero temperature of adult Aedes albopictus
|
Transcriptional reprogramming of macrophages upon Mycobacterium tuberculosis ( Mtb ) infection is widely studied; however , the significance of alternate splicing ( AS ) in shaping cellular responses to mycobacterial infections is not yet appreciated . Alternate splicing can influence transcript stability or structure , function and localization of corresponding proteins thereby altering protein stoichiometry and physiological consequences . Using comprehensive analysis of a time-series RNA-seq data obtained from human macrophages infected with virulent or avirulent strains of Mtb , we show extensive remodeling of alternate splicing in macrophage transcriptome . The global nature of this regulation was evident since genes belonging to functional classes like trafficking , immune response , autophagy , redox and metabolism showed marked departure in the pattern of splicing in the infected macrophages . The systemic perturbation of splicing machinery in the infected macrophages was apparent as genes involved at different stages of spliceosome assembly were also regulated at the splicing level . Curiously there was a considerable increase in the expression of truncated/non-translatable variants of several genes , specifically upon virulent infections . Increased expression of truncated transcripts correlated with a decline in the corresponding protein levels . We verified the physiological relevance for one such candidate gene RAB8B; whose truncated variant gets enriched in H37Rv infected cells . Upon tweaking relative abundance of longer or shorter variants of RAB8B transcripts by specialized transduction , mycobacterial targeting to lysosomes could be promoted or blocked respectively , which also resulted in corresponding changes in the bacterial survival . Our results show RAB8B recruitment to the mycobacterial phagosomes is required for phagosome maturation . Thus the abundance of truncated RAB8B variant helps virulent Mtb survival by limiting the RAB8B levels in the cells , a mechanism which we subsequently verified in human primary macrophages . Taken together we demonstrate alternate splicing as a new locus of intervention by Mtb and provide attractive alternative to exploit for novel drug targets against Mtb .
Infection of host macrophages with Mycobacterium tuberculosis results in significant alterations in the macrophage physiology . This includes altered regulation of common microbicidal processes like phagosome maturation , autophagy , apoptosis and other innate immune functions [1 , 2] . While the early responses like phagosome maturation are regulated through the perturbation of immediate signaling events like activation of signaling through PI3kinases and Ca2+ dependent kinases ( PKCs , CAMKII ) or pro-survival pathways [3 , 4] , long term shaping of the macrophage response largely depends on the modulation of transcriptional machinery resulting in a modified protein complement and physiology of the cells [5] . In order to track the changes in gene expression in the macrophages upon Mtb infections , several studies in the past reported microarray analysis of total RNA from infected macrophages , providing a great deal of information on the shaping of cellular responses upon infection [6–14] . Thus it was reported that genes involved in immune regulation , inflammation , metabolism and cell survival were differentially regulated upon Mtb infections [6 , 11 , 12] . In one such study , we previously showed distinct expression and activity of the tyrosine kinase Src in the survival of virulent Mtb strain H37Rv [5] . While the global expression analysis using microarray does provide details of cellular responses , an apparent non-correlation between gene expression and protein translation in the eukaryotes has always confounded the inferences drawn from these analyses [15–17] . The potential reasons for this discrepancy include a longer time lag in the eukaryotes between transcription and translation as well as the involvement of complex post-transcriptional mechanisms like splicing , poly-adenylation , RNA-editing , etc . [17] . Splicing incorporates lots of heterogeneity in the transcripts from a single gene by selectively retaining or excluding specific exons in the final processed mRNA . This leads to the generation of several alternately spliced variants of a single gene [18 , 19] . The alternate spliced variants of a transcript may have different structures , different functions , different sub-cellular localizations and stability owing to presence or absence of specific exons coding for specific protein modules [20] . Many of the alternate spliced variants do not get translated for reasons like presence of premature stop codons in each of the reading frames . The non-translatable transcripts may eventually get degraded through a process called non-sense mediated decay ( NMD ) [21] . Other truncated transcripts may get translated into a truncated protein with significant alterations in activity , binding , localization , etc . The role of alternate splicing in regulating the immune responses of cells is increasingly getting recognized . A series of studies show alternate splicing as a common mechanism that influences several molecules of the TLR signaling pathway and thereby regulates the inflammatory responses of the cell [22 , 23] . Several viral pathogens are known to alter AS events in the cells to facilitate viral replication and control cell cycle [24 , 25] . Splicing of eukaryotic transcripts occurs through a dynamic multi-step macro-molecular complex called spliceosome , considered as one of the most complex molecular machineries in the cell [26] . At different steps of spliceosome assembly and progression one or more of snRNPs ( small nuclear ribonucleoproteins ) like U1 , U2 , U4/U6 and U5 are involved each consisting of one or more snRNAs complexed with a vast number of accessory proteins . Their recruitment , catalysis and release are highly regulated to achieve high specificity of splicing as well as high flexibility to accommodate alternate splicing [27 , 28] . Involvement of a large set of proteins and multi-step regulation renders spliceosome functions vulnerable to cues where a large number of genes get differentially regulated . Intriguingly , while the global changes in gene expression upon infection of macrophages with Mtb are extensively studied there is no clear description of whether spliceosome components too get differentially regulated in the infected macrophages . The 3’UTR of the mammalian transcripts are yet another regulatory feature , mostly controlling the stability of the transcript [29 , 30] . The 3’UTR contains several potential miRNA binding targets and therefore based on the utilization of a proximal or distal poly-adenylation site a transcript may specifically exclude or include a particular miRNA site respectively [30] . The significance of alternate poly-adenylation has been shown in the case of cancers where individuals showed large variations in the poly-A site utilization [30–32] . Recently an alternate 3’UTR was shown to serve as a scaffold to regulate membrane protein localization [33] . Regulation of miRNA upon Mtb infections is an intense area of investigation . Several studies in the past have shown specific miRNA expression upon Mtb infection [34–38] . It is therefore very likely that the cell could have evolved the APA strategy to counter the miRNA-mediated decay in the infected macrophages as well . Here we followed the time-resolved transcriptome data of THP-1 macrophages that were infected with H37Rv ( virulent ) or H37Ra ( non-virulent ) strains of Mycobacterium tuberculosis . We report extensive alternate splicing and alternate poly-adenylation events at the global scale in infected macrophages . Importantly , the transcript variants formed due to these events significantly contributed in deciding the fate of cellular response to infections .
Total RNA isolated from H37Ra and H37Rv infected THP-1 macrophages at 0 , 6 , 12 , 24 , 36 and 48 hours post-infection were used to make cDNA libraries followed by sequencing using Illumina Hiseq2000 platform ( see methods for detail ) . The experimental set-up is schematically shown in Fig 1A . Quality control for the paired end raw sequence data set was performed using FASTQC kit . The reads obtained were of very high quality as more than 95% of reads across the conditions crossed the Phred score of Q20 while more than 88% of reads were above Q30 threshold ( S1A Table ) . A score of Q20 corresponds to incorrect base call at 1in 100 while the score of Q30 means incorrect call at 1 in 1000 . These values correspond to overall base call accuracy of 99% ( Q20 ) and 99 . 9% ( Q30 ) respectively . Reads with a score higher than Q30 were taken further for downstream analysis . The downstream analysis flowchart is shown in Fig 1B . For each sample , approximately 180 million paired-end reads of 101 bp was used for genome-guided alignment using Tuxedo pipeline . Alignment of raw reads on human genome build Hg19 was carried out using splicing aware Tophat aligner [39] . For each sample , more than 70% of reads aligned on Hg19 ( S1B Table ) . Of the aligned reads , ~3% aligned to intronic regions , ~2% aligned to intergenic regions and the remaining ~94% aligned to exons . The absolute quantitation ( fragments per kilobase per million reads or FPKM ) of genes and transcripts and their differential regulation as compared to the uninfected control was obtained using Cufflink and Cuffdiff package [39 , 40] . The analysis through Cufflink-Cuffdiff package also provides statistical measures for identifying every regulated gene . The entire list of regulated genes across the time points in both H37Ra and H37Rv infected cases is provided in S2 Table . We also characterized a few global properties of expression profiling before analyzing the genes that were differentially expressed . First , the dispersion analysis of reads aligning to the genome revealed an absence of any sample specific bias in alignment ( S1A Fig ) . Further , we tested deviation of individual gene expression around the median of the particular sample . All the 12 samples showed a nearly identical median; however there were some infection and time point specific alterations in the deviation , like more down-regulated genes in H37Ra infected macrophages at 0 and 6 hours but more down-regulated genes in H37Rv infected macrophages at 12 , 24 and 36 hours post-infection ( S1B Fig ) . The density distribution plot , shown here for 36-hour time point , shows relative differences in the gene level expression between H37Ra and H37Rv infected cells with respect to uninfected cells ( Fig 1C ) . Similar plots for each of the time points are shown in supplemental S2 Fig . Between H37Ra and H37Rv infected macrophages we got several unique and overlapping sets of genes showing specific regulation across different time points . Fig 1D comprises a list of numbers of all such uniquely or commonly regulated genes across the points . The comparison groups included identifying genes that were commonly up or down regulated ( rows 1 and 2 in Fig 1D ) , genes which show exactly contrasting pattern like up in H37Rv and down H37Ra or vice versa ( rows 3 and 4 in Fig 1D ) , cases where genes are up or down regulated in H37Rv and un-regulated in H37Ra and vice-versa ( rows 5 , 6 , 7 and 8 in Fig 1D ) and total number of genes showing up or down regulated pattern in one case irrespective of the strain case ( rows 9 , 10 , 11 and 12 in Fig 1D ) . This analysis was performed at each of the time points ( Fig 1D ) . A few simple observations emerged through this analysis , e . g . there were more common up-regulated genes than down-regulated ones . Secondly , the number of genes showing exactly contrasting expression between H37Ra and H37Rv infected cells was very small ( <100 throughout except for 36 hours time point , rows 3 and 4 combined in Fig 1D ) . The functional class analysis of these differentially regulated genes followed expected patterns , as shown for 48 hours time point in Fig 1E and remaining time points in S3 Table . Gene enrichment analysis revealed significant enrichment of genes belonging to metabolism , gene regulation , trafficking , immune , inflammation and chemokine/cytokine signaling suggesting , expectedly , massive perturbation of macrophage innate immune function . The functional classes overlapped with our previously published microarray experiments [5] . We were keen to understand transcript level expression pattern in the infected macrophages . Most of the genes in the human genome have known transcript variants and isoforms . We modified the transcript-specific GTF file ( see methods ) and followed the Cufflink-Cuffdiff package to obtain transcript-specific expression data . We first checked there was no sample specific bias in the dispersion of reads alignment at the transcript level . We also compared transcript level distribution of expression in each of the samples with respect to the uninfected control . All samples showed near normal distribution of transcript level expression ( S3A Fig ) . The whole list of isoform-specific expression across the groups and time points is provided in supplemental S4 Table . At 36 hours , H37Rv infected cells transcript expression distinctly diverged from the uninfected cells , however , continued to show normal distribution suggesting significant regulations at this time point ( S3A Fig ) . We were curious to see whether transcript level expression profile differed from the pattern observed at the gene level . We checked transcript expression for several genes like CORO1B , ACSL1 , PGK1 , ATG13 , IL1B , RAB8B , BRI3 and COX7A2 ( Fig 2A ) . The selection of genes was driven by empirical observation of differential transcript expression as well as due to their easy association with innate immune regulation and cellular responses to Mtb or other infections [41–44] . The most common observation was that different transcripts of a given gene showed wide variations in terms of expression and greatly differed from the patterns observed at the corresponding gene level ( Fig 2B ) . For most of the cases , the expression of transcripts was more contrasting and prominent between H37Ra and H37Rv infected samples at later time points specifically at 24 and 36 hours post-infection , as discussed in the next section . In cDNAs prepared from independent experiments , we verified expression patterns observed at the transcript levels for select isoforms of each of the genes discussed in Fig 2A through Q-PCR analysis . The selection of transcript for validations was mainly driven by two factors: it should show large differences at FPKM level , which could allow them to be picked up by RTPCR analysis . Secondly , we focused more on shorter transcripts of each gene . The significance of shorter transcripts is discussed in more detail in subsequent results sections . Thus we picked ENST00000545736 , ENST00000505492 , ENST00000491291 , ENST00000579280 , ENST00000496280 , ENST00000558990 , ENST00000456357 and ENST00000472311 for CORO1B , ACSL1 , PGK1 , ATG13 , IL1B , RAB8B , BRI3 and COX7A2 respectively ( marked as arrowheads in Fig 2B ) . The expression pattern between uninfected , H37Ra infected and H37Rv infected largely matched between the independent Q-PCR experiments ( Fig 2C ) and RNA-seq data ( Fig 2D ) . While the variation at the isoform level expression was an interesting observation , it also raised questions about the suitability of earlier approaches using microarray . Different isoforms of a gene vary in their functionality due to reasons like inclusion/exclusion of certain exons , stability of the transcript and differential targeting [20] . Variations in the isoform level expression arise due to differential alternate splicing ( AS ) events and therefore we decided to analyze the global pattern of AS upon infection . For a gross estimation of alternate splicing events in H37Ra and H37Rv infected macrophages , we compared the junction read counts that originated from exon-exon boundaries across different samples . While the alignment of reads to exon junctions is purely coincidental , we did observe gross differences in the total number of reads corresponding to the exon-exon boundaries between H37Ra and H37Rv infected cells across all the samples ( S3B Fig ) . For a more statistically qualified understanding of alternate splicing in these macrophages , we followed the robust Bayesian analysis framework MATS ( Multivariate Analysis of Transcript Splicing ) [45] . AS events were classified into five major groups: Alternative 5’ splice site ( A5SS ) , Alternative 3’ splice site ( A3SS ) , Skipped exon ( SE ) , Mutually exclusive exons ( MXE ) and Retained exon ( RI ) as reported earlier [45] . In order to get greater insights into the extent of alternate splicing in these macrophages , we calculated “percent splicing index ( psi ) ” score ( ψ-score ) for each of the transcript variants with respect to the uninfected control samples [20] . A very stringent cut-off for difference of psi-score from uninfected sample ( 0 . 5 ) was taken as significant differential splicing events , which were induced upon infection with H37Ra or H37Rv [20] . In Fig 3A , the number of each of the 5 possible AS events with a psi-score difference of more than 0 . 5 with respect to uninfected is listed ( Fig 3A ) . The complete list of AS events and corresponding psi-scores across all the samples and time points is provided in S5 Table . Switch like events were detected where the difference in psi-score was exactly 1 or -1 . We next plotted ψ-scores of transcripts in H37Ra infected cells ( X-axis ) versus that of H37Rv infected cells ( Y-axis ) across the course of infection ( Fig 3B ) . The plots clearly showed that in addition to the AS events being specific to infection with Mycobacterium tuberculosis , all data points at the top right in grey in Fig 3B ( green box ) show significant AS in both H37Ra and H37Rv infection , a large number of them were specific to the infecting strains , indicating differential regulation between H37Ra and H37Rv infected macrophages differing by 0 . 5 or more ( in red and blue respectively , Fig 3B ) . More switch like events were detected in H37Rv infected samples compared to H37Ra infected samples , which can be easily visualized in Fig 3B as alignment of several red dots on the top left boundary of the psi-score plots at 0 , 6 , 12 and 36 hours . In H37Ra infected cells , 48 hours sample showed very high number of switch like events ( blue dots at bottom , right boundary; Fig 3B ) . A list comprising genes that show psi-score of 1 in H37Rv infected macrophages is provided in Table 1 . While the list in Table 1 does not necessarily captures the whole set of genes , which show AS upon infection with H37Rv , it nevertheless reflects that even for the AS cases as specific as having a psi-score of 1 , large number of genes with apparently varying functions are regulated through this means ( Table 1 ) . The exact numbers of common and differential splicing events between H37Ra infected and H37Rv infected macrophages across all time points are shown in Fig 3C . Strain-specific splicing patterns got strongly expressed at later time points specifically at 24 , 36 and 48 hours post-infection as the total number of genes showing significant splicing and unique to the infection was consistently higher than 1500 at these time points ( first two rows combined , Fig 3C ) . In addition , we also analyzed psi-score distribution for those genes , which were significantly regulated in our analysis in Fig 1 . These six additional psi-score plots in supplemental S4 Fig highlight that as we go later in the course of infection , more genes that are differentially regulated at gene level start showing significant alternate splicing as well ( Fig 3D and S4 Fig ) . To understand which biological functions were majorly targeted through AS upon Mtb infections , we performed a gene ontology analysis of the genes showing significant alternate splicing unique to either H37Ra or H37Rv infected cells . The large number of functional classes associated with the respective gene list is shown in S6 Table . To infer from the GO analysis , we manually checked for functional classes and classified them into one of the following categories like metabolism , trafficking , redox , gene regulation , cell cycle , lipid metabolism , RNA processing , DNA damage/repair , protein transport , apoptosis , ubiquitination etc . and clubbed the smaller redundant functional classes into large parent functional class . The pi-charts of genes belonging to these functional classes at each of the time points are shown in Fig 3E . Genes belonging to certain classes like metabolism , gene regulation and trafficking or organelle organization undergo alternate splicing throughout the course of infection , irrespective of the infecting strain . Though there were differences in the list of genes within each functional class between H37Ra and H37Rv infection cases ( S6 Table ) . At 48 hours post infection , H37Ra infected cells had more functional perturbations , as evident with more functional classes at 48 hours ( Fig 3E ) . Similarly , at 12 , 24 and 36 hours post-infection , H37Rv infected cells showed maximum functional perturbations , as evident by the number of functional classes in the pie-chart ( Fig 3E ) . Some interesting functional classes , which were more perturbed via splicing in H37Rv infected cells include protein transport ( 12 and 36 hours ) , redox ( 36 hours ) , RNA processing ( 12 , 24 and 36 hours ) , lipid metabolism ( 24 hours ) , translation ( 36 hours ) and ubiquitination ( 12 , 24 and 36 hours; Fig 3E ) . Together , it was clear that infection induced alternate splicing of transcripts was widespread and therefore had potential to dramatically alter the host response to infection . Various transcripts/isoforms of a given gene vary in the length of 3’ untranslated regions ( UTRs ) . Searching through the Ensembl database revealed even a single transcript could have several different 3’UTR lengths . Fig 4A shows the number of transcripts and their known numbers of 3’UTR as reported in the Ensembl database . A global analysis comparing the 3’UTR length versus the corresponding FPKM revealed that UTR length expression distinctly varied between uninfected and H37Ra or H37Rv infected macrophages ( Fig 4B ) . For a better understanding of the 3’UTR size variation and their respective expression levels , we classified UTRs into size ranges and looked for their expression in terms of FPKM . There were infection and strain-specific variations in the number of transcripts expressing longer or shorter 3’UTR ( S5 Fig ) . The 3’-UTR length may be regulated through the use of proximal or distal poly-adenylation signals ( poly-A signal ) present in the transcript; a process also termed as alternate poly-adenylation [30] . Higher use of proximal poly-A sites will result in more transcripts with smaller UTR while distal poly-A uses will result in longer UTR . We next calculated percent distal poly-A site uses index ( PDUI ) score to compare the relative UTR length versus expression using DaPars algorithm [46] . DaPars algorithm has an inbuilt system to notify statistically significant events . The table with PDUI score and corresponding p-values is provided as S7 Table . It turned out; there was a distinct shift in the PDUI score in H37Rv infected macrophages , suggesting an overall increase in the uses of distal poly-A sites and increased expression of transcripts with longer 3’-UTR ( Fig 4C ) . As expected , transcripts with longer UTR showed more differential regulation as compared to transcripts with shorter UTR length ( Fig 4C ) . In the literature , it is reported that AS and APA events may be linked , suggesting transcripts undergoing AS also have increased chances of undergoing APA [29] . In our data , we could observe nearly 10–15% of transcripts that showed APA also showing AS . These numbers were considerably higher than overall percent significant AS events ( 0 . 8–1 . 6% ) and percent significant APA events ( 0 . 2–2 . 6% ) alone thereby suggesting AS and APA to some extent indeed may be linked events . Gene ontology enrichment of the list present in S7 Table showed genes from a majority of functional classes undergoing APA including those involved in splicing , phagocytosis , immunity , apoptosis and metabolism ( S8 Table ) again emphasizing the global APA events as a consequence of Mtb infection . Having witnessed considerable changes in the splicing pattern in macrophages post-infection , we were curious to test whether genes belonging to the different stages of spliceosome complex could also undergo alternate splicing . We downloaded the list of genes involved in splicing events from the spliceosome database [47] . Off nearly 131 genes involved in splicing , a vast majority did not show any significant regulation in either virulent or avirulent infection at any of the time points studied ( Fig 5A ) . Among few that were differentially regulated , PRPF19 was significantly down regulated across all the time points in both H37Rv and H37Ra infected macrophages ( Fig 5A ) . Few more genes , which showed down regulation at 48 hours in H37Ra infected cells , included LENG1 , NOSIP , PPIL1 , RBM8A , SNRPB , SNRPC and SNRPF . Out of these LENG1 and PPIL1 were also down regulated in H37Rv infected macrophages at 48 hours post-infection ( Fig 5A ) . We next compared the expression of the spliceosome-associated genes at transcript levels . At the transcript level the scenario was grossly different with a large number of transcripts showing marked regulation in expression with respect to the control in both H37Ra and H37Rv infected macrophages ( Fig 5B ) . The list of transcripts associated with spliceosome genes and their corresponding expression level at each of the time points is provided in S9 Table . The range of transcript expression varied between ~15fold down regulation to ~12 fold up regulation ( Fig 5B ) . While alternate splicing may result in varied lengths of the final product , we were especially intrigued with the frequency of shorter or truncated transcripts , almost universally present for most of the genes . These transcripts usually spanned lesser than 1000bps and multiple variants of these transcripts were present for a large number of genes . Shorter transcript variants in many instances do not get translated or when translated , do not form a functional protein . Transcripts with premature stop codon get degraded through nonsense-mediated decay ( NMD;[48 , 49] . To test how the expression of these truncated transcripts was regulated , we plotted transcript length versus corresponding fold change in expression for each of the transcripts of genes associated with spliceosomes across all the conditions ( Fig 5C ) . A cursory examination of the plots revealed that more number of transcripts shorter than 1000bp was differentially regulated than those transcripts that were longer than 1000bp . As shown in Fig 5C , for each of the condition/time-points the plots clearly reveal this bias . For comparison when we plotted the entire transcriptome data for any condition , this bias was visibly lost or at least declined ( Fig 5D ) . To get a numerical representation of this bias in the relative enrichment of transcript length among highly regulated transcripts , we calculated the ratio of the number of shorter transcripts differentially regulated to that of the number of longer transcripts differentially regulated for each experimental groups . This ratio was calculated separately for transcripts associated with spliceosome genes or for the entire transcriptome ( Fig 5E ) . The plots in Fig 5E show that among the spliceosome genes there is a far greater propensity for the enrichment of differentially regulated shorter transcripts compared to the whole transcriptome . The differences were also statistically significant as we confirmed through a hyper-geometric analysis ( Fig 5E ) . It was true for each of the time points studied in both H37Ra and H37Rv infected macrophages ( Fig 5E ) . Finally , we captured the transcript variants and their relative expression on a protein functional association network between the genes involved at different stages of splicing like complex A , complex B , complex Bact , complex C and complex P [28] . For all the proteins known to be part of the spliceosome complexes as published earlier ( 28 ) their interaction pattern was obtained from STRING database to create a functional association network of spliceosome components . Since different stages of spliceosome assembly involve several common and unique molecules , we partitioned the network into smaller circular sub-clusters , each of them common to a set of splicing complexes ( Fig 5F ) . Thus , the first circle on the top , left of the network included molecules that were part of each of the five spliceosome stages A , B , Bact , C and P ( Fig 5F ) . Similarly , other circles were created specifying the spliceosome stages they represented . Following that , the expression value ( either maximum or minimum , represented by colors red or green respectively ) and length of the corresponding transcript ( represented by the size of the nodes ) were incorporated into the network to visualize how the spliceosome machinery was perturbed upon infection . We used expression data at 48 hours post-H37Rv infected macrophages for this analysis . As is visually evident from Fig 5F , more transcripts of shorter length were up or down regulated at multiple steps of the spliceosome function ( Fig 5F ) . This analysis corroborated with the earlier observation that shorter transcripts corresponding to the spliceosome related genes were more differentially regulated as compared to their longer counterparts . Together Fig 5 highlights that the observed variations in the pattern of splicing in infected macrophages influence the spliceosome components . As seen above , infection with Mtb led to an increase in the expression of truncated transcripts for splicing related genes . Functional analysis of highly spliced variants too revealed several genes belonging to key functional classes like immune regulation and response to stress getting enriched in the list of alternate spliced genes . We decided to investigate the influence of alternate splicing particularly that of truncated transcript on shaping the cellular response to infection . We verified the increased expression of shorter transcripts in the infected macrophages by performing isoform-specific real-time PCR analysis of genes like RAB8B , ACSL1 and Dynamin-1 ( Fig 6 and S6 Fig ) . Selection of these genes for subsequent analysis was with keeping in mind their known association with the aspects of trafficking ( RAB8B and Dynamin-1 ) as well as lipid metabolism ( ACSL1 ) , both key functional aspects in the virulence [42 , 44] . To visualize the extent of alternate splicing , we constructed SASHIMI plots for RAB8B , ACSL1 and Dynamin-1 ( Fig 6 and S6 Fig ) . SASHIMI plots integrate the probability of a splicing event keeping into account reads corresponding to the exon-exon junctions [50] . As shown in Fig 6 and S6 Fig , in each of the three cases analyzed , considerable differences were observed in terms of the number of reads corresponding to exon-exon junctions between uninfected or H37Ra or H37Rv infected macrophages . The comparative exon organization plots for the shorter transcript with respect to the full-length transcripts are shown in Fig 6B and S6 Fig ) . The shorter transcript of RAB8B ( ENST00000558990 ) had completely different exon composition with respect to the full length RAB8B ( ENST00000321437 ) transcript ( Fig 6B ) . At the gene level , expression of RAB8B , ACSL1 and Dynamin-1 was minimum in case of uninfected macrophages , intermediate in case of H37Ra infection and highest in the case of H37Rv infection ( Fig 6C and S6 Fig ) . We then calculated percent contribution of each of the known transcripts of RAB8B , ACSL1 and Dynamin-1 ( Fig 6D and S6 Fig ) . Surprisingly , a large proportion of the increase in expression witnessed in H37Rv infected macrophages at the gene level was contributed by an increase in the expression of the corresponding truncated transcripts in RAB8B and ACSL1 ( Fig 6D and S6 Fig ) . In the case of Dynamin-1 , even H37Ra infected macrophages showed increased expression of the truncated transcript ( S6 Fig ) . Using isoform-specific primers , designed from independent exon-exon boundaries , we then validated through real-time PCR , increased expression of the truncated isoforms of RAB8B and ACSL1 ( Fig 6E and S6 Fig ) . For real-time experiments , we used both β-Actin and 18srRNA as controls in independent experiments . Expectedly , as seen in the case of RNAseq data , there was no difference in the expression of longer transcript between UI , H37Ra or H37Rv cells ( Fig 6F ) . We were keen to understand whether and how increased expression of the truncated isoforms could influence cellular responses to infections . The truncated RAB8B transcript ENST00000558990 does not get translated since it has a premature stop codon . It rather undergoes nonsense-mediated decay ( NMD ) , a process that keeps quality control of mRNAs [51] . NMD takes place while the mRNA is still part of the translational machinery . We , therefore , isolated the polysome fraction from uninfected , H37Ra infected and H37Rv infected THP-1 macrophages , ( Fig 6G ) . From the total RNA isolated from the polysome fraction , we checked for the presence of truncated transcripts . In the case of RAB8B , we could observe a nearly six-fold increase in the level of truncated transcript in the polysome fraction from H37Rv infected macrophages as against the uninfected control ( Fig 6H ) . Even in the case of H37Ra infection , there was nearly two-fold increase in the enrichment of truncated isoform in the polysomal fraction . We next verified the effect of alternate splicing on RAB8B protein levels and whether that influenced host response to Mtb infections . Estimation of relative protein abundance of RAB8B revealed that increased expression of the truncated transcript correlated with a sharp decline in the protein levels in H37Rv infected macrophages ( Fig 7A ) . It was true for ACSL1 proteins as well ( S7A Fig ) . How increased expression of truncated transcript led to a decline in RAB8B protein level remains less understood . However , one possibility is that it can compete with the full-length variants for translational machinery . RAB8B belongs to the family of small monomeric GTPases , involved in the process of intracellular trafficking [44 , 52] . This protein also seems to influence the maturation of Mtb-containing phagosomes and autophagosomes [44] . We compared recruitment of RAB8B to GFP-expressing H37Ra and H37Rv phagosomes in THP-1 macrophages at 48 hours post-infection . While H37Ra showed nearly 36% co-localization with RAB8B , it was significantly lower ( 22% ) in the case of H37Rv infections ( Fig 7B ) . H37Ra is known to get readily targeted to lysosomes unlike H37Rv [53] . It , therefore , seemed plausible to hypothesize that by decreasing the RAB8B protein levels through this unusual means of post-transcriptional regulation , the virulent strain of Mtb evades getting targeted to the lysosomes and therefore promotes its survival . To further test this hypothesis we cloned the full-length or the truncated transcript cDNAs of RAB8B into the pMSCV-PIG vector where the expression cassette contains GFP as part of a bicistronic construct separated by an IRES element ( Fig 7C ) . Lentiviruses were made by transfecting these constructs into HEK293T cells along with the helper plasmids and then transduced into THP-1 macrophages . The purpose of this experiment was to test whether by overexpressing shorter or longer transcripts of RAB8B , survival of Mtb could be regulated within THP-1 macrophages . We were able to achieve more than 70% transduction of long and short RAB8B transcripts as well as vector controls in THP-1 macrophages at the time of infection as well as 48 hours post-infection ( S7B Fig ) . The real-time analysis confirmed increased expression of both shorter and longer transcripts upon transduction ( Fig 7D ) . Western blot analysis of transduced THP-1 macrophages , interestingly , correlated with the pattern observed above in H37Rv infected macrophages . Thus , whereas cells transduced with longer RAB8B transcript showed a higher level of RAB8B proteins , those transduced with shorter RAB8B transcript showed a diminished level of RAB8B proteins with respect to vector control ( Fig 7E ) . We next infected these macrophages with H37Ra or H37Rv and monitored their survival . In vector control transduced cells , H37Ra and H37Rv survival followed the expected pattern , resulting in higher H37Rv CFU at 48 hours post infection compared to H37Ra ( Fig 7F ) . There was no difference in the uptake of bacteria in THP-1 macrophages under any of the three conditions ( Fig 7F ) . In cells transduced with shorter RAB8B transcript , both H37Ra and H37Rv showed higher CFU compared to vector control ( Fig 7F ) . Similarly overexpressing longer RAB8B isoform led to diminished CFU for both strains with respect to the vector control ( Fig 7F ) . It is important to note here that the effect of overexpressing longer transcript on H37Ra survival ( decline of nearly 11% with respect to vector control ) and that of shorter transcript on H37Rv survival ( increase in CFU by ~15% ) were although significant but of lower magnitude than vice versa i . e . ~60% increase in H37Ra CFU in cells transduced with shorter transcript while ~30% decline in H37Rv CFU in cells transduced with longer transcript ( Fig 7F ) . It possibly reflected already high RAB8B protein levels in H37Ra infected cells and low RAB8B protein levels in H37Rv infected cells . In fact , H37Rv infected macrophages that were also transduced with shorter RAB8B transcript showed further diminished RAB8B protein level with respect to H37Rv infected cells alone , thereby confirming the same ( S7C Fig ) . It was now evident , by lowering RAB8B protein level via the unusual means of post-transcriptional regulation virulent Mtb strain was aiding its survival within macrophages . To understand how RAB8B protein levels influenced Mtb survival within macrophages , we analyzed lysosomal targeting of Mtb in cells that were transduced with the long or short isoform of RAB8B as a bi-cistronic construct with GFP . Expression of GFP allowed us to analyze only those cells , which were transduced . We found levels of RAB8B protein correlated with the lysosomal targeting of Mtb . Thus more H37Rv localized to lysosome when cells were transduced with RAB8B long transcript compared to vector control ( Fig 8A ) . Similarly lesser H37Ra localized to lysosomes when transduced with short RAB8B transcript compared to vector control ( Fig 8A ) . Due to the large variation observed in the confocal microscopy data , mostly due to differential level of expression of the transduced copy of RAB8B , we could not get significant differences between vector control and long isoform group although they followed the expected trend . Transduction of shorter transcript , however , showed consistently significant difference from both vector control and long isoform ( Fig 8A ) . Thus , levels of RAB8B protein influenced phagosomal maturation and survival of Mtb within infected macrophages . The levels of RAB8B protein itself , as shown above , was regulated by the pattern of alternate splicing . To further confirm the specific role of RAB8B in regulating Mtb killing , we knocked down RAB8B ( using siRNA specific to the longer isoform ) in THP-1 macrophages that were infected with either H37Ra or H37Rv . We observed nearly 80% knockdown of RAB8B at protein level ( Fig 8B ) . RAB8B knockdown led to decreased targeting of both H37Ra ( ~60% ) and H37Rv ( ~20% ) to lysosomal compartments as compared to scramble siRNA treated control ( Fig 8C ) . Finally , we were also able to observe a significant increase in the survival of both H37Ra and H37Rv in RAB8B knockdown cells ( Fig 8D ) . In agreement with the findings in the previous section , increase in bacterial CFU upon RAB8B knockdown with respect to control was more pronounced in the case of H37Ra ( ~65% ) than that of H37Rv ( ~50% , Fig 8D ) . As noted earlier , alternate patterns of splicing and poly-adenylation has been shown to be associated with different diseases including cancers [54 , 55] . It was , therefore , important to test that truncation of RAB8B transcript upon H37Rv infection was not a consequence of the transformed phenotype of THP-1 cells , a leukemic cell line , rather represented a true macrophage response upon bacterial infection . We tested some of the genes analyzed in Fig 2 from monocyte-derived macrophages ( MDMs ) obtained from PBMCs of healthy volunteers . The pattern of splicing of genes IL1B , ACSL1 and ATG13 upon infection with H37Ra or H37Rv followed the trend observed in THP-1 macrophages , showing maximum expression in H37Rv infected cells than others ( Fig 9A ) . For RAB8B , we confirmed its splicing patterns from two different donors . While both the donors show a similar increase in RAB8B truncated transcripts upon H37Rv infection , one of them also showed an increase in truncated RAB8B upon H37Ra infection , however levels in H37Ra infected MDMs were always lower than those in H37Rv infected ones ( Fig 9B ) . In fact , in the second donor , longer RAB8B transcript increased manifold upon H37Ra infection , suggesting strong phagosome maturation flux in that individual ( Fig 9B ) . All subsequent validations were done using MDMs from donor 1 . We next confirmed , increase in RAB8B shorter transcript expression correlated with a decline in RAB8B protein level in MDMs upon H37Rv infection ( Fig 9C ) . In agreement with the results in THP1 macrophages , more H37Ra ( ~40% ) localized to the RAB8B compartment in the MDMs compared to H37Rv ( less than 30% , Fig 9D ) . We were also able to knockdown RAB8B longer isoform in MDMs using siRNAs ( S7D Fig ) . Upon siRNA-mediated knockdown of RAB8B in the MDMs , both H37Ra and H37Rv showed an increase in CFU ( Fig 9E ) . Expectedly the increase was more pronounced and significant in the case of H37Ra infection as well as reduced co-localization to lysosomes ( Fig 9E ) . Similarly , upon RAB8B knockdown bacterial co-localization to lysosome declined , which was again more pronounced and significant in the case of H37Ra infection compared to H37Rv infection ( Fig 9F ) . Thus infection induced alternate splicing of RAB8B is a specific response of macrophage , which helps the survival of virulent strain in the infected macrophages .
Alternate processing of pre-messenger RNA results in the production of several variants of mRNA of a given gene [20] . Alternate splicing is attributed to play a crucial role in the generation of biological complexity , and arguably , therefore , mis-regulation in the alternate splicing machinery is associated with several human diseases [54] . In fact , it is estimated that nearly 15–50% of human disease mutations directly influence the splice site selection [54 , 55] . The role of alternate splicing in the regulation of cellular response , especially in the immune responses , is steadily getting evident . In innate immune regulation , the role of alternate splicing has been extensively studied in the context of TLR4 signaling [22] . The converging theme across several studies in this context was how alternately spliced variants of a transcript result in non-functional and/or truncated variations of the protein , resulting in switching off of the immune signaling [23] . Infection with bacterial pathogens including Mycobacterium is known to induce a sharp change in the macrophage gene expression profile [5 , 14] . However , most such studies involved microarray analysis and relied on gene level interpretations . Here we performed RNA-seq analysis and comprehensively investigated transcript variants that were regulated upon infection in macrophages . At gene level classification , functional enrichment analysis did reveal similar functional classes in our study , as reported earlier , like metabolism , immune regulation or inflammation , etc . [5] . Eukaryotic gene expression is a complex process , where primary transcripts are processed at multiple steps including capping , cleavage and polyadenylation , cytosolic export , splicing , editing , etc . [56] . Most of these steps are regulated and could significantly impact the overall stoichiometry and functioning of corresponding translated products [56] . A comparative analysis of gene-level and isoform level expression of several genes showed discordance between them for most of the genes . Thus certain isoforms of a given gene were more enriched in expression and that the identity of enriched isoforms varied between H37Ra infected , H37Rv infected and uninfected macrophages . It strongly indicated that infection with Mtb was altering the global patterns of alternate splicing in the macrophages . The psi-score analysis , which factors in the frequency of exons spliced in among different isoforms of a given gene , indeed confirmed infection and strain-specific regulation of alternate splicing . To the best of our knowledge , infection induced alternate splicing at the global level has never been shown in the past . While the altered pattern of splicing itself was novel observation , the functional consequences of it became more evident as we noted for several genes , the truncated isoforms were highly regulated in the infected macrophages . Many of the truncated transcripts are known to have pre-mature stop codons , and they do not get translated [57] . These transcripts are subsequently degraded through a process called non-sense mediated decay [57] . How increased abundance of truncated transcripts influence cellular response is not exactly clear , however in one such study in an unrelated system , it was shown that non-sense mediated decay cramps the nuclear cap binding protein ( NCBP ) in an inactive form thereby influencing the cytosolic export of newly formed transcripts [58] . It will be interesting to see whether similar mechanisms are operational in higher eukaryotes as well . It is important to note here that we used a highly stringent cut-off ( 0 . 5 ) to select AS events . In the literature , a cut-off of 0 . 2 is considered significant to identify AS events . At that cut-off , the number of genes showing AS will considerably increase . Thus AS as a consequence of Mtb infection may be even more wide-spread than being presented in this study . The global nature of alternate splicing triggered by Mtb infection was also evident by our analysis that revealed extensive alternate splicing of genes belonging to spliceosome complex . Since many AS events depend on recognition/non-recognition of weak splice sites [59] , loss of specific functional domains of a protein or decrease in the concentration of certain spliceosomal protein due to AS could influence the overall splice site selection and therefore AS . Moreover , it is understood that many important binary interactions between spliceosome complex proteins are weak in nature and require several additional factors to get stabilized [60] . Thus alternate spliced products of spliceosome proteins are likely to vary in the stabilizing properties , further favoring extensive alternate splicing . Given the multiplicity of factors involved , understanding the exact mechanism of AS regulation upon Mtb infection would require a more focused study . We were also intrigued with yet another form of transcript length regulation via alternate polyadenylation [30] . The length of 3’-UTR of transcripts can influence the stability of the transcripts via exclusion/inclusion of miRNA binding sites which typically target the 3’UTRs [30] . Such regulation during transformation has been reported where several genes involved in cell cycle/survival were shown to have a shorter 3’UTR and longer stability resulting in highly efficient translation . Regulation of alternate polyadenylation upon infection has never been discussed in the past . We identified several APA events that were specific to virulent infections . In general , virulent infections resulted in more elongation of transcripts ( increased use of a distal poly-A site ) . That could be another means to destabilize the transcripts and eventually influence the corresponding protein levels in the cells . Our studies with RAB8B isoforms in regulating cellular responses to infection were quite revealing . While there were several molecules of interest in the AS list , we followed the effect of AS on RAB8B for two reasons . Firstly , RAB8B is known to participate in the phagosome maturation pathway involving RAB7 and TBK1 [44 , 52] . Secondly , it showed an interesting pattern where increased expression of truncated transcripts in virulent infection led to a decline in the protein level . Phagosome maturation arrest is an established mechanism through which virulent strains of Mtb are known to evade lysosomal targeting [53 , 61] . The role of small GTPases like RAB7 is established in the maturation process , and exclusion of RAB7 from Mtb phagosomes is one of the core mechanisms helping the bacteria to evade lysosomal targeting [61] . This study adds another dimension to the regulation of phagosome maturation in Mtb-infected macrophages , whereby levels of RAB8B , a molecule closely associated with RAB7 , gets significantly depleted as a consequence of alternate splicing . That the RAB8B protein levels could be regulated by the relative expression of shorter and longer isoforms was confirmed using transduction studies as well as knockdown studies . Having identified and established the AS-mediated regulation of macrophage response in THP-1 macrophage , which are monocytic leukemia cells , it was critical to test whether similar mechanisms are operational in primary macrophages as well . That is partly because genetic variations in the splice-site are linked to several diseases in humans including cancer [62 , 63] . In macrophages derived from peripheral blood of healthy donors ( MDMs ) , we could confirm infection specific alterations in the expression of IL1B , ACSL1 , ATG13 and RAB8B isoforms as noted in THP-1 macrophages . We further verified that selective RAB8B splicing influenced the outcome of infection , confirming it to be a true macrophage response . Even with a very limited analysis of RAB8B AS upon H37Rv infection in macrophages from two different donors , it was evident that there is an enormous possibility of inter-individual variations in the patterns of splicing . Therefore it is likely that depending on the extent of AS and APA regulation , different individuals in the population may show different susceptibility to the disease , which is an exciting concept to follow-up in a subsequent study . While we characterized the biological consequences only for one molecule in this study , as a proof-of-concept , it represents an important breakthrough , implicating for the first time global alternate splicing particularly in the context of Mtb infections . To assume how each of the alternate spliced and alternate poly-adenylated transcripts could together regulate the overall outcome of infection looks both challenging and exciting . It provides new opportunities to explore host responses to infection through the window of post-transcriptional regulations . Identification of some universal splicing code , which remains elusive till date , has been the focus of the intense investigation by several groups [19 , 54 , 57] . These results may provide newer tools to explore the spliceosome components , assembly , cis-regulatory elements and their perturbation by external factors like virulent bacterial factors . Interestingly , in some recent reports involvement of mycobacterial factors in regulating host epigenetics has been reported [64 , 65] and therefore provides sufficient indications that certain bacterial factors may also directly or indirectly influence the splicing and polyadenylation machinery . An interesting point to note here is that out of five different kinds of AS events , RI events are unique as retained introns can greatly influence the folding of nascent polypeptides . A shift in the RI event post-infection could therefore potentially activate the unfolded protein response ( UPR ) , which is now increasingly getting recognized as yet another arm of host innate defense mechanism [66 , 67] . We believe further investigations on this line would add an exciting dimension to the host-pathogen interactions during Mtb infection of macrophages . We did not identify any potential regulator of global splicing either from the host or the pathogen in this study . Specific splicing factors are increasingly getting recognized for their anti-cancer properties [63] . Thus there appears enormous opportunity to identify new targets , both from the host and pathogen , for developing potential anti-TB drugs . In conclusion , we report here global perturbation of alternate splicing and alternate polyadenylation upon Mtb infection in the macrophages . As is the case with other known perturbations like regulation of autophagy , apoptosis , gene regulation , phagosome maturation arrest etc . [5 , 41 , 42 , 53 , 68] , Mtb is also able to extract benefit out of the key cellular function of alternate splicing . A better understanding of how Mtb can achieve these regulations could result in designing novel approaches for intervention .
All experiments involving human primary macrophages from healthy volunteers was approved by the Institutional Ethics Committee ( approval number: ICGEB/IEC/2016/03 ) THP-1 derived macrophages were obtained by treating THP-1 cells with 32 nM PMA in 10% FBS in RPMI 1640 for 24 hours followed by its removal and another 24 hours in 10% FBS in RPMI1640 . The THP-1 were then infected with H37Ra and H37Rv strains of Mtb for 4 hours in 10% FBS in antibiotic free RPMI 1640 followed by plain RPMI wash and another 2 hours in 200 μg/ml amikacin to kill any leftover extracellular bacilli . Amikacin was then washed and the cells were kept in 10% FBS in antibiotic free RPMI 1640 for the indicated time points . Media was replaced after every 24 hours . See the S10 Table . RNA was extracted from Mycobacterium tuberculosis infected THP1 cells using MDI RNA Miniprep kit ( MTRK250 ) according to manufacturer’s guidelines . Briefly , 100 ng of total RNA was used to prepare amplified cDNA using Illumina TruSeq Kit as per Manufacturer recommended protocol . The produced double-stranded cDNA was subsequently used as the input to the Illumina library preparation protocol starting with the standard end-repair step . The end-repaired DNA with a single ‘A’- base overhang is ligated to the adaptors in a standard ligation reaction using T4 DNA ligase and 2 μM-4 μM final adaptor concentration , depending on the DNA yield following purification after the addition of the ‘A’-base . Following ligation , the samples were purified and subjected to size selection via gel electrophoresis to isolate 350 bp fragments for amplification in preparation for cluster generation . The prepared cDNA library was sequenced for 101-bp paired-end reads using the Hiseq 2000 platform . The image analysis , base calling and quality score calibration were processed using the Illumina Pipeline Software v1 . 4 . 1 according to the manufacturer’s instructions . Reads were exported in FASTQ format and has been deposited at the NCBI Sequence Read Archive ( SRA ) under accession number SRA047025 . Paired end RNA seq reads each of 101bp length from each time point were mapped independently against human genome build hg19 downloaded from Ensemble ( http://asia . ensembl . org ) using tophat version 2 . 0 . 12 ( http://tophat . cbcb . umd . edu ) with the following options “-p 24 -G Human_ENSEMBL_Coding . gtf ˮ where Human_ENSEMBL_Coding . gtf contains the Ensemble coding transcripts in GTF file format . No novel junctions or novel insertion-deletion were taken in account by passing the parameter “-no-novel-juncˮ and “-no-novel-indelˮ respectively . Gene and isoform level expression were calculated by using isoform expression method by running cuffdiff ( http://cufflinks . cbcb . umd . edu/ ) on the alignment files from tophat and Ensemble coding genes . For isoforms , multiple mapping reads and sequence bias were corrected by using “-uˮ option of cuffdiff . FPKM ( fragments per kilobase of transcript per million mapped read ) was used to normalize the variability in number of reads produced for each run and RNA fragmentation during library construction . The count variance was modeled as non linear function of mean counts using negative binomial distribution . Differential exon splicing pattern and inclusion level was modeled for each sample compared to uninfected sample using robust Bayesian statistical framework MATS ( multivariate analysis of transcript splicing ) . Exon-exon junction database was constructed from the Ensembl transcript annotation ( release 64 ) file to be used in MATS . A threshold of 0 . 5 was taken as cutoff to identify significant inclusion level of exons between the samples . Switch like events were considered where inclusion level difference was perfect 1 . Further p-value and FDR ( False discovery rate ) were determined by Markov chain Monte Carlo ( MCMC ) method by simulating over samples . The quantitative visualization of splice junction of the genes showing significant psi score difference was done in IGV ( Integrated genomic viewer ) along with uninfected sample as sashimi plot . Dynamic APA ( alternative polyadenylation ) usage by each sample compared to uninfected sample was identified by the DaPars ( De novo identification of dynamic APA ) algorithm . To explain the localized read density change DaPars employs a linear regression model to determine the optimal fitting point . Alignment files were converted to wig file format using RSEM software . PDUI ( percentage distal usage index ) difference of 0 . 3 with FDR < 0 . 05 which passed the filter were considered as significant . The filtered result was visualized using integrated genomic viewer software . cDNA was made with iScript cDNA synthesis kit ( BioRad #170–8891 ) using mix of random hexamers and oligo-dT primers . qRT-PCR was performed using SsoFast EvaGreen Supermix ( BioRad #1725201 ) and transcript expression were normalized to actin or 18s rRNA . A list of all primers used for RT-PCR in this study is provided in S10 Table . 1 . 5X106 cells per well were plated for western blotting experiments . After mentioned time point , cells were washed with ice cold PBS before their incubation with Buffer A solution ( 20mM HEPES , 10mM NaCl , 1 . 5mM MgCl2 , 0 . 2mM EDTA and 0 . 5%v/v Trition-X-100 ) with 1X Protease Arrest ( G-Biosciences ) for 15 minutes on ice for lysis . Cell lysate was centrifuged at 4°C at 6000g for 10 minutes and supernatant was collected . Protein quantification was done using BSA as standard in Bradford’s assay . Protein sample was mixed with 6X loading dye and subjected to SDS PAGE and transferred to nitrocellulose membrane for immunoblotting . Blocking was done for an hour with Odessey blocking buffer ( LI-COR Biosciences ) in 1:1 dilution with 1X PBS at room temperature . Blots were immunoblotted with primary and then with secondary antibody made in blocking buffer . Blots were imaged with Odessey Infra Red Imaging system ( LI-COR Biosciences ) . Long ( ENST00000321437 ) and short ( ENST00000558990 ) form of Rab8B transcript were amplified from the cDNA and cloned in pMSCV PIG ( Puro IRES GFP ) ( www . addgene . org/21654 ) between XhoI and EcoRI site of MCS . HEK293T cells were maintained at 30–50% confluency overnight before transfection in complete 10% FBS DMEM media at 5% CO2 and 370 C in a humidified chamber . Lentiviral transfer plasmid was mixed with two packaging plasmid encoding Gag , Pol and other Rev genes in jetPRIME buffer ( Polyplus #712–60 ) in the ratio 3:1:1 . Final transfection mix was prepared by adding jetPRIME at a concentration of 1μl/μg DNA , vortexed for 10 seconds and incubated for 10 minutes at room temperature . Transfection mix was added dropwise and incubated for 48 hours before lentiviral harvest . Supernatant was filtered with 0 . 45μm filter and concentrated using Retro-X concentrator ( Clontech #631456 ) . 1X106 THP1 cells were seeded in each well of 6-well plate in 500μl complete RPMI 1640 media and transduced by lentiviral vectors at a multiplicity of infection of 10:1 in the presence of 8μg/ml polybrene . Spinoculation was done for 1 hour at 1200g . Media was replenished with complete media after 6 hours of transduction . For immunofluorescence experiments 20X104 cells were plated on autoclaved coverslips one per well in a 24 well plate . Cells were infected with PKH67 labeled or GFP expressing bacteria . The cells were fixed at their requisite time points using 4% v/v paraformaldehyde in 1X PBS for 20 minutes at room temperature . The cells were then incubated with blocking buffer for immunostaining ( 0 . 02% Triton-X100 , 3% w/v BSA in 1X PBS ) for 1 hour . The cells were then washed thrice with 1X PBS for 5 minutes each and incubated with 1:100 anti-Rab8B antibody in blocking buffer for 3 hours . Cells were washed twice with 1X PBS each for 5 minutes and treated with 1:300 anti-rabbit 568nm for 1 hour . The cells were then washed thrice with 1X PBS for 5 minutes each before mounting the coverslips on slides with Pro-Long anti-fade reagent ( Life Technologies ) . All the incubation and staining was done at room temperature . NIS-Elements ( Nikon ) was used for acquiring the images as Z-stacks using Nikon EclipseTi-E laser scanning confocal microscope equipped with a 60X/1 . 4 NA Plan Apochromat DIC objective . For FACS experiments 1 X 106 transduced THP1 macrophages were plated per well in a 6 well plate and infected with Far-red labeled bacteria . The time and staining concentrations of Far-red dye was performed as per manufacturer’s directions . The cells were fixed at their requisite time points using 4% v/v paraformaldehyde in 1X PBS for 20 minutes at room temperature . Cells were scrapped and data was acquired using BD FACSDiva acquisition software in BD FACS Canto II flow cytometer ( BD ) . Data was analyzed and plotted using R packages flowcore and flowViz . Heparinized blood was immediately diluted in 1:1 ratio by volume with DPBS . Diluted blood was layered on Ficoll-paque ( Himedia ) and centrifuged at 2000 rpm for 45 min . Interface containing PBMC was isolated carefully and washed twice with DPBS . Cells were diluted in RPMI1640 media containing 10% FBS to a concentration of 1x106 cells/ml . Cells were put in a 6 well tissue culture plate and incubated for 3 hours in a humidified 5% co2 chamber at 37°C . Non adherent cells were removed and two washes with RPMI was done . Complete media containing 5ng/ml recombinant human M-CSF ( R&Dsystems , 216-MC/CF ) was added and cells were allowed to differentiate for 4 days into macrophages in a humidified 5% co2 chamber at 37°C . siGenome human Rab8B siRNA ( SmartPool of 4 different siRNA , M-008744-01-0005 ) was obtained from Dharmacon ( Dharmacon , GE Healthcare ) . Cells were transfected at final concentration of 50nM siRNA using Dharmafect-2 transfection reagent according to manufacturer's protocol . After indicated time period cells were lysed in 50ul of 0 . 06% SDS containing lysis buffer at room temperature for 10 min . 1:10 and 1:50 dilution of lysate was prepared and plated in duplicates on 7H11 agar plates supplemented with 10% OADC . Square plates ( 12cm x 12 cm ) were used for plating by track dilution method . 10μl of the diluted sample was spotted and allowed to flow by tilting the plate at 450 angle . Plates were dried and incubated for 18–20 days in a humidified incubator at 37°C before colonies were counted .
|
Eukaryotic gene expression is a complex process where several intermediary processing steps of transcripts are required before translation can take place . Change in gene expression is a fundamental means through which cells adapt to environmental cues or stimuli like infections . The professional phagocytes macrophages are known to undergo drastic alteration in gene expression upon inflammatory stimuli or infection with pathogens like Mycobacterium tuberculosis . However , whether the alterations in gene expression are also influenced by possible perturbations of one or more of the intermediate steps in RNA processing is not known . Here taking advantage of next-generation RNA sequencing approach , followed by robust computational analysis , we show infection of macrophages by Mycobacterium tuberculosis results in massive alterations in the splicing pattern in the host . Since alternate splicing can influence transcript stability , stability of the translated products , loss/gain of function , interacting partners and sub-cellular localization , its implications on host response to infection could be overwhelming . We subsequently performed targeted experiments to confirm that alternate spliced variants of host genes indeed helped infecting Mtb strains to survive better within the macrophages . Regulators of splicing , either from host and/or pathogen , therefore , constitute an attractive set of targets to develop novel therapeutic strategies to control tuberculosis .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"blood",
"cells",
"medicine",
"and",
"health",
"sciences",
"immune",
"cells",
"vesicles",
"gene",
"regulation",
"immunology",
"phagosomes",
"infectious",
"disease",
"control",
"cellular",
"structures",
"and",
"organelles",
"bacteria",
"small",
"interfering",
"rnas",
"infectious",
"diseases",
"white",
"blood",
"cells",
"animal",
"cells",
"gene",
"expression",
"actinobacteria",
"biochemistry",
"rna",
"spliceosomes",
"cell",
"biology",
"nucleic",
"acids",
"mycobacterium",
"tuberculosis",
"genetics",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"macrophages",
"non-coding",
"rna",
"organisms"
] |
2017
|
Alternate splicing of transcripts shape macrophage response to Mycobacterium tuberculosis infection
|
A pervasive case of cost-benefit problem is how to allocate effort over time , i . e . deciding when to work and when to rest . An economic decision perspective would suggest that duration of effort is determined beforehand , depending on expected costs and benefits . However , the literature on exercise performance emphasizes that decisions are made on the fly , depending on physiological variables . Here , we propose and validate a general model of effort allocation that integrates these two views . In this model , a single variable , termed cost evidence , accumulates during effort and dissipates during rest , triggering effort cessation and resumption when reaching bounds . We assumed that such a basic mechanism could explain implicit adaptation , whereas the latent parameters ( slopes and bounds ) could be amenable to explicit anticipation . A series of behavioral experiments manipulating effort duration and difficulty was conducted in a total of 121 healthy humans to dissociate implicit-reactive from explicit-predictive computations . Results show 1 ) that effort and rest durations are adapted on the fly to variations in cost-evidence level , 2 ) that the cost-evidence fluctuations driving the behavior do not match explicit ratings of exhaustion , and 3 ) that actual difficulty impacts effort duration whereas expected difficulty impacts rest duration . Taken together , our findings suggest that cost evidence is implicitly monitored online , with an accumulation rate proportional to actual task difficulty . In contrast , cost-evidence bounds and dissipation rate might be adjusted in anticipation , depending on explicit task difficulty .
Suppose that you are given a job whose payoff is proportional to the effort made within a limited time , say for instance the number of Christmas cards sold at the end of the day . Maximizing your payoff would require running from house to house , but this effort would induce such fatigue that you decide to walk from time to time . This sort of situation can be examined through economic decision theory , which would suggest you to write down the expected costs and benefits , and try to figure out whether the effort is worthy . If the cost of a given effort is anticipated to increase with fatigue [1] , [2] , then you will find an optimal duration that can be determined before engaging any action . Yet the literature on exercise performance has developed a different perspective on this issue [3] , [4] , which would suggest that you start by running , and only stop when some physiological variable , for instance in cardiovascular function ( such as heart beat rate ) or in muscular metabolism ( such as lactate concentration ) , attains a given limit [5] , [6] . In other words , effort cessation would be a reaction to homeostatic failure , and would not require any explicit anticipation of effort cost . These two extreme perspectives have obvious limitations . The physiological view does not account for the effect of expectations that might pre-configure behavioral performance [4] , [7] , [8] . The economic view does not integrate the constraints imposed by physiological reactions , which might be difficult to anticipate [9] . Here , we intend to overcome these limitations by integrating the two perspectives into the same computational model . Furthermore , we have built this model so as to explain the duration not only of effort exertion but also of rest ( recovery from fatigue ) . Let us assume that a single waning and waxing variable triggers decisions to stop and restart effort exertion when reaching bounds ( see Figure 1A for a graphical presentation ) . As this variable linearly accumulates during effort and dissipates at rest , it can be seen as a simple reflection of physiological reactions that predict the proximity of homeostatic failure . Alternatively , it can be interpreted as tracking cost increase with fatigue , by integrating past effort over time . Thus , the basic architecture of the model ( the accumulation-to-bound principle ) can account for implicit , online adaptation to actual effort costs , complying with physiological constraints . On this basis , the modulation of the model latent parameters ( slopes and bounds ) could allow for anticipatory adjustments , depending on explicit costs and benefits ( see Figure 1B for a graphical illustration ) . To dissociate the effects of actual and expected effort costs , we developed seven variants of a paradigm that was employed in a previous paper [10] to identify the neural underpinnings of the modeled variable , which we termed cost evidence ( see Figure 2 for an overview ) . The task involved participants squeezing a handgrip with a given force , knowing that their payoff will be proportional to their effort duration . Cost evidence can be manipulated by varying either an imposed duration or an imposed force ( task difficulty ) . In a first study , we used three tasks that impose variable durations in order to verify that the behavior is adapted on the fly due to internal constraints ( bounds ) . In a second study , we demonstrate that explicit ratings of subjective exhaustion do not follow the cost-evidence variable that accounts for the decision to stop effort exertion . In a third study , we used three other tasks that vary the difficulty in order to dissociate the effects of expected and actual costs .
In our previous paper [10] , we suggested that the alternation of effort and rest periods observed in the Effort Allocation Task was well explained by a waning and waxing accumulation signal . However , this cost-evidence signal that we localized in the brain could be epiphenomenal , in the sense that it would not reflect any causal mechanism triggering the decisions to stop and restart effort . In this first study , we wished to verify that the level of cost evidence imposes actual constraints on subsequent behavior , as predicted by the accumulation-to-bound principle . We therefore tested the predictions of the accumulation model on the behavior that followed an effort whose duration was imposed . The difficulty was not manipulated in this study , for two reasons: firstly , the effect of difficulty was already shown in the previous paper [10] and will be further investigated in the following studies , and secondly , manipulation of difficulty only applies to effort periods , whereas manipulation of duration can be equally applied to both effort and rest periods . Predictions of the accumulation model are that 1 ) prolonging effort should decrease the next effort period ( if compensatory resting is not allowed ) , 2 ) prolonging rest should increase the next effort period ( up to a maximum corresponding to full recovery ) , and 3 ) prolonging effort should increase the next rest period ( if compensatory resting is allowed ) . These three predictions were tested in different groups of participants ( n = 36 in total ) , using three variants of the Effort Allocation Task . These three Adaptation Tasks had the same structure , with first an imposed effort ( between start and stop signals ) , second a rest period ( either fixed or free ) and third a free effort exertion . Difficulty of both efforts was fixed at 60% of the maximal force , and payoff was proportional to the duration of the last effort , which was the main dependent measure . Data were regressed at the individual level against a linear model that included the factor of interest ( the imposed duration ) and several potential confounds ( see methods ) . The statistical significance of regressors was estimated at the group level using two-sided one-sample t-tests . Results are given as standardized effect size ( beta ) ± inter-subject standard error of the mean . In this task , cost evidence was increased by prolonging the first effort period ( from 1 to 10s ) , then the second effort duration was observed after a fixed 2-s rest ( Figure 3A ) . To ensure that the rest duration was well controlled , we checked that initiation delay of the second effort after the go signal was not significantly impacted by the duration of the first effort ( 6 . 0 10−2±3 . 2 10−2 , df = 11 , p = 0 . 09 ) , by cumulated duration of efforts produced in the current session ( 1 . 1 10−2±3 . 4 10−2 , df = 11 , p = 0 . 76 ) , and by the session number ( −2 . 8 10−2±2 . 3 10−2 , df = 11 , p = 0 . 25 ) . Critically , the second effort was significantly shortened by prolonging the first effort ( −8 . 29 10−1±2 . 3 10−1 , df = 11 , p = 0 . 0037 ) . Next we examined the shape of the transfer function from imposed to observed effort duration . The model predicts that this link should be negative , except if resting is long enough to fully dissipate the accumulated cost . We therefore compared a model with pure negative correlation ( no saturation , #1 ) to models with an upper plateau ( over shortest efforts ) , followed by a decrease . We tried two possibilities for this saturation effect: first a constant followed by a linear decrease ( #2 ) and second a negative exponential ( #3 ) . The latter was implemented because it provides a better fit of plateau effects when data are noisy ( see methods ) . Bayesian model selection revealed that the pure linear model was far better than the two saturation models in the family comparison ( model 1 versus models 2 & 3 ) , with an expected frequency ef = 0 . 81 ( which is much higher than chance level - 1/2 ) and an exceedance probability xp = 0 . 96 ( confidence that the model is more frequently followed than the others ) . Thus , the result supports linear accumulation of cost evidence , which limits subsequent effort production due to the existence of an upper bound . However , we found no evidence for the existence of a lower bound in cost dissipation , probably because our rest period was not long enough . This limitation was overcome in the next task , where rest period was systematically varied . This task ( Figure 3B ) was very similar to Task 1 , except that effort duration was now fixed ( to 7 s ) and rest duration was systematically varied ( from 1 to 12 s ) . We checked again that subjects were not delaying effort initiation to compensate for variations in the imposed rest duration ( −3 . 7 10−2±2 . 1 10−2 , df = 11 , p = 0 . 10 ) . In addition we found that the initiation delay was slightly affected by the cumulated duration of past efforts ( 1 . 2 10−2±5 . 0 10 −3 , df = 11 , p = 0 . 03 ) , but not by the session number ( −1 . 1 10−2±1 . 4 10−2 , df = 11 , p = 0 . 46 ) . Critically , observed effort was significantly prolonged by longer rest ( 6 . 9 10−1±1 . 9 10−1 , df = 11 , p = 0 . 0035 ) . Next we tested the existence of a saturation , meaning that beyond a certain rest duration , cost evidence is entirely dissipated and subsequent effort cannot be further prolonged . As was done for the previous task , we compared three models for the link between rest and effort duration: 1 ) a linear effect ( no saturation ) , 2 ) a linear effect bounded by an upper plateau ( over longest rests ) , 3 ) an exponential asymptotic plateau . Bayesian model selection showed that the saturation family was now more plausible ( models 2 and 3 versus model 1 , chance level is 1/2 , ef = 0 . 79 , xp = 0 . 94 ) . Direct comparison between models 2 and 3 revealed that the asymptotic saturation was more likely than the linear plateau ( xp = 0 . 98 ) . Thus , the results confirmed that prolonging rest after a first effort augments the capacity to produce a second effort , as if cost evidence was dissipated . Moreover , the saturation effect suggests the existence of a threshold after which prolonging rest is useless , which would correspond to a lower bound for cost-evidence dissipation . This task ( Figure 3C ) was quite similar to Task 2 , except that participants were not asked to resume their effort immediately at the go signal , but only when they felt ready to do so . There were therefore two dependent variables of interest: rest duration and subsequent effort duration . Critically , rest duration was significantly increased by prolonging the imposed effort duration ( 6 . 5 10−1±1 . 3 10−1 , df = 11 , p = 0 . 0005 ) . We expected that participants would rest long enough to fully dissipate the first effort cost , which hence would have no impact on the second effort duration . This was not the case: prolonging the first effort significantly shortened the second effort ( −4 . 7 10−1±1 . 4 10−1 , df = 11 , p = 0 . 006 ) . Thus , subjects did not wait long enough to compensate for the imposed effort cost . This partial recovery might be related to the fact that the total time allowed for rest and effort was limited to 20s , so that participants may have shortened rest to make sure there would be enough time for effort ( even if in reality , 20s was largely enough to fully dissipate and accumulate cost again ) . So far , our results suggest that effort duration is not entirely planned in advance but adapted on the fly so as to keep cost evidence within pre-defined bounds . The next study was designed to assess whether our participants could explicitly report the cost evidence that was monitored by their brain in order to regulate their behavior . The first study only manipulated the duration of effort or rest periods . Yet our model posits that cost evidence accumulation during effort depends on task difficulty . Therefore , cost-evidence level should reflect the interaction of task difficulty and effort duration . The logic of this second study was first to examine whether introspective reports would reflect the interaction of difficulty and duration , and then to verify that behavioral choices were indeed driven by this interaction , For introspective reports we asked a new group of 18 participants to perform a Cost Rating Task , in which they had to rate their degree of exhaustion after effort exertion . Note that we could have directly inserted cost ratings within the Effort Allocation Task , but subjects in this case might have artificially aligned their behavior to their explicit reports ( or vice-versa ) . Another issue with this possibility was that effort duration would not have been sufficiently varied , at least not orthogonally to effort difficulty , since subjects would have stopped their effort when cost evidence ( difficulty times duration ) reached a pre-defined bound . We chose to frame the question in terms of exhaustion because debriefing of previous studies revealed that exhaustion is the intuitive term that subjects spontaneously use to describe the sensation that makes them cease their effort . The precise question was ‘Have you exhausted your resources ? ’ and the response scale was ranging from ‘not at all’ to ‘completely’ . In this Cost Rating Task , both effort duration ( from 3 to 7 s ) and task difficulty ( from 40 to 60% of maximal force ) were imposed and varied experimentally ( Figure 4A ) . To keep similarity with the Effort Allocation Task , we also manipulated the incentive level . Yet we acknowledge that the comparison between tasks has limitations , first because they implement different range of forces and durations , second because they are performed by different subjects , who might have different sensitivity to effort cost . On each trial , the payoff was calculated as the incentive multiplied by the fraction of the imposed duration that subjects spent squeezing at the required target force level or higher . As participants were asked to be as accurate as possible , this fraction was almost 100% ( mean over subjects: 98 . 7% , extreme subjects: 94 . 6% and 99 . 9% ) . The difference between required and produced force levels did not vary significantly across conditions ( multiple regression analysis and two-sided t-test with df = 17; incentive: 4 . 1 10−3±3 . 7 10−3 , p = 0 . 28; duration: −3 . 4 10−6±4 . 6 10−3 , p = 0 . 99; difficulty: −1 . 8 10−3±3 . 3 10−3 , p = 0 . 59; interactions between these factors: all p>0 . 21 ) , suggesting that effort production was well controlled by the experimental design . Cost ratings were not significantly impacted by incentives ( 1 . 4±0 . 86 , df = 17 , p = 0 . 1 ) , and marginally by the initial position of the cursor on the scale ( 1 . 8±0 . 9 , df = 17 , p = 0 . 056 ) . Critically , cost ratings increased with both duration ( 1 . 9±0 . 79 , df = 17 , p = 0 . 028 ) and difficulty ( 3 . 2±0 . 49 , df = 17 , p = 5 10−6 ) , without significant interaction between these factors ( p = 0 . 96 ) . We then fitted cost ratings with a linear combination of regressors meant to capture the impact of duration and difficulty . We considered three possibilities: main effects of duration and difficulty , non-linear effects ( power functions ) of duration and difficulty , and interaction between duration and difficulty . Including or not each possibility in the linear combination made a total of eight models , which we compared using Bayesian model selection ( Figure 4C ) . This analysis confirmed the absence of significant interaction between duration and difficulty , since the best model was simply additive ( chance level is 1/8 , ef = 0 . 48; xp = 0 . 93 ) . In principle , this additive effect could arise from half the subjects reporting duration and the other half reporting difficulty . This would imply that the effect sizes of these factors are anti-correlated across subjects . We found the opposite result ( Pearson rho: 0 . 82 , df = 16 , p = 3 10−5 ) , suggesting that subjects who were good at perceiving duration were also good at perceiving difficulty . Yet they reported the addition of the two dimensions , and not their product , as should be the case if they were simply introspecting cost evidence . We next re-analyzed the behavioral choices observed in our Effort Allocation Task ( Figure 4B ) that involved subjects ( n = 38 ) squeezing a handgrip in order to accumulate as much money as possible [10] . The payoff was calculated as the monetary incentive multiplied by the time spent above a target force level ( which indexed task difficulty ) . Both the incentive ( 10 , 20 or 50 cents ) and difficulty levels ( 70 , 80 or 90% of maximal force ) were varied across trials such that we could assess their effects on effort allocation . Incentive levels were sufficient for subjects to initiate the effort and to reach the target , but difficulty levels were too demanding for subjects to sustain their effort throughout trials , which lasted 30 seconds . Instead , they freely alternated effort and rest periods within trials ( as can be seen in Figure 1A ) . We used the normalized cumulative distribution of effort durations to calculate the probability of stopping the effort after a given duration at a given difficulty level . This probability was fitted with a sigmoid function of cost-evidence level , which accounts for higher cost evidence making effort cessation more likely . Cost evidence was then modeled with the same linear combinations as used for fitting cost ratings . Results of Bayesian model selection ( Figure 4D ) showed that the most plausible model was pure interaction ( chance level is 1/8 , ef = 0 . 62 , xp = 0 . 988 ) . The Cost Rating and Effort Allocation tasks thus elicited distinct forms of cost evidence , with additive versus multiplicative effect of effort difficulty and duration . The critical difference is the shape of iso-value lines of cost evidence in the duration by difficulty space , with straight lines for explicit report and convex lines for effort cessation ( compare Figures 4E and 4F ) . To directly compare the curvature of cost evidence inferred from introspective reports and behavioral choices , we fitted a model with constant elasticity of substitution ( CES ) between duration and difficulty ( see methods ) . This model has a free parameter that captures the curvature of cost in the duration by difficulty space , which should be equal to one in the absence of interaction , and below one in the case of a convex interaction . We found that the curvature parameter was significantly below one in the Effort Allocation Task ( median: 0 . 52 , SEM: 0 . 06; two-sided sign-test of the median against 1: p = 6 . 7 10−8 ) but not in the Cost Rating Task ( median: 1 . 01 , SEM: 0 . 12; sign-test of the median against 1: p = 1 ) , with a significant difference between tasks ( p = 3 10−6 , two-sided Wilcoxon rank sum test for equal medians ) . When debriefing the Cost Rating Task , participants unambiguously reported having noticed variations in both difficulty and duration . When asked whether one of these two factors had a greater impact on their ratings , 13 subjects favored the duration , 3 favored the difficulty , and 2 could not favor one or the other , describing something like an interaction . However , comparison of standardized effect size revealed a greater impact of difficulty on ratings ( paired t-test on duration minus difficulty effect size: −1 . 3±0 . 48 , df = 17 , p = 0 . 016 ) . Among the 16 subjects who favored a main effect , 12 got it wrong ( the other factor had a higher impact on their ratings ) , which is more than expected by chance ( binomial test , p = 0 . 028 ) . To summarize , the costs reported in subjective ratings do not have the same shape as the costs inferred from behavioral choices . What subjects report is an addition of duration and difficulty , whereas what drives their behavior is an interaction between the two . Furthermore , at a meta-cognitive level , subjects have poor insight into the factors that modulate their sensation of exhaustion . The two studies presented so far are compatible with a completely implicit and automatic model , in which decisions to cease and resume effort production are controlled by an internal variable fluctuating between bounds that might be determined by physiological constraints . In this last study , we explored whether explicit information about cost could impact the mechanics driving decisions to start and stop effort exertion . In our previous paper [10] , we had observed that task difficulty shortened effort duration , which could reflect cost evidence ( difficulty times duration ) reaching the upper bound , but did not affect rest duration . We hypothesized that the last observation could arise from task difficulty not being made explicit to participants . Indeed , monetary incentives , contrary to difficulty levels , were explicitly presented with coin images at trial start and affected both effort and rest durations ( with longer effort and shorter rest for higher incentive ) . We therefore tested whether providing explicit information about difficulty level would change the way participants process cost evidence . We constructed three variants of the Effort Allocation Task , which were administered to three different groups of participants ( n = 67 in total ) . In all tasks , incentives ( coin images ) were explicitly displayed before and during trials , which had a fixed duration ( 30s ) that was specified to participants prior to the experiment . The Implicit Task ( Figure 4B ) is the task used in our previous paper [10] , with no visual cue for difficulty level . In the Explicit Task , the only change is that difficulty level ( percentage of maximal force: 70 , 80 or 90% ) was announced before the beginning of trials , on the same screen as incentive level . In the Dissociation Task , we kept the explicit cues , but they were no longer predictive of the actual task difficulty . To maintain sufficient statistical power , only two difficulty levels were used ( 75 and 85% ) , in a full factorial design ( two cued difficulties crossed by two actual difficulties ) . This design was meant to disentangle the effects of implicit versus explicit cost processing . Monetary incentives were also manipulated in all tasks and crossed with the three ( Implicit and Explicit Tasks ) or four ( Dissociation task ) cells corresponding to variations in difficulty . We only used two incentive levels ( 10 versus 20c ) in the Dissociation task to avoid combinatorial inflation . In every task , the effect of experimental factors ( incentive , actual and cued difficulty ) on the duration of effort and rest epochs were estimated in separate multiple linear regressions followed by two-sided one-sample t-tests . Note that because they must add up to 30s , the cumulative durations of effort and rest are anti-correlated . However , this dependency was broken first because the last rest epochs were discarded from the analysis , since they are interrupted by trial ending , and second because we considered the single epoch durations , which are not predictable from the cumulative durations , since they depend on the number of alternations between effort and rest . The remaining correlation was rather low ( Pearson rho: −0 . 15±0 . 026 in the main Implicit Effort Allocation Task ) and probably due to opposite effects of experimental factors ( see below ) . As previously shown [10] , in the Implicit Task ( Figure 5 , left ) , effort duration was both longer for higher incentive ( 1 . 5±0 . 26 , df = 37 , p = 8 . 1 10−7 ) and shorter for higher difficulty ( −1 . 1±0 . 13 , df = 37 , p = 1 . 6 10−10 ) . In contrast , rest duration was shorter for higher incentive ( −0 . 37±0 . 08 , df = 37 , p = 2 . 0 10−5 ) but was not modulated by the difficulty ( 0 . 03 , ±0 . 03 , df = 37 , p = 0 . 32 ) . Interactions were included in the regression model , but the incentive x difficulty interaction was not significant , neither for effort or for rest duration ( all p>0 . 084 ) . All significant results were replicated in the Explicit Task ( Figure 5 , middle ) : effort duration was both longer for higher incentive ( 2 . 2±0 . 53 , df = 13 , p = 1 . 1 10−3 ) and shorter for higher difficulty ( −1 . 8±0 . 24 , df = 13 , p = 6 . 0 10−6 ) , and rest duration was shorter for higher incentive ( −0 . 4±0 . 09 , df = 13 , p = 9 . 7 10−4 ) . The novel result is that rest duration was now increased by higher difficulty ( 0 . 31±0 . 08 , df = 13 , p = 1 . 6 10−3 ) , which was correctly cued at trial start . The difference in standardized effect sizes between Implicit and Explicit Tasks was also significant ( p = 1 . 2 10−4 , unpaired t-test , df1: 37 , df2: 13 ) . All interactions remained non-significant , neither for effort or rest duration ( all p>0 . 1 ) . Thus , the difficulty in the Explicit Task , which was both expected and experienced during effort exertion , affected both effort and rest durations . The results obtained with the Implicit and Explicit Tasks are compatible with the actual difficulty affecting effort duration , and the expected difficulty affecting rest duration . In the Implicit Task , there was no explicit cue , so subjects did not expect any particular difficulty level , and consequently only effort duration ( not rest duration ) was affected by task difficulty . In the Explicit Task , both effort and rest durations were modulated because the actual difficulty was fully expected . However , as the explicit cues were perfectly valid , we could not formally demonstrate with this task that rest duration is not concerned with actual difficulty , or that effort duration is not concerned with expected difficulty . To complete our demonstration , we intended to dissociate the two effects within the same task . In the Dissociation Task ( Figure 5 , right ) , the levels of actual and cued difficulty were manipulated independently . As in the Implicit and Explicit tasks , higher incentive increased effort duration ( 0 . 42±0 . 16 , df = 14 , p = 0 . 022 ) and shortened rest duration ( −0 . 22±0 . 06 , df = 14 , p = 1 . 5 10−3 ) . Effort duration was affected by the actual ( −0 . 47±0 . 18 , df = 14 , p = 0 . 021 ) but not by the cued difficulty ( 0 . 07±0 . 15 , df = 14 , p = 0 . 64 ) . The difference in standardized effect size was at significance limit ( −0 . 54±0 . 25 , df = 14 , p = 0 . 050 , paired t-test ) . We also verified that the effect of cued difficulty on effort duration in the Dissociation Task was significantly lower than the ( actual ) difficulty effects observed in the Implicit ( p = 4 . 3 10−7 , unpaired t-test , df1: 37 , df2: 14 ) and Explicit ( p = 2 . 3 10−6 , unpaired t-test , df1: 14 , df2: 13 ) tasks . Conversely , rest duration was affected by the cued ( 0 . 22±0 . 06 , df = 14 , p = 1 . 7 10−3 ) but not by the actual ( 0 . 03±0 . 06 , df = 14 , p = 0 . 63 ) difficulty . The difference in standardized effect size was as well significant ( −0 . 19±0 . 08 , df = 14 , p = 0 . 045 , paired t-test ) . We also verified that the effect of cued difficulty on rest duration was higher in the Dissociation Task than the ( actual ) difficulty effect observed in the Implicit Task ( p = 0 . 002 , unpaired t-test , df1: 37 , df2: 14 ) , and that the effect of actual difficulty in the Dissociation Task was lower than the ( cued ) difficulty effect observed in the Explicit Task ( p = 0 . 008 , unpaired t-test , df1: 14 , df2: 13 ) . Thus , within- and between-task comparisons both support a double dissociation between the actual and cued difficulty effects on effort and rest durations . As some critical p-values were near 0 . 05 type I error rate , we conducted a permutation test to ensure the reliability of the parametric t-distribution in our small sample . This permutation-based t-distribution yielded the same exact p-values up to the 3rd decimal . Second and third order interaction terms between incentive , cued and actual difficulty were included in the model , but none of them was significant neither for rest or effort duration ( all p>0 . 18 ) . We also checked that there was no interaction of cued difficulty with time , which could potentially reflect a progressive discount of the cue effect ( as subjects would learn that cues are not predictive of actual difficulty ) . Time was modeled at three nested scales ( rest or effort period position within a trial , trial position within a session , and session number ) . Two-way interactions with cued difficulty were estimated for each time scale: none of them was significant ( all p>0 . 25 ) . We compared different versions of our accumulation model to identify how the latent parameters ( A: amplitude between bounds , SE: accumulation slope during effort , and SR: dissipation slope during rest ) were affected by the experimental factors ( I: Incentive , Da: actual difficulty , Dc: cued difficulty ) . We started with the formalization that we proposed in our previous publication [10] to account for the behavior observed in the Implicit Task . All models were built as a set of three equations that defines each latent parameter as a linear combination of the different factors ( see methods ) . Only models that can produce the behavioral results ( significant effect on effort or rest duration ) were included in the space covered by Bayesian Model Selection . In the Implicit Task , this left 24 possible models ( see Figure 6A ) with one that was much more plausible than the others ( chance level is 1/24 , ef = 0 . 30 , xp = 0 . 90 ) . For the novel tasks ( Explicit and Dissociation ) , we explored two possibilities for integrating the additional factor ( cued difficulty ) . The first possibility was to integrate it as an additive term , just as was done with actual difficulty ( see Figure 6B and 6C ) . Note that these purely linear models do not enable dissociating the effects of actual and expected difficulty in the Explicit Task . The second possibility was to integrate cued difficulty as a hyperbolic discounter of incentives , which is quite standard in the literature for capturing temporal discounting [11]–[13] . Thus , for the novel tasks that manipulate expected difficulty , we included the hyperbolic equivalent of our linear models ( see Figure 6D ) . With this hyperbolic version , we can dissociate the effect of actual and expected difficulty ( the former is linear , the latter hyperbolic ) even in the Explicit Task where the two factors are confounded . Family comparison revealed that there was far more evidence in favor of a hyperbolic rather than linear discount of incentives by cued difficulty , in both the Explicit and Dissociation tasks ( chance level is 1/2 , ef>0 . 91 , xp>0 . 999 ) . Among the 78 possible hyperbolic models , a best model was identified with xp = 0 . 90 ( chance level is 1/78 , ef = 0 . 13 ) in the Dissociation Task and with xp = 0 . 82 ( chance level is 1/78 , ef = 0 . 14 , ) in the Explicit Task . Crucially , the best hyperbolic model identified in the Explicit and Dissociation tasks was the same model , which also corresponded to the best model identified in the Implicit Task ( where modulation by cued difficulty is necessarily absent ) . This best model is written as follows ( Te and Tr being effort and rest duration , α , β , γ the coefficients and I , Da and Dc the incentive , actual difficulty and cued difficulty levels ) :A graphical interpretation of the model with a summary of the observed effects is provided in Figure 1 . In short , incentives impacted both the amplitude between bounds and the dissipation rate , resulting in longer effort and shorter rest for higher incentives . The effect of task difficulty was computationally dissociable: higher actual difficulty accelerated the accumulation , resulting in shorter effort , whereas higher expected difficulty slowed the dissipation , resulting in longer rest .
In our previous paper [10] , we addressed the issue of how the brain allocates effort production over time , in a situation where the payoff depends on the total effort duration . We found a neural signal that was ramping up and down during effort and rest periods and that could , in principle , trigger the decisions to stop and restart effort production . Here we provide evidence that the core accumulation-to-bound mechanism is reactive and implicit . Indeed , participants adapted their behavior on the fly when we implicitly manipulated both the duration ( Study 1 ) and the difficulty ( Study 3 ) of effort exertion . However , when asked to rate their degree of exhaustion ( Study 2 ) , subjects did not report the cost evidence signal that was shown to drive their behavior . In addition , we suggest that some latent parameters of the accumulation-to-bound process are susceptible to anticipatory adjustment based on explicit information . Indeed , we found that expected benefit and difficulty could modulate the distance between bounds or the dissipation rate during rest . The dissociation of implicit and explicit cost processing could reconcile the perspectives offered by sport physiology on the one hand , and economic theory of choice on the other hand . The implicit part of the model - monitoring cost evidence and triggering decisions when bounds are attained , accords well with the literature on exercise performance [4] , [8] . Although it was developed to explain how athletes pace their running on treadmills , we can borrow the notion that behavioral changes are reactions to physiological variables reaching homeostatic borders . Results of Study 1 show that bounds between which the cost-evidence signal fluctuates are true limits that determine the decisions to stop and restart effort exertion . On the contrary , the explicit part of the model – adjusting the behavior depending on expected benefit and difficulty – is consistent with the literature on value-based decision-making [14] , [15] . It is quite remarkable in Study 3 that the computational effect of actual ( implicit ) difficulty during effort was simply linear , as in a passive accumulation , whereas the effect of expected ( explicit ) difficulty was hyperbolic , as in economic models of temporal discounting [13] , [16] . It should be acknowledged that mixtures of anticipatory calculations and on-line adaptations are frequently used in motor control theory [17] , [18] , for instance to explain how movement trajectory can be adjusted to internal noise and to unexpected target displacement . However , these models have not integrated the conflict between costs and benefits until very recently [19] , [20] . Finally , we note that the perspectives offered by the literatures on exercise performance and value-based choice only explain the duration of effort; without further specification they say nothing about the duration of rest . Our model accounts for the timing of both effort and rest , within the same accumulation framework . Examining whether the accumulation mechanism is optimal or not would go beyond the scope of this study . It can nonetheless be seen has a heuristic mechanism that certainly has advantages . Physiologically , it ensures that effort production does not put the body at threat , avoiding for instance damage to the muscles . In this view the signal would indicate the likelihood of physiological damage , and the upper bound would implement a threshold on that risk . Economically , it ensures that costs do not overcome benefits . In this view , the signal would indicate the cost , and the upper bound the benefit of the potential effort at the next time point . Mixing predictive and reactive processes also presents advantages . Online monitoring of effort consequences allows refining cost estimation , which is usually uncertain beforehand , as in our implicit version of the task . Anticipatory estimation allows deciding whether or not to engage the action , and scaling energy expenditure to expected costs and benefits . In our case , this means spending more time at work and less time at rest when the net value of effort is higher . The two behaviors , effort and rest , are not equivalent though . While monitoring cost evidence during effort might be a passive process ( mechanically integrating difficulty over time ) , dissipating cost evidence during rest seems more active . Indeed , the dissipation rate was susceptible to modulation by explicit information ( monetary incentive and cued difficulty ) . Moreover , the observation that subjects do not report cost evidence was only made in Study 2 , relative to the effort period . It remains possible that during rest , subjects are fully aware of the cost-evidence level , and hence of how much effort they would be able to produce next . We could have tried to test whether their introspective reports integrate duration with cued difficulty after a given rest , but asking the question in this case would have been awkward . Using dissipation as well as accumulation in order to explain behavioral choices is a major difference between our model and the standard evidence accumulation models . Classically , accumulation of evidence is meant to improve the estimation of a stationary noisy input , whether external , as in perceptual decision-making [21]–[23] , or internal , as in value-based decision making [24]–[26] . The fact that the cost evidence variable dissipates at rest rules out the possibility that this signal simply reflects an integration of the force produced throughout the trial ( which can only increase ) . It is likely that the signal reflects an input that is already dynamical ( and not stationary ) . This might be true not only at the theoretical level , if we interpret it as signaling the potential effort cost or the proximity of exhaustion , but also at the biological level . For instance , our cost evidence signal could relay the accumulation and dissipation of a by-product of effort exertion , which could integrate several variables such as lactate concentration , stretch of muscle fibers or heart beat rate . Alternatively , the cost-evidence signal could reflect increase in the efferent drive needed to overcome fatigue and maintain motor output [27] . Using combined fMRI and MEG recordings , we localized the cost-evidence signal in proprioceptive areas ( posterior insula ) . This localization would incline us to situate the input in the afferent proprioception coming from the muscles [28] . However , the fact that subjects had a poor introspection into that signal argues against the idea that it represents the neural counterpart of a common and intuitive sensation such as fatigue . Yet the fact that cost evidence dissipation could be modulated depending on expected benefit and difficulty suggests that other neural processes occur during rest than passive transmission of effort-induced physiological perturbation . First , the dissipation of cost evidence could be linked to the preparation of the next effort . Such preparation is reflected by motor signals such as the readiness potential [29] , [30] or the de-synchronization of beta oscillations [31] , [32] . We showed in a previous publication [33] that the last process is modulated by incentive level; it could therefore mediate the effect of motivation on cost dissipation in the posterior insula . Second , the dissipation of cost evidence could be accentuated by analgesic mechanisms . The posterior insula region that signals cost evidence is also involved in pain perception [34] , [35] and placebo effect [36] . The placebo effect suggests that the brain has an endogenous means to control pain , possibly through the opioid system [37]–[39] . Another possibility would be serotonin , which contributes to the analgesia induced by common pain killers such as acetaminophen [40] and to the sensation of fatigue during effort [1] , [5] . Thus , through opioids or serotonin , the brain might be able to regulate cost-related signals depending on motivation level . Before concluding , we must acknowledge some limitations and inconsistencies . First , the situation explored in our paradigm is highly restricted . Notably , subjects are only allowed to adjust the duration of their effort and not the intensity , which we can usually adjust in ecological situations . We have conducted a series of studies where the payoff was based on effort intensity [41]–[43] , but we still have to explore a situation where the two dimensions can vary . Second , the model does not account for a number of observations , for instance the fact that fatigue accumulates at longer time scales . Indeed , we observed in some tasks an effect of trial and/or session number on effort duration , which could mean that cost was not fully dissipated after each rest period , or that slower effort-induced perturbations were accumulated elsewhere and imposed constraints on performance . Third , it cannot be formally concluded that subjects have no explicit access to the cost-evidence signal driving effort allocation , as one can always object that the inability to report a variable is due to the question not being appropriately formulated . This objection should nevertheless be tempered since our negative result is not an absence of effect: the question about exhaustion did elicit subjective ratings that were sensitive to the cost parameters ( effort duration and difficulty ) but not in the way that is relevant to cost monitoring ( multiplicative and not additive ) . It remains interesting that , although participants spontaneously explain why they stopped their effort in terms of exhaustion , they failed to report the variable ( difficulty times duration ) that exhausted their resources . Despite these limitations , the results of the present studies taken together provide strong evidence that costs are implicitly monitored in order to adapt effort duration on the fly , which can be dissociated from anticipatory adjustments depending on explicit costs and benefits . Moreover , this dissociation was computationally tractable and might be of clinical relevance . It suggests the existence of two different kinds of apathy: effort could be limited because the expected cost is over-estimated , or because the actual effort-induced cost is inflated . The first category ( perhaps in depression disorders ) would rest a lot but would encounter little difficulty in maintaining their effort once it is engaged , whereas the other ( perhaps in chronic fatigue ) would easily initiate efforts but then would rapidly renounce .
The study was approved by the Pitié-Salpétrière Hospital ethics committee ( protocol number: 106-07 ) . All subjects were recruited via email within an academic database and gave inform consent prior to participating in the study . The study was approved by the Pitié-Salpétrière Hospital ethics committee . All subjects were recruited via email within an academic database and gave inform consent prior to participating in the study . There was no restriction of handedness , excepted for the original ( Implicit ) Effort Allocation Task , in which participants were all right handed for neuroimaging purposes . Other inclusion criteria were: age between 20 and 39 years , absence of self-reported psychiatric or neurological history and of current psycho-active substance consumption . In all studies , participants were told that they would win the money accumulated during the task . In the previous study ( Implicit task ) , the payoff was eventually rounded up to a fixed amount ( 100€ ) credited by bank transfer . In all new studies participants were paid in cash at the end of the experiment . The payoff was partitioned into a fixed amount and variable amount depending on the money won during the task . For the Cost Rating Task , the amount earned during the task was eventually down-scaled ( divided by 2 . 48 ) to fit in a budget of 30€ per subject while maintaining the correspondence between payoff and incentive during the task . Participants were informed about this correction prior to the experiment . The Implicit task was performed in a MRI scanner for half the subjects and under a MEG helmet for the other half . One subject in the MRI group was excluded from the analysis because of calibration issues . For the Adaptation Task 3 , one participant was excluded because of calibration issues and another for cheating ( repeated , direct manipulation of the air tube ) . For the Dissociation Task , one participant was excluded due to an instruction issue: she could not understand the meaning of the percentage displayed on the screen , which indicated the difficulty level in proportion of the maximal force . Two other participants were excluded due to calibration issues . The task-specific information is summarized in Table 1 . We used homemade power grips composed of two plastic or wood cylinders compressing an air tube when squeezed . The tube was connected to a transducer converting air pressure into voltage . Thus , grip compression resulted in the generation of a differential voltage signal , linearly proportional to the force exerted . The signal was amplified and digitized by a signal conditioner ( CED 1401 , Cambridge electronic design , UK ) for Implicit , Explicit and Dissociation tasks , and by a homemade device for the Adaptation Tasks and Cost Rating task . The digitized signal was read by a Matlab program ( The MathWorks Inc . , USA ) . In the Adaptation Tasks ( 1 to 3 ) , the effort onsets were identified on-line and used to update the screen displayed to the participants . The effort onset was determined as the first sample exceeding 20% of the participant maximal force . In the Effort Allocation Tasks ( ‘implicit’ , ‘explicit’ and ‘dissociation’ ) , effort onsets and offsets were identified off-line with an algorithm using the same two criteria for all conditions: 1 ) force temporal derivative higher than one standard deviation and 2 ) force level lower ( for effort onset ) or higher ( for effort offset ) than half the maximal force . The first rest period started with coin presentation and the subsequent effort and rest periods were defined by force onsets and offsets . For all tasks , we measured the maximal force for each hand before starting task performance , following published guidelines [1] . Participants were verbally encouraged to squeeze continuously as hard as they could , until a growing line displayed on a computer screen reached a target . The growing rate was proportional to the force produced to motivate subjects to squeeze hard . Maximal force was set to the average of data points over the last half of the squeezing period exceeding the median . Then subjects were provided a real-time feedback about the force produced on the handgrip , which appeared as a fluid level moving up and down within a thermometer , the maximal force being indicated as a horizontal bar at the top . Subjects were asked to produce a force such that fluid level would reach this horizontal bar and to state whether it truly corresponded to their maximal force . If not , the calibration procedure was repeated . The procedure was slightly simplified for the Adaptation Tasks and Cost Rating Task: 1 ) the rate of the growing bar was held constant and not indexed on the participants' exerted force level , 2 ) the duration during which the participants had to squeeze as hard as they could was fixed to 5 s , and 3 ) all data points were used for the estimate ( and not the last half of recorded levels ) . All tasks were presented on a computer screen , and were programmed with Matlab using Cogent 2000 ( Wellcome Department of Imaging Neuroscience , London , UK ) for the Implicit and Explicit Tasks , and Psychtoolbox ( http://psychtoolbox . org ) for the Dissociation Task , Adaptation Tasks , and Cost Rating Task . To perform model selection , models were first estimated for each subject following a variational Bayes approach under the Laplace approximation [45] , [46] , using a toolbox by Jean Daunizeau [47] ( available at http://code . google . com/p/mbb-vb-toolbox/ ) . Note that all the models developed here are deterministic: they are meant to provide a mechanistic link from factors of interest ( monetary incentive or task difficulty ) to observations ( effort or rest duration ) . The aim of model estimation was to find the distribution of free parameters that best fitted the observations , and not to explain their stochasticity . The variational Bayes algorithm not only estimates linear and non-linear models but also calculates their evidence based on a free-energy approximation [45] . The evidence of a model is the probability of observing the data given this model . This probability corresponds to the marginal likelihood , which is the integral over the parameter space of the model likelihood weighted by the prior on free parameters . This probability increases with the likelihood ( which measures the accuracy of the fit ) and is penalized by the integration over the parameter space ( which measures the complexity of the model ) . The model evidence thus represents a trade-off between accuracy and complexity and can guide model selection [48] . Model selection was performed with a group-level random-effect analysis of the log-evidence obtained for each model and subject , using Gibbs sampling in SPM8 ( Statistical Parametric mapping , Wellcome Department of Imaging Neuroscience , London , UK ) [48] . This procedure estimates the expected frequency ( denoted ef ) and the exceedance probability ( denoted xp ) for each model within a set of models , given the data gathered from all subjects . Expected frequency quantifies the posterior probability , i . e . the probability that the model generated the data for any randomly selected subject . This quantity must be compared to chance level ( one over the number of models or families in the search space ) . Exceedance probability quantifies the belief that the model is more likely than all the other models of the set , or in other words , the confidence in the model having the highest expected frequency [48] . Family-level inference was conducted similarly to model-level inference after defining a partition within the model space as described in [49] and implemented in SPM8 . We first defined a class of models that can a priori produce the results that we intended to explain . These models were then submitted to a Bayesian model selection in order to identify the most plausible model among all the possible models . The model space was defined by simplifying a full model , starting with the Implicit Task . The model is based on accumulation-dissipation processes: cost evidence ramps up during effort to a bound that triggers effort cessation , and ramps down during rest to a bound that triggers effort resumption . As for simplicity the fluctuations were modeled as linear , the effort and rest durations ( Te and Tr ) are just the ratios between the amplitude A ( distance between bounds ) and the accumulation or dissipation slope ( Se and Sr ) . In the full model , the free latent parameters A , Se , Sr can vary across trials around their mean values ( αm , βm , γm ) , depending linearly on experimental factors: in this case the incentive I and the difficulty D . The full model is thus:Simpler models can be designed by setting one or more weights to zero . As there are 6 weights ( 3 latent parameters times 2 experimental factors ) , all combinations give a total of 26 = 64 models . However , some of these models are not worth considering as they cannot account for the effect that we want to explain . The most extreme case is when all weights are null: such a model cannot produce any of the effect of incentive and difficulty that we observed in the data . After discarding all models with which at least one of the significant results reported in Figure 5 could not be produced ( shown with red in Figure 6A ) , the search space was restricted to 24 models . Note that predicting an effect which was not significant in our data was not a criterion for rejection . To illustrate the logic of model selection , we can take the case of the opposite effects of incentives on effort and rest duration ( Te and Tr ) . This behavioral pattern cannot be accounted for by models which would have no incentive effect at all or an effect on one latent parameter only ( A , Se or Sr ) . These models can therefore be excluded from the model space retained for Bayesian comparison . On the contrary , modulations of both A and Se ( #1 ) , or both A and Sr ( #2 ) , or all three parameters ( #3 ) can a priori produce the observed incentive effects on Te and Tr; these models should therefore be compared . Model #1 predicts that incentive effect on Tr should be linear ( because it only affects the numerator A ) and the effect on Te non-linear ( because it also affects the denominator Se ) . Model #2 generates the opposite predictions . The data can thus disambiguate between these models , depending on which effect is non-linear . Model #1 and #2 are special cases of model #3 which is more complex and thereby , more likely to provide a better fit . Yet model #3 will be preferred over model #1 and #2 only if the improvement of fit surpasses the inflation of complexity in the calculation of model evidence . Note that this example is a simplification of our model comparison , as all factors ( not just incentives ) should be considered at the same time , which also makes predictions on the pattern of interaction between factors . However , detailing the specific predictions of all the models included in the Bayesian comparison would require much more length than that allowed in a research paper . The same approach was applied for defining linear models of the Explicit Task , leading to a search space of 16 models ( Figure 6C ) . For the Dissociation Task , another modulator was included as there were two types of difficulty ( cued and actual ) . The full model has therefore 9 weights , which gives 29 = 512 possible models , which were reduced to 144 models after rejection of irrelevant models ( Figure 6B ) . Since these models provided poor fit and unclear evidence in favor of a particular model , we also tested a class of hyperbolic models for the Explicit and Dissociation Tasks . As opposed to the linear formulation , the discount of incentive by cued difficulty was assumed to be hyperbolic , as in some economic models of temporal discounting . The full hyperbolic model is:The D term that denoted difficulty in the first linear model has been decomposed into Da and Dc , denoting actual and cued difficulty in the Dissociation Task . Note that in the Explicit Task , De and Dc have exactly the same values . The model can nonetheless be estimated unambiguously in this task since the effect of Da is linear whereas that of Dc is hyperbolic . Also note that with hyperbolic formulation , there are dependencies between weights since a null numerator prevents the denominator from impacting the model fit . Thus we discarded models with a null numerator and a non-null weight at the denominator ( this is shown with red in Figure 6D ) . After discarding the models that were not able to produce all the significant results shown in Figure 5 , the search space was eventually restricted to 78 models for the Explicit and Dissociation Tasks .
|
Imagine that ahead of you is a long time of work: when will you take a break ? This sort of issue – how to allocate effort over time – has been addressed by distinct theoretical fields , with different emphasis on reactive and predictive processes . An intuitive view is that you start working , stop when you are tired , and start again when fatigue goes away . Biologically , this means that decisions are taken when some physiological variable reaches a given bound on the risk of homeostatic failure . In a more economic perspective , fatigue translates into effort cost , which must be anticipated and compared to expected benefit before engaging an action . We proposed a computational model that bridges these perspectives from sport physiology and decision theory . Decisions are made in reaction to bounds being reached by an implicit cost variable that accumulates during effort , at a rate proportional to task difficulty , and dissipates during rest . However , some latent parameters ( bounds and dissipation rate ) are adjusted in anticipation , depending on explicit costs and benefits . This model was supported by behavioral data obtained using a paradigm where participants squeeze a handgrip to win a monetary payoff proportional to effort duration .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"medicine",
"and",
"health",
"sciences",
"decision",
"making",
"social",
"sciences",
"neuroscience",
"cognitive",
"neuroscience",
"cognitive",
"psychology",
"human",
"performance",
"cognition",
"behavior",
"consciousness",
"pain",
"psychology",
"pain",
"management",
"motivation",
"psychophysics",
"biology",
"and",
"life",
"sciences",
"sensory",
"perception",
"cognitive",
"science",
"motor",
"reactions"
] |
2014
|
How the Brain Decides When to Work and When to Rest: Dissociation of Implicit-Reactive from Explicit-Predictive Computational Processes
|
Like other domesticates , the efficient utilization of nitrogen resources is also important for the only fully domesticated insect , the silkworm . Deciphering the way in which artificial selection acts on the silkworm genome to improve the utilization of nitrogen resources and to advance human-favored domestication traits , will provide clues from a unique insect model for understanding the general rules of Darwin's evolutionary theory on domestication . Storage proteins ( SPs ) , which belong to a hemocyanin superfamily , basically serve as a source of amino acids and nitrogen during metamorphosis and reproduction in insects . In this study , through blast searching on the silkworm genome and further screening of the artificial selection signature on silkworm SPs , we discovered a candidate domestication gene , i . e . , the methionine-rich storage protein 1 ( SP1 ) , which is clearly divergent from other storage proteins and exhibits increased expression in the ova of domestic silkworms . Knockout of SP1 via the CRISPR/Cas9 technique resulted in a dramatic decrease in egg hatchability , without obvious impact on egg production , which was similar to the effect in the wild silkworm compared with the domestic type . Larval development and metamorphosis were not affected by SP1 knockout . Comprehensive ova comparative transcriptomes indicated significant higher expression of genes encoding vitellogenin , chorions , and structural components in the extracellular matrix ( ECM ) -interaction pathway , enzymes in folate biosynthesis , and notably hormone synthesis in the domestic silkworm , compared to both the SP1 mutant and the wild silkworm . Moreover , compared with the wild silkworms , the domestic one also showed generally up-regulated expression of genes enriched in the structural constituent of ribosome and amide , as well as peptide biosynthesis . This study exemplified a novel case in which artificial selection could act directly on nitrogen resource proteins , further affecting egg nutrients and eggshell formation possibly through a hormone signaling mediated regulatory network and the activation of ribosomes , resulting in improved biosynthesis and increased hatchability during domestication . These findings shed new light on both the understanding of artificial selection and silkworm breeding from the perspective of nitrogen and amino acid resources .
The silkworm , Bombyx mori , is the only fully domesticated insect species , originating from its wild ancestor , B . mandarina , approximately 5000 years ago . During the domestication process , the domestic silkworm evolved rapidly under human-preferred selection . Deciphering the way in which artificial selection acts on the silkworm genome to produce human-favored domestication traits will provide clues from a unique insect model for understanding Darwin's theory of artificial selection [1] . Recently , through genome-wide screening of selection signatures in a large batch of domestic and wild silkworms , we identified candidate domestication genes that enriched nitrogen and amino acid metabolism pathways , specifically in glutamate and aspartate metabolism . Knockout of two involved genes resulted in abnormal metamorphosis and decreased cocoon yield [2] . These findings suggest that , like domestic plants and animals , domestic silkworms also tend to have efficient utilization of nitrogen resources to adapt to human-preferences [2–4] . In addition to the glutamate and aspartate metabolism , which is an ammonia re-assimilated system [5] , we further wonder whether other kinds of nitrogen supplies are also affected by artificial selection . If this is the case , how have they contributed to silkworm phenotypic changes during domestication ? Insect storage proteins ( SPs ) are another important resource of amino acids and nitrogen . Specifically , SPs are repositories of stored amino acids that belong to a special conserved arthropod hemocyanin superfamily [6] . Most insects have at least two main types of storage proteins , i . e . , arylphorin and methionine-rich storage proteins; some species have other atypical SPs [7] . SPs have been cloned or predicted in many insect species , including Lepidoptera moths and butterflies [8–11] . Insect SPs are believed to serve as a source of amino acids and nitrogen for pupae and adults during metamorphosis and reproduction [12] , however there is little solid functional evidence of their biological significance [10 , 13] . In plants , storage proteins are mainly reserved in seeds , where , along with other nutrients such as oil and starch , they supply energy for seed germination and growth [14 , 15] . Particularly in crops , seed SPs act to provide energy for humans and animals and thus are of great interest and a target for breeding and improvement [14–16] . In the domestic silkworm , previous studies preliminarily characterized the gene and protein expression patterns of four SPs [8 , 17–19] . SP1 is female-biased expressed when entering the last instar , and only accumulates in the female pupa [8 , 20 , 21] . It has been suggested that SP1 contributes to adult female characters and is related to the synthesis of vitellogenin ( Vg ) , the precursor of yolk protein [17] . SP2 couples with SP3 to form a heterohexamer and has inhibitory effects on cell apoptosis [18 , 19] . SSP2 was a heat resistant protein and suggested has a cell-protective function [19] . The determination as to whether or not SPs are also important in silkworm domestication , as is the case in domesticated plants , awaits a thorough exploration of their biological and evolutionary significance . Development of genomics and genome-editing techniques provide tools for efficiently deciphering the evolutionary and functional significance of particular genes [22 , 23] . In this study , we conducted a genome-wide identification of the silkworm SPs . Taking advantage of the genomic data resource of a batch of representative domestic and wild silkworms [2] , we performed selection signature screening of the silkworm SPs followed by functional verification via the CRISPR/Cas9 knockout system and comprehensive comparative ova transcriptomes of wild-type and mutant silkworms as well as domestic and wild silkworms . Our findings suggest that artificial selection on SP1 contributes to increased egg hatchability during silkworm domestication , possibly by promotion of vitellogenin , influence of hormone synthesis and egg development and eggshell formation . These results provide a novel case with functional evidence for the determination of a regulatory framework on a silkworm domestication gene , revealing that artificial selection acting on the nitrogen and amino acid supply is also required for improved silkworm reproduction .
In total , we identified 8 SPs in the silkworm genome by means of a blast search . Among which SP2 were not annotated in the gene list . Of these , SP1 exhibited the highest methionine content ( 10 . 98% ) ( Table 1 ) . Phylogenetic analysis showed that SP1 was located in one distinct clade , whereas other SPs were in another , indicating an obvious divergence between SP1 and the remaining SPs ( Fig 1A ) . SP1 is located on chromosome 23 while the other SPs are clustered on chromosome 3 , suggesting possible tandem duplication events during evolution . Interestingly , by screening artificial selection signatures on the genomic region bearing SP1 and the other SPs respectively , we detected a strong selection signature in the SP1 region of the domestic silkworm ( see Material and methods ) , since there was notably reduced nucleotide diversity in the domestic silkworm group ( Fig 1B and 1C ) . Furthermore , we detected strong differentiation in allelic frequency upstream of SP1 ( Fig 1D ) . Correspondingly , SP1 was differentially expressed in the ova of domestic and wild silkworms , with higher expression in the domestic one ( Fig 1E , S1 Fig ) . We also detected 11 SNPs that caused amino acid changes in the coding sequence of the gene ( S2 Fig ) ; the biological significance of these SNPs requires further evaluation . These results suggest that artificial selection acting on SP1 during silkworm domestication may affect the function of this gene in domestic silkworms . At the very least , we can infer that selection may favor higher expression in the domestic silkworm . To explore the possible phenotypic influence of artificial selection of SP1 acting on the domestic silkworm , we first investigated the biological role of this gene in the silkworm using CRISPR/Cas9 knockout system . For the single guide RNA ( sgRNA ) design , we selected highly specific targets in the first exon , close to the translation starting site; namely , S1 and S2 ( Fig 2A ) . We chose another site S3 close to the end of the first exon , more than 60 bp downstream from S1 and S2 ( Fig 2A and Table 2 ) to obtain a potentially large fragment deletion by injecting the pool of three gRNAs . After mutation screening of the injected eggs ( G0 generation ) , the gRNAs targeting the above three sites successfully guided DNA editing and generated a variety of mutation types , including 4–9 bp deletions or small insertions followed by a large deletion ( Fig 2B ) . Through screening of the exuviae of the fifth instar larvae in the G0 cocoons , we successfully identified 26 mosaic mutant G0 moths . We then generated pairwise crosses of those G0 mutants with similar mutant genotypes from the G1 populations . After mutation screening of the G1 eggs , we selected two populations with large deletions for further feeding and mutation screening ( see Material and methods ) . Finally , in the G2 generation we obtained two types of homozygous mutants , i . e . , MU1 and MU2 ( Fig 2C ) . In MU1 , there was an 8 bp insertion followed by a 63 bp deletion in the SP1 coding sequences . In MU2 , there was a 4 bp insertion followed by a 65 bp deletion . The mutations occurred at +29 and +26 bp of the first SP1 exon in MU1 and MU2 , respectively ( Fig 2C ) , resulting in reading frame shift mutations and severe premature termination close to the translation starting site , with stop signals at +10 aa and +37 aa of the SP1 protein ( Fig 2D ) . We selected and maintained the MU1 population for assay on phenotypes related to reproduction and metamorphosis , such as the number of eggs , hatching rate , pupa weight , and cocoon weight . Compared with the wild-type , which exhibited hatching rates of approximately 90% , the hatching rates of the SP1 mutants were dramatically lower , with a mean value of about 40% ( Fig 2E ) , although neither the number of eggs produced nor the whole pupa weight or cocoon shell weight were noticeably affected ( Fig 2E ) . Given that the data were obtained from large replicates ( 83 replicates for the hatchability assay and 240 replicates for the pupa and cocoon weights ) , the results are robust . Loss-of-function mutation resulted in significantly decreased expression of SP1 and Vg in the ova , based on the RNA-seq data ( Fig 2F ) . These results suggest that in the silkworm SP1 may positively affect the expression of ova Vg and contribute to silkworm egg development . Given that knockout of SP1 resulted in a reduced hatching rate ( Fig 2E ) and that it is female-specific expressed in the pupa and adult stages , we suspect that it plays an important role in ova development , thus contributing to an efficient hatching process . During domestication , artificial selection preferred higher expression of SP1 , thus may improve the silkworm hatching rate . As expected , we found that the hatching rate of the domestic silkworm was significantly higher than that of the wild one ( Fig 3A ) . No obvious differences in egg production was detected between wild and domestic silkworms ( Fig 3A ) . The lower hatching rate of the wild silkworm has also been reported in other studies [24 , 25] . We further tested expression of Vg in the ova and discovered that consistently , it was significantly higher expressed in the domestic silkworm than in the wild silkworm ( Fig 3B , S1 Fig ) . Promotion of SP1 expression in the domestic silkworm thus results in the corresponding up-regulation of Vg , which further contributes to increased hatchability during silkworm domestication . In order to further explore the regulation network and possible molecular mechanisms of female-specific SP1 on egg hatchability , we generated comprehensive ova comparative transcriptome analyses between the wild-type and the mutant , as well as the domestic and wild silkworm ( Bombyx mandarina ) , with 4 . 87~9 . 15 Gb RNA-seq data for each sample ( S1 Table ) . We chose ova instead of fertilized eggs as target because silkworm SP1 is female-biased expressed when entering the last instar and only accumulated in the female [8 , 20 , 21] . Comparative transcriptomics in this target tissue would directly focus mechanism of SP1 on female reproductivity and avoid potential influence from the male . In total , there were 561 genes identified as differentially expressed genes ( DEGs ) in the SP1 knockout mutants ( MU1 ) compared to the wild-type silkworm , with significantly more down-regulated genes ( 341 ) than up-regulated ( 220 ) ( p = 0 . 0003 , Chi-squared test with Yates' continuity correction ) ( Fig 4A and S2 Table ) . As expected , we found many more DEGs ( 2882 ) between the wild and domestic silkworms , since wild silkworms are much more genetically and phenotypically different from the domestic one , compared with the silkworm mutant from the wild-type . It is interesting that in the 2882 DEGs , there were also significantly more lower expressed genes ( 1761 ) than higher expressed ( 1121 ) ( p = 2 . 2e-16 , Chi-squared test with Yates' continuity correction ) ( Fig 3A and S3 Table ) in the wild silkworm . These results suggest that transcriptome repression in ova might be an output of SP1 depletion in the SP1 mutant ( Fig 2F and Fig 4A ) and a low expressional level of SP1 in the wild silkworm ( Fig 1E; Fig 4A and S1 Fig ) . We identified 302 common genes in the two sets of DEGs . KEGG enrichment analysis indicated that these common DEGs were significantly enriched in pathways related to cell proliferation , such as ECM-receptor interaction and folate biosynthesis , as well as the hormone synthesis pathway , which is important in adult ovary development and female production [26 , 27] ( Fig 4B ) . Gene ontology ( GO ) enrichment analysis indicated that the common genes were enriched in reproduction related biological processes , such as chorion-containing eggshell formation ( Fig 4B ) . These genes were also enriched in the molecular functions of the structural constituents of chorion ( Fig 4B ) . In fact , they are all annotated as chorionic proteins , including 3 chorion class CB protein M5H4-like genes ( BGIBMGA009720 , BGIBMGA009719 , BGIBMGA009715 ) as well as a chorion class B protein PC10 gene ( BGIBMGA009721 ) . All these chorion like genes showed significantly higher expression in the domestic silkworm , compared with both the mutant and the wild silkworm ( Fig 4B , S2 Table , S3 Table ) . This pattern was also supported by Real-Time PCR validation in the domestic and wild silkworm ( S1 Fig ) . Genes in ECM-receptor interaction pathway include collagens and integrins ( S3 Fig ) and those in folate biosynthesis include folylpolyglutamate synthase , which involves in 7 , 8-Dihydrofolate ( DHF ) and 5 , 6 , 7 , 8-Tetrahydrofolate ( THF ) , substrates for subsequent one carbon pool mediated by folate ( S4 Fig ) . The enriched hormone synthesis pathway includes genes functioning in both juvenile and molting hormones ( S5 Fig ) . Extend to all the enriched genes , it is notable that most of these enriched genes were relatively highly expressed in the domestic silkworm , compared with SP1 mutant and the wild silkworm ( Fig 4B ) . We further generated enrichment analyses on DEGs on the two sets of DEGs independently and observed consistent pattern ( Tables 3 and 4 ) . Functional enrichment analysis of DGEs between wild-type silkworm and SP1 mutant silkworm revealed a significantly enriched the KEGG pathway “ECM-receptor interaction” as well as other three pathways with marginal significances , insect hormone biosynthesis , Glycine , serine and threonine metabolism and Folate biosynthesis ( Table 3 ) . The enriched GO terms included eggshell formation process and structural constituent of chorion . Most of the genes in these two GO terms were down-regulated in the mutant ( Table 3 ) . Consistently , These GO items were also in the top rank with the lowest p values when analyzing the DEGs between wild and domestic silkworm ( Table 4 ) . Nearly all of the genes involved in these KEGG and GO terms showed significant lower expression level in the wild silkworm , i . e . , up-regulated in the domestic silkworm ( Table 4 ) . These results further supported that in the wild silkworm , low expression level of SP1 may be associated with suppressed expression of genes in the eggshell formation process . DEGs between domestic and wild silkworms were significantly enriched in function of structural constituent of ribosome . The related genes are mostly ribosome proteins ( S6 Fig ) . We also noted that the related biological processes , such as amide and peptide biosynthesis , was also in the top rank with the lowest p value ( Table 4 ) . The related genes were also up-regulated in the domestic silkworm . During domestication , there might be other factors that contribute to improved hatchability , such as , improved amide and peptide biosynthesis and activated ribosome activities in the ovaries .
Nitrogen resources are very important for silkworm domestication . The domestic silkworm tends to efficiently utilize nitrogen resources to yield protein outputs to adapt to human-preference , such as the economically important product , the cocoon . In this study , we discovered that artificial selection could directly act on a nitrogen resource gene , i . e , storage protein 1 ( SP1 ) , to improve silkworm hatchability . SPs are also target loci of breeding in crops [28] . However , with edible crops , human can directly benefit from the nutrients of these improved SPs [16] , whereas in the silkworm , the SPs benefit is in the form of increased silkworm reproductive capacity . Among all the SPs identified , SP1 is quite divergent and somewhat unique from the others , both in terms of genomic location and phylogenetic position . A similar pattern was also observed in other Lepidoptera species , such as the tobacco hornworm , Manduca sexta [29] , suggesting that SP1 may have evolved independently , while the other types of SPs might have experienced duplication during Lepidoptera evolution . Methionine-rich SP1 seems to be of special interest , since methionine is reported to be an important amino acid in the trade-off between growth and reproduction [30] . In Drosophila , dietary methionine restriction extends lifespan [30] , while in grasshoppers , a reduced reproduction-induced increase in expression methionine-rich protein occurred during life extension [31] . Similarly , in the beet armyworm , silencing of SP1 by RNA interference ( RNAi ) decreases larval survival , which indicates the role of the methionine-rich SP in growth and metamorphosis[13] . We therefore added a new evidence that different to grasshopper [31] and the beet armyworm[13] , but similar to Drosophila[30] , silkworm methionine-rich SP1 functions in the reproduction process but does not obviously affect growth . Given that in cocoon-producing silk moths , other nitrogen utilization system such as the glutamate /glutamine cycle , have been reported to be vital in metamorphosis silk-cocoon production [2 , 5 , 32] , we suspect that the strategy of nitrogen resource allocation via storage proteins may have diverged or modified during Lepidoptera insect evolution . In the silkworm , the function of SP1 is limited to influencing the egg hatching rate . Artificial selection acted only on SP1 rather than other SPs , suggesting the importance of SP1 for human-preferred domestication traits , i . e . , increased hatchability . Ova comparative transcriptome analyses further illustrated a framework of regulatory network of SP1 on hatchability . Firstly , there are many genes near the bottom of the regulatory network , including vitellogenin ( Vg ) , chorion proteins , structural component proteins in the extracellular matrix ( ECM ) -interaction pathway such as collagen and integrins , and synthetase in folate biosynthesis are all generally repressed in both the SP1 mutant and the wild silkworm . Thus , artificial selection acts on SP1 for increased hatchability , possibly associated with the influence of those genes , pathway or biological processes , and finally contributes to an improved performance of ovary . Vg is the main nutrient for silkworm egg formation and embryonic development[33] . It appears and accumulates at the stage when SP1 rapidly declines and disappears in the fat body , shortly before the emergence of the adult silkworm [17 , 34] . SP1 may supply amino acids for the synthesis of Vg , as previously reported in Plutella xylostella [35] . We therefore suspect that deficiency of SPs might directly trigger an as yet unknown regulatory pathway for the expression or synthesis of Vg . Chorion proteins are the major component of the silkworm eggshell and perform the essential function of protecting the embryo from external agents during development , while simultaneously allowing gas exchange for respiration . Eggshell ( chorion ) is constructed by the ovarian follicle cells . The follicle cell epithelium surrounds the developing oocyte and , in the absence of cell division , synthesizes a multilayer ECM [36] . Eggshell ECM was usually linked by integrins , a family of transmembrane receptor proteins to the cytoskeleton of the oocyte . Via a series of signal transductions , ECM-integrins function in oocyte movement , differentiation , and proliferation [36] . Integrins were reported to function in formation of actin arrays in the egg cortex [37] and they were also involved in tracheole morphogenesis which affects respiration [38] . Repression of these genes are directly associated with deficient development and function of the ovary . Loss of function of SP in the mutant or low expression level in the wild silkworm of SP might influence development and function of the ovary , further reducing the expression of Vg and chorion genes . Secondly , folate is known to be important for human fetal development [39] . In insects , folate also plays an important roles in egg development , possibly promoting the biosynthesis of nucleic acids in the ovaries , and evoking mitoses in cells of the collicular epithelium [40–42] . Last and interestingly , we found that the hormone synthesis pathway was also repressed in response to SP1 deficiency . Recent advances in hormone signaling indicate that in the adult insect , juvenile ( JH ) and molting hormones may cooperate to promote Vg expression and oocyte development [27 , 43 , 44] . Therefore , hormone signaling pathway might function in the regulatory network of SP in these downstream genes , although there are still black boxes in the regulation connections of these genes , which require further in-depth experimental exploration . Notably , increased hatchability during domestication may not be solely attributed to the increased expression of SP1 and the associated downstream genes , given that artificial selection acts on hundreds of gene loci in the silkworm genome [2 , 45] and that the ova comparative transcriptome between wild and domestic silkworms identified many more genes than that between SP1 mutant and wild-type silkworm . We observed significantly enriched pathway and structural constituent of ribosome , the protein translation machinery and the biological processes involved in nitrogen metabolism and , are generally up-regulated in the domestic silkworm compared with the wild one ( Table 4 ) . These results again supported the importance of nitrogen and amino acids in silkworm domestication , not only for silkworm protein output [2] , but also for productivity . Similar to other domesticates , hatchability of silkworm eggs directly determines the quantity of offspring , and thus it is an important productivity trait for human to favorably select during domestication . Based on the above results and the discussion , we propose that artificial selection , favors higher expression of SP1 in the domestic silkworm , which would subsequently up-regulate the genes or pathways vital for egg development and eggshell formation . On the other hand , artificial selection consistently favors activated ribosome activities and improved nitrogen metabolism in the ova , as it might act in the silk gland for increased silk-cocoon yield [2] . In result , the domestic silkworm demonstrates improved egg hatchability compared with it wild ancestor .
A multivoltine silkworm strain , Nistari , was used in all experiments . Larvae were reared on fresh mulberry leaves under standard conditions at 25°C . The wild silkworms were collected in Zhejiang Province , China and maintained as laboratory population in our lab . The genomic single nuclear polymorphic data ( SNP ) file ( the VCF ) for the domestic and wild silkworm obtained from DEYAD platform ( https://doi . org/10 . 5061/dryad . fn82qp6 ) [2] . Reference genome and the annotation file used for RNA-seq data mapping were obtained from the Ensemble database ( http://metazoa . ensembl . org/Bombyx_mori/Info/Index ) . The reference sequences of B . mori storage proteins ( SP1 , SP2 , SSP2 and SP3 ) were retrieved from the NCBI GenBank . These sequences were used as query , searching for homologs in the B . mori genome by tblastn with e-value <10−7 . Other insect homologs of the silkworm SPs were searched in GenBank ( https://blast . ncbi . nlm . nih . gov/ ) by BLASTP with an e-value <10−7 . We selected sequences from several representative Lepidoptera species and Drosophila melanogaster as candidate proteins for further analyses . The sequences of the SP1 homologs were aligned using MEGA 6 . 0 software [46] . A gene tree was constructed using MrBayes-3 . 1 . 2 with GTR + gamma substitution model [47] . The gene-ration number was set as 1000000 and the first 25% was set as burn-in . Other parameters were set as default . Based on the available whole genomic single nuclear polymorphic data ( SNP ) of domesticated and wild silkworm populations [2] , ( https://doi . org/10 . 5061/dryad . fn82qp6 ) , we screened the selection signatures of the silkworm SPs , according to Xiang et al’s pipeline [2] . Specifically , data from 19 samples of the early domesticated group ( i . e . , trimoulting local strains [CHN_L_M3] ) of the domestic silkworm Bombyx mori and 18 samples of wild silkworm B . mandarina were used . The SNP data of the two chromosomes that SP1 ( Chromosome 23 ) and the cluster of the other SPs ( Chromosome 3 ) were located were used to screen for the domestication signature . Chr . 23 is 20 , 083 , 478 bp in length and 2 , 046 , 397 SNPs were identified . Chr . 3 is 14 , 662 , 804 bp in length and 1 , 448 , 852 SNPs were identified , based on the published data . Allelic frequency and SNP annotation were calculated using in-house Perl scripts . For the detection of selection signature during silkworm domestication , we set a very stringent threshold to screen out regions significantly deviated from the overall distribution . We only used windows within the top 1% of selective signatures ( the corresponding p value of a Z test < 0 . 001 ) and applied Fst ( fixation index ) between the two groups to represent the selective signatures , taking the highest 1% value as the cutoff . The selection in the domestic silkworm group ( i . e . , the early domesticated group ) was further confirmed by limiting π at a relatively low level ( the lowest 5% ) . The 20 bp sgRNA targets immediately upstream of PAM were designed by the online platform CRISPRdirect ( http://criSpr . dbcls . jp/ ) [48] . The sgRNA DNA template was synthesized by PCR , with Q5 High-Fidelity DNA Polymerase ( NEB , USA ) . The PCR conditions were 98°C for 2 min , 35 cycles of 94°C for 10 s , 60°C for 30 s , and 72°C for 30 min , followed by a final extension period of 72°C for 7 min . The sgRNA were synthesized based on the DNA template in vitro using a MAXIscript T7 kit ( Ambion , Austin , TX , USA ) according to the manufacturer’s instructions . The Cas9 construct was a kind gift provided by the Shanghai Institute of Plant Physiology and Ecology ( Shanghai , China ) . The Cas9 vector was pre-linearized with the NotI-HF restriction enzyme ( NEB , USA ) . The Cas9 mRNA was synthesized in vitro with a mMESSAGE mMACHINE T7 kit ( Ambion , Austin , TX , USA ) according to the manufacturer’s instructions . All related primers are shown in Table 2 . Fertilized eggs were collected within 1 h after oviposition and microinjection was within 4 h . The Cas9-coding mRNA ( 500 ng/μL ) and total gRNAs ( 500 ng/μL ) were mixed and injected into the preblastoderm Nistari embryos ( about 8 nl/egg ) using a micro-injector ( FemtoJet , Germany ) , according to standard protocols ( Tamura , 2007 ) . The injected eggs were then incubated at 25°C for 9–10 d until hatching . To calculate the effect of Cas9/sgRNA-mediated gene mutation in the injected generation ( G0 ) , we collected ~10% of the eggs ( 64 out of 600 ) 5 d after injection to extract genomic DNA for PCR , with primers Sp1-F and Sp1-R ( Table 2 ) . The amplified fragments were cloned into a pMD19-T simple vector ( Takara , Japan ) and sequenced to determine mutation type . When the injected G0 silkworms pupated , we collected silkworm exuviae from fifth instar larvae in each cocoon . Genomic DNA was extracted using a TIANamp Blood DNA Kit ( Tiangen Biotech , Beijing ) according to the manufacturer’s instructions . Individual mutation screening was generated with PCR at 94°C for 2 min , 35 cycles of 94°C for 30 s , 57°C for 30 s , and 72°C for 45 s , followed by a final extension period of 72°C for 5 min . The PCR products were cloned into the pMD19-T simple vector ( Takara , Japan ) and sequenced . Mosaic mutant moths were obtained from the above mutation screening of exuviae DNA from fifth instar larvae . Moths with the same mutation site were pairwise crossed with each other to acquire G1 offspring . About 7 d after the G1 eggs were laid , we collected ~30 eggs from each offspring population from one parental pair and pooled them to extract genomic DNA for mutation screening by PCR . The amplified fragments were cloned into a pMD19-T simple vector ( Takara , Japan ) and sequenced to determine the exact mutation type . Two G1 offspring populations with large deletions in BmSp1 were selected for further breeding . At the pupa stage , 20 randomly selected individuals within each population were subjected to mutation screening of exuviae DNA . Homozygous mutant moths with the same identified mutant genotype were crossed to acquire G2 offspring . Mutation effects on proteins were evaluated using MEGA 6/0 software[46] through codon alignment of the wild-type and the mutant . On the fourth day of pupation ( P4 ) , we weighed and recorded the whole cocoon weight , pupa weight , and cocoon shell weight . In total , data from 240 SP1-MU1 mutants and 110 wild-type silkworms were recorded respectively . Offspring of the homozygous mutants and wild-type silkworms were incubated at 25°C for 9–10 d until hatching . The number of eggs produced and hatched from each female moth were recorded respectively . The egg hatching rates were then determined . Eighty-three replicates were set for the SP1 mutant and 17 for the wild-type populations , respectively . These assays were also generated for 20 wild silkworms . Student’s t-test was used to analyze the significance of the differences . For comparisons of datasets with unbalanced size , Student’s t-test with FDR ( false discovery rate ) correction was used . Specifically , for the cocoon- and pupal- related traits , we divided the samples of SP1 mutant to two groups , consisting of 120 samples for each , and generated t-test with the wild type respectively . The average p value followed by FDR correction was used to verify the significance . As for the analysis of the number of eggs and hatching rates , we divided the samples of SP1 mutant to 4 groups , consisting of 20 or 21 samples for each , and generated t-test with the comparable data from the wild-type silkworms , respectively . Average p value followed by FDR correction was used to verify the significance . Ova from newly emerged virgin moth of the domestic wild type silkworm , SP1 mutant and the wild silkworm were dissected and collected for RNA extraction with three replicates set for each . Total RNA were isolated using TRIzol ( Invitrogen ) . For each sample , RNA were sent to Novogene Bioinformatics Institute ( Beijing , China ) for cDNA library construction and RNA-seq by Illumina Hiseq 2500 ( Illumina , San Diego , CA , USA ) with 125 bp paired-end reads according to the manufacturer’s instructions . Raw data were filtered with the following criteria: ( 1 ) reads with ≥ 10% unidentified nucleotides ( N ) ; ( 2 ) reads with > 10 nt aligned to the adapter , allowing ≤ 10% mismatches; and ( 3 ) reads with > 50% bases having phred quality < 5 . The clean data were mapped to the Bombyx mori reference genome using Tophat with 2 nt fault tolerance and analyzed using Cufflinks [49] . The relative expression value of each gene was calculated using the widely used approach , i . e . , fragments per kilobase of exon per million pair-end reads mapped ( FPKM ) [49] , using Cuffdiff In order to identify differentially expressed genes ( DEGs ) , Cuffdiff was further used to perform pairwise comparisons between wild-typed and SP1 mutant samples , as well as the wild and domestic silkworm , respectively , with corrected P-value of 0 . 05 <5 and Log2-fold change>1 . KEGG and GO enrichment analyses of DEGs were performed with an online platform ( http://www . omicshare . com/tools/ ) , using all the expressed genes ( FPKM >1 ) in the ova of virgin moth of Bombyx mori as background . We used real-time PCR to evaluate the results of RNA-seq data . The Ova of the domestic wild type silkworm and the wild type were dissected from newly emerged virgin moths . Total RNA was digested with DNase I ( Takara ) to remove the remaining DNA . For Complimentary DNA synthesis , 1ug of total RNA was used in the ReverAid First Strand cDNA Synthesis kit ( Takara ) . Primers for real-time PCR were as follows: 5′ -GGCTTCACTGTCACCAGCACTT-3′ ( BGIBMGA009715_f ) and 5′ -ACCACAGCCGTAAGACACCAGA-3′ ( BGIBMGA009715_r ) for BGIBMGA009715; 5′ -GGGCTTATGATGCCGTAGGA-3′ ( BGIBMGA009719_f ) and 5′ -CGGTGGGAGTTATTGGTGATGT-3′ ( BGIBMGA009719_r ) for BGIBMGA009719; 5′ -ACCAGCATATCACCAATAGCACC-3′ ( BGIBMGA009720_f ) and 5′ -ATCGCCGCAGCCATACAGAA-3′ ( BGIBMGA009720_r ) for BGIBMGA009720; 5′ -GGCTTCATCTATCATCGCTCCAC-3′ ( BGIBMGA009721_f ) and 5′ -GCCACACCCATACGCCACTTCT-3′ ( BGIBMGA009721_r ) for BGIBMGA009721; 5′ -GGCAATTATAGCCGCCGTGTCC-3′ ( Vg_f ) and 5′ -GGCCAGGACTCTTTACCCGGAT-3′ ( Vg_r ) for Vg; 5′ -GACTCGTCGTGTAATGGAAAGC -3′ ( SP1_f ) and 5′ -ATGTGGGCAAGAGCATACCG -3′ ( SP1_r ) for SP1 and 5′ -CAGGCGGTTCAAGGGTCAATAC-3′ ( RP49_f ) and 5′-TACGGAATCCATTTGGGAGCAT-3′ ( RP49_r ) for the internal control , the ribosomal protein 49 ( Bmrp49 , AB48205 . 1 ) . Real-time PCR was performed in three duplicates with SYBR Green PCR Mix ( Bio-Rad ) and subjected to the Roche LightCycler 480 Real-Time PCR System . The messenger RNA quantity of each gene was calculated with the 2-ΔΔCT method and normalized to the abundance of RP49 .
|
Like other domesticates , nitrogen resources are also important for the only fully domesticated insect , the silkworm . Deciphering the way in which artificial selection acts on the silkworm genome to improve the utilization of nitrogen resources , thereby advancing human-favored domestication traits , will provide clues from a unique insect model for understanding the general rules of Darwin's theory on artificial selection . However , the mechanisms of domestication in the silkworm remain largely unknown . In this study , we focused on one important nitrogen resource , the storage protein ( SP ) . We discovered that the methionine-rich storage protein 1 ( SP1 ) , which is divergent from other SPs , is the only target of artificial selection . Based on functional evidence , together with key findings from the comprehensive comparative transcriptome , we propose that artificial selection favored higher expression of SP1 in the domestic silkworm , which would influence the genes or pathways vital for egg development and eggshell formation . Artificial selection also consistently favored activated ribosome activities and improved amide and peptide biosynthesis in the ova , like what they may act in the silk gland to increase silk-cocoon yield . We highlighted a novel case in which artificial selection could directly act on a nitrogen resource protein associated with a human-desired domestication trait .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"invertebrates",
"animal",
"types",
"moths",
"and",
"butterflies",
"artificial",
"selection",
"domestic",
"animals",
"animals",
"invertebrate",
"genomics",
"silkworms",
"zoology",
"nutrient",
"and",
"storage",
"proteins",
"proteins",
"insects",
"animal",
"genomics",
"arthropoda",
"biochemistry",
"eukaryota",
"genetics",
"biology",
"and",
"life",
"sciences",
"biosynthesis",
"genomics",
"evolutionary",
"biology",
"evolutionary",
"processes",
"organisms"
] |
2019
|
Artificial selection on storage protein 1 possibly contributes to increase of hatchability during silkworm domestication
|
The ortholog conjecture implies that functional similarity between orthologous genes is higher than between paralogs . It has been supported using levels of expression and Gene Ontology term analysis , although the evidence was rather weak and there were also conflicting reports . In this study on 12 species we provide strong evidence of high conservation in tissue-specificity between orthologs , in contrast to low conservation between within-species paralogs . This allows us to shed a new light on the evolution of gene expression patterns . While there have been several studies of the correlation of expression between species , little is known about the evolution of tissue-specificity itself . Ortholog tissue-specificity is strongly conserved between all tetrapod species , with the lowest Pearson correlation between mouse and frog at r = 0 . 66 . Tissue-specificity correlation decreases strongly with divergence time . Paralogs in human show much lower conservation , even for recent Primate-specific paralogs . When both paralogs from ancient whole genome duplication tissue-specific paralogs are tissue-specific , it is often to different tissues , while other tissue-specific paralogs are mostly specific to the same tissue . The same patterns are observed using human or mouse as focal species , and are robust to choices of datasets and of thresholds . Our results support the following model of evolution: in the absence of duplication , tissue-specificity evolves slowly , and tissue-specific genes do not change their main tissue of expression; after small-scale duplication the less expressed paralog loses the ancestral specificity , leading to an immediate difference between paralogs; over time , both paralogs become more broadly expressed , but remain poorly correlated . Finally , there is a small number of paralog pairs which stay tissue-specific with the same main tissue of expression , for at least 300 million years .
The ortholog conjecture is widely used to transfer annotation among genes , for example in newly sequenced genomes . But has been difficult to establish whether and how much orthologs share more similar functions than paralogs [1 , 2] . The most widely accepted model is that orthologs diverge slower , and that the generation of paralogs through duplication leads to strong divergence and even change of function . It is also expected that in general homologs diverge functionally with time . The test of these hypotheses poses fundamental questions of molecular evolution , about the rate of functional evolution and the role of duplications , and is essential to the use of homologs in genome annotations . Surprisingly , there are several studies which have reported no difference between orthologs and paralogs , or even the opposite , that paralogs would be more functionally similar than orthologs . Tests of the ortholog conjecture using sequence evolution found no difference after speciation or duplication in positive selection [3] , nor in amino acid shifts [4] . The debate was truly launched by Nehrt et al . [5] who reported in a large scale study , based on expression levels similarity and Gene Ontology ( GO ) analysis in human and mouse , that paralogs are better predictors of function than orthologs . Of note , methodological aspects of the GO analysis of that study were criticized by several other authors [6 , 7] . Using a very similar GO analysis but correcting biases in the data , from 13 bacterial and eukaryotic species , Altenhoff et al . [8] found more functional similarity between orthologs than between paralogs based on GO annotation analysis , but the differences were very slight . An early comparison of expression profiles of orthologs in human and mouse reported that they were very different , close to paralogs and even to random pairs [9] . Further studies , following Nehrt et al . [5] , found little or no evidence for the ortholog conjecture in expression data . Rogozin et al . [10] reported that orthologs are more similar than between species paralogs but less similar than within-species paralogs based on correlations between RNA-seq expression profiles in human and mouse . Wu et al . [11] found only a small difference between orthologs and paralogs . Paralogs were significantly more functionally similar than orthologs , but by classifying in subtypes they reported that one-to-one orthologs are the most functionally similar . The analysis was done on the level of function by looking at expression network similarities in human , mouse , fly and worm . On the other hand , the ortholog conjecture has been supported by several studies of gene expression . Contra Yanai et al . [9] , several studies have reported good correlations between expression levels of orthologs , between human and mouse [12] , or among amniotes [13] . Moreover , some studies have reported changes of expression following duplication , although without explicitly testing for the ortholog conjecture: duplicated genes are more likely to show changes in expression profiles than single-copy genes [14 , 15] . Chung et al . [16] reported through network analysis in human that duplicated genes diverge rapidly in their expression profile . Recently Assis and Bachtrog [17] reported that paralog function diverges rapidly in mammals . They analysed among other things difference in tissue-specificity between a pair of paralogs and their single copy ortholog in closely related species . They conclude that divergence of paralogs results in increased tissue-specificity , and that there are differences between tissues . Finally , several explicit tests of the ortholog conjecture have also found support using expression data . Huerta-Cepas et al . [18] reported that paralogs have higher levels of expression divergence than orthologs of the similar age , using microarray data with calls of expressed/not expressed in human and mouse . They also claimed that a significant part of this divergence was acquired shortly after the duplication event . Chen and Zhang [7] re-analysed the RNA-seq dataset of Brawand et al . [13] and reported that expression profiles of orthologs are significantly more similar than within-species paralogs . Thus while the balance of evidence appears to weight towards confirmation of the ortholog conjecture , functional data has failed so far to strongly support or invalidate it . Even results which support the ortholog conjecture often do so with quite slight differences between orthologs and paralogs [8 , 10] . Yet expression data especially should have the potential to solve this issue , since it provides functional evidence for many genes in the same way across species , without the ascertainment biases of GO annotations or other collections of small scale data . Part of the problem is that the relation between levels of expression and gene function is not direct , making it unclear what biological signal is being compared in correlations of these levels . Another problem is that the comparison of different transcriptome datasets between species suffers from biases introduced by ubiquitous genes [19] or batch effects [20] . In our analysis we have concentrated on the tissue-specificity of expression . Tissue-specificity indicates in how many tissues a gene is expressed , and whether it has large differences of expression level between them . It reflects the functionality of the gene: if the gene is expressed in many tissues then it is "house keeping" and has a function needed in many organs and cell types; tissue-specific genes have more specific roles , and tissue adjusted functions . Recent results indicate that tissue-specificity is conserved between human and mouse orthologs , and that it is functionally informative [21] . Moreover , tissue-specificity can be computed in a comparable manner in different animal datasets without notable biases , as long as at least 6 tissues are represented , including preferably testis , nervous system , and proportionally not too many parts of the same organ ( e . g . not many parts of the brain ) . Are there major differences between the evolution of tissue-specificity after duplication ( paralogs ) or without duplication ( orthologs ) ? We analyse the conservation of one-to-one orthologs and within-species paralogs with evolutionary time , using RNA-seq datasets from 12 species .
We compared orthologs between 12 species: human , chimpanzee , gorilla , macaque , mouse , rat , cow , opossum , platypus , chicken , frog , and fruit fly . Overall 7 different RNA-seq datasets were used , including 6 to 27 tissues ( see Materials and Methods ) . Three comparisons were performed with the largest sets as focal data: 27 human tissues from Fagerberg et al . , 16 human tissues from Bodymap , and 22 tissues from mouse ENCODE [22–24] . For all analyses we used tissue-specificity of expression as described in Materials and Methods . The first notable result is that tissue-specificity is strongly correlated between one-to-one orthologs . The correlations between human and four other species are presented in Fig 1A for illustration . This confirms and extends our previous observation [21] , which was based on one human and one mouse datasets . Correlation of tissue-specificity varies between 0 . 74 and 0 . 89 among tetrapods , and is still 0 . 43 between human and fly , 0 . 38 between mouse and fly . The latter is despite the very large differences in anatomy and tissue sampling between the species compared , showing how conserved tissue-specificity can be in evolution . The correlation between orthologs decreases with divergence time ( Fig 2 ) . The decline is linear . An exponential model is not significantly better: ANOVA was not significantly better for the model with log10 of time than for untransformed time for any dataset ( p > 0 . 0137 , q > 1% ) . The trend is not caused by the outlier fly data point: removing it there is still a significant decrease of correlation for orthologs ( see S1 Fig ) . Results are also robust to the use of Spearman instead of Pearson correlation between tissue-specificity values . The correlation between within-species paralogs is significantly lower than between orthologs ( ANOVA p<0 . 0137 , q<1% for all datasets ) ( Fig 2 ) . Moreover , there is no significant decline in correlation with evolutionary time ( neither linear nor exponential ) for paralogs . This may indicate almost immediate divergence of paralogs upon duplication , although other scenarios are possible ( see Discussion ) . The results are consistent using human or mouse as focal species ( Fig 2A and 2B ) . Results are also consistent using a different human RNA-seq dataset ( Fig A in S1 Fig ) . This main analysis is based on the correlation of tissue-specificity for orthologs called pairwise between species . The number of orthologs used in the analysis is thus variable ( available in Table B in S1 Table ) . An additional analysis was also performed using the same orthologs for all tetrapods , 4785 genes ( Fig B-D in S1 Fig ) . Correlations of these "conserved orthologs" are not significantly different from those observed over all orthologs . The analysis was also performed on all the datasets with tissue-specificity calculated without testis ( Fig E-G in S1 Fig ) . The correlation between orthologs becomes significantly lower ( ANOVA p = 0 . 000178 ) , while between paralogs it does not change significantly ( ANOVA p = 0 . 846 ) . Even though the correlation between orthologs becomes weaker there is still a significant difference between orthologs and paralogs ( ANOVA p = 1 . 299e-07 ) . The same analysis was also performed removing 4 other main tissues ( brain , heart , kidney and liver ) ( Fig H-K in S1 Fig ) . For the brain the correlation between orthologs becomes significantly lower ( ANOVA p = 0 . 000289 ) , but stays higher than for paralogs; for other tissues there is no significant difference . For paralogs the correlation never changes significantly . We also performed the analysis removing genes on sex chromosomes ( Fig L-N in S1 Fig ) . This analysis was done without frog , as sex chromosome information is not available . This does not change significantly the correlations between either orthologs ( ANOVA p = 0 . 856 ) or paralogs ( ANOVA p = 0 . 755 ) . In general paralogs have lower expression and are more tissue-specific than orthologs ( Fig O in S1 Fig ) , which is consistent with the dosage-sharing model [25 , 26] . Young paralogs are very tissue-specific , and get more ubiquitous with divergence time ( Fig 1B and Fig P in S1 Fig ) ; this is true for all datasets , and for τ calculated with or without testis . We also tested for asymmetry by comparing paralog pairs to the closed possible non duplicated outgroup; e . g . , we compared each Eutheria specific paralog to the non-duplicated opossum outgroup ( one-to-two ortholog; Fig 3 ) . We observe that the higher expressed paralog has a stronger correlation with the outgroup , thus appears to keep more the ancestral tissue-specificity , while the lower expressed paralog has a lower correlation and appears to become more tissue-specific ( Fig 3 ) , which is consistent with a form of neo-functionalization . When both orthologs of a pair are tissue-specific ( τ > 0 . 8 ) , they are most often expressed in the same tissue ( Fig 4 ) . The same is observed when both paralogs are tissue-specific and are younger than the divergence of tetrapods . But for Euteleostomi and Vertebrata paralogs , if both are tissue-specific then they are as likely to be expressed in the different as in same tissues; most of these are expected to be ohnologs , i . e . due to whole genome duplication . This analysis was performed on the Brawand et al . ( 2011 ) dataset , because it has the most organisms with the same 6 tissues . This result does not change after removing testis ( Fig Q in S1 Fig ) , nor changing the τ threshold from 0 . 8 to 0 . 3 ( Fig R-S in S1 Fig ) . Also after removing all tissue-specific genes ( τ > 0 . 8 ) , the difference between orthologs and paralogs is smaller but stay significant ( ANOVA p = 0 . 001 ) ( Fig T in S1 Fig ) .
Our results show that most genes have their tissue-specificity conserved between species . This provides strong new evidence for the evolutionary conservation of expression patterns . Using tissue-specificity instead of expression values allows easy comparison between species , as bias of normalisation or use of different datasets has little effect on results [21] . All of our results were confirmed using three different focus datasets , from human or mouse , and thus appear to be quite robust . The conservation of expression tissue-specificity of protein coding genes that we find is high even for quite distant one-to-one orthologs: the Pearson correlation between τ in human or mouse and τ in frog is R = 0 . 74 ( respectively R = 0 . 66 ) over 361 My of divergence . Even between fly and mammals it is more than 0 . 38 . Moreover , this tissue-specificity can be easily compared over large datasets without picking a restricted set of homologous tissues ( e . g . in [7 , 13] ) . The correlation between orthologs is strongest for recent speciations , and decreases linearly with divergence time . This decrease shows that we are able to detect a strong evolutionary signal in tissue-specificity , which has not always been obvious in functional comparisons of orthologs ( e . g . [5 , 8] ) . Correlation between within-species paralogs is much lower than between orthologs . Whereas the expression of young paralogs has been recently reported to be highly conserved [17] , we find a large difference between even very young paralogs in tissue-specificity . In Assis and Bachtrog [17] , the measure of tissue-specificity is not clearly defined , but it seems to be TSI [27] , which performed poorly as an evolutionarily relevant measure in our recent benchmark [21]; they also treated female and male samples as different "tissues" , confounding two potentially different effects . The low correlation that we observed for young paralogs does not decrease significantly with divergence time . It is possible that on the one hand paralogs do diverge in tissue-specificity with time , and that on the other hand this trend is compensated by biased loss of the most divergent paralogs . It is also possible that we lack statistical power to detect a slight decrease in correlation of paralogs , due to low numbers of paralogs for many branches of the phylogeny . The most likely interpretation is that for small-scale paralogs ( defined as not from whole genome duplication [28] ) there is an asymmetry , with a daughter gene which lacks regulatory elements of the parent gene upon birth; further independent changes in tissue-specificity in each paralog would preserve the original lack of correlation . In any case , we do not find support for a progressive divergence of tissue-specificity for paralogs . The overall conservation of tissue-specificity could be due to a subset of genes , and most notably sex-related genes . Indeed , the largest set of tissue-specific genes are testis-specific [21] . To verify the influence of sex-related genes , we performed all analyses without testis expression data , or without genes mapped to sex chromosomes . After removing testis expression from all datasets the correlation between paralogs does not change significantly , while between orthologs is gets significantly weaker . The lower correlation of orthologs suggests that testis specific genes are conserved between species , and as they constitute a high proportion of tissue-specific genes , they contribute strongly to the correlation . Removing sex chromosome located genes does not change results significantly . After removing testis expression the differences of conservation of tissue-specificity between orthologs and paralogs stay significant . Overall , it appears that tissue-specificity calculated with testis represents a true biological signal , and given its large effect it is important to include this tissue in analyses . In general paralogs are more tissue-specific and have lower expression levels . This could be explained if ubiquitous genes are less prone to duplication or duplicate retention . Yet we do not observe any bias in the orthologs of duplicates towards more tissue-specific genes ( Fig 3; see also S1 Fig ) . With time both paralogs get more broadly expressed ( Fig 1 and Fig P in S1 Fig ) . In the rare case where both paralogs are tissue-specific , small-scale young paralogs are expressed in the same tissue , while genome-wide old paralogs ( ohnologs ) are expressed in different tissues ( Fig 4 ) . With the data available , we cannot distinguish the effects of paralog age and of duplication mechanism , since many old paralogs are due to whole genome duplication in vertebrates , whereas that is not the case for the young paralogs . In many cases the higher expressed paralog has a similar tissue-specificity to the ancestral state , while the lower expressed paralog is more tissue-specific ( Fig 3 ) . We have studied gene specificity without taking in account alternative splicing , or the possibility that different transcripts are expressed in different tissues , because it is still difficult to call transcript level expression reliably [29] . This would probably not change our main observations , that tissue-specificity is conserved among orthologs , diverges with evolutionary time , and follows the ortholog conjecture . Of note , recent results have not supported an important role of alternative splicing for differences in transcription between tissues [30 , 31] . The overall picture that we obtain for the evolution of tissue-specificity is the following . In the absence of duplication , tissue-specificity evolves slowly , thus is mostly conserved , and tissue-specific genes do not change their main tissue of expression ( Figs 2 and 4 ) . After small-scale duplication ( i . e . , not whole genome ) paralogs diverge rapidly in tissue-specificity , or already differ at birth . This difference is mostly due to the less expressed paralog losing the ancestral specificity , while the most expressed paralog keeps at first closer to the ancestral state , as estimated from a non-duplicated outgroup ortholog ( Fig 3 ) . But over time , even the most expressed paralog diverges much more strongly than a non-duplicated ortholog . While paralog divergence is rapid , in the small number of genes which stay tissue-specific for both paralogs the main tissue of expression is mostly conserved , for several hundred million years ( i . e . origin of tetrapods , Fig 4 ) . With increasing age of the paralogs , they both tend to become more broadly expressed ( Fig 1 and Fig P in S1 Fig ) while keeping a low correlation . For whole genome duplicates we have less information , because of the age of the event in vertebrates and the lack of good outgroup data . The main difference is that when two genome duplication paralogs are both tissue-specific , they are often expressed in different tissues ( Fig 4 ) . We have used tissue-specificity to estimate the conservation of function , rather than Gene Ontology annotations or expression levels . We believe that this metric is less prone to systematic errors , whether annotation biases for the Gene Ontology , or proper normalisation between datasets and choice of few tissues for expression levels . Our results confirm the Ortholog Conjecture on data which is genome-wide and functionally relevant: orthologs are more similar than within-species paralogs . Moreover , orthologs diverge monotonically with time , as expected . On the contrary , even young paralogs show large differences .
Additional Supplementary files are available on Figshare: https://dx . doi . org/10 . 6084/m9 . figshare . 3493010 . v2
|
From specific examples , it has been assumed by comparative biologists that the same gene in different species has the same function , whereas duplication of a gene inside one species to create several copies allows them to acquire different functions . Yet this model was little tested until recently , and then has proven harder than expected to confirm . One of the problems is defining "function" in a way which can be easily studied . We introduce a new way of considering function: how specific is the activity ( "expression" ) of a gene ? Genes which are specific to certain tissues have functions related to these tissues , whereas genes which are broadly active over many or all tissues have more general functions for the organism . We find that this "tissue-specificity" evolves very slowly in the absence of duplication , while immediately after duplication the new gene copy differs . This shows that indeed duplication leads to a strong increase in the evolution of new functions .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"Methods"
] |
[
"vertebrates",
"mice",
"animals",
"mammals",
"primates",
"analysis",
"of",
"variance",
"amphibians",
"mathematics",
"statistics",
"(mathematics)",
"genome",
"analysis",
"mammalian",
"genomics",
"research",
"and",
"analysis",
"methods",
"mathematical",
"and",
"statistical",
"techniques",
"gene",
"expression",
"statistical",
"methods",
"gene",
"ontologies",
"evolutionary",
"genetics",
"animal",
"genomics",
"rodents",
"gorillas",
"genetics",
"biology",
"and",
"life",
"sciences",
"apes",
"physical",
"sciences",
"genomics",
"evolutionary",
"biology",
"amniotes",
"computational",
"biology",
"frogs",
"organisms"
] |
2016
|
Tissue-Specificity of Gene Expression Diverges Slowly between Orthologs, and Rapidly between Paralogs
|
Understanding the roles of neutrophils and macrophages in fighting bacterial infections is a critical issue in human pathologies . Although phagocytic killing has been extensively studied , little is known about how bacteria are eliminated extracellularly in live vertebrates . We have recently developed an infection model in the zebrafish embryo in which leukocytes cannot reach the injected bacteria . When Escherichia coli bacteria are injected within the notochord , both neutrophils and macrophages are massively recruited during several days , but do not infiltrate the infected tissue presumably because of its tough collagen sheath . Nevertheless , the bacteria are killed during the first 24 hours , and we report here that neutrophils , but not macrophages are involved in the control of the infection . Using genetic and chemical approaches , we show that even in absence of phagocytosis , the bactericidal action relies on NADPH oxidase-dependent production of superoxide in neutrophils . We thus reveal a host effector mechanism mediated by neutrophils that eliminates bacteria that cannot be reached by phagocytes and that is independent of macrophages , NO synthase or myeloperoxidase .
The innate immune system is the first line of defence of the host . It includes large phagocytes ( such as macrophages and granulocytes ) equipped with a battery of weapons to destroy the invader within minutes or hours . Since the seminal work of Elie Metchnikoff [1] , the defence mechanisms relying on leukocytes remain a challenging subject . When microbes penetrate the epithelial barrier , macrophages and neutrophils are rapidly recruited and upon contact , engulf the bacteria into a vacuole called a phagosome that fuses with intracellular granules or lysosomes to form a lytic vacuole in which bacteria may be killed by a wide variety of mechanisms involving chemicals and enzymes [2 , 3] . Non-oxidative effectors include antimicrobial proteins , while the oxygen-dependent mechanism , also known as the respiratory burst , involves the generation of reactive oxygen species ( ROS ) [4 , 5 , 6] . ROS production inside the phagocytic vacuole involves NADPH oxidase and the major ROS , superoxide ( O2- ) and hydrogen peroxide ( H2O2 ) , can directly or indirectly promote the death of the microbe , according to the nature of the pathogens [7 , 8] . Nitric oxide ( NO ) , produced by NO synthase , can contribute to microbicidal activity and is essential for the defence against intracellular organisms such as Salmonella enterica and mycobacteria [9 , 10] . Many microbes manage to survive within macrophages after phagocytosis . While some cope with the phagolysosomal conditions ( S . enterica serovar Typhimurium [11] ) , others like Listeria , Shigella and some mycobacteria [12 , 13 , 14] are able to block the maturation of the phagosome or even to escape from these compartments . Host cells , however , have developed counter strategies to fight cytosolic bacteria including directing them to autophagosomes [15] . While microbe killing inside the phagosome has been extensively studied , it is less well understood how phagocytes are capable of killing microbes extracellularly in whole organisms . Neutrophils can fight bacterial pathogens without phagocytosis either by release of toxic granule contents ( degranulation ) [16] or by expelling neutrophil extracellular traps ( NETs ) , which are networks of extracellular fibres built upon expulsion of chromatin [17] . However , events such as these are very hard to disentangle from phagocytosis-mediated killing in the full context of tissue infection . Thanks to its transparency and genetic amenability , the zebrafish embryo is a useful model for the study of host/pathogen interactions in vivo . The zebrafish model has been used to evaluate the respective roles of neutrophils and macrophages in eliminating invading bacteria [10 , 18 , 19]; this relies not only on the nature of the invading microbe , but also on the route and anatomical site of infection . One striking observation was that macrophages are very efficient at engulfing microbes from body fluids ( “flypaper” strategy ) while neutrophils may be very efficient at clearing surface associated microbes in a “vacuum-cleaner”-like behaviour [20] . We have recently developed an infection model in the zebrafish embryo in which the bacteria are trapped in a tissue in which macrophages and neutrophils cannot enter . When non-pathogenic Escherichia coli ( E . coli ) bacteria are injected in the notochord , the swollen rod that provides axial stiffness to the developing embryo , they slide between notochord cells and the thick cylindrical collagen sheath that encases the cord . Although unable to thread their way through this envelope , neutrophils and macrophages are massively recruited all along the infected notochord where they stay in a highly activated state for days . Interestingly , these inaccessible bacteria are cleared within the first 24 hours [21] . Here we address the mechanisms of E . coli clearance in the notochord infection model where professional phagocytes cannot directly encounter the injected bacteria . We first investigate whether macrophages or neutrophils are involved in this clearance and then investigate the nature of the molecules instrumental for bacterial killing .
We previously showed that K12 Escherichia coli cells injected in the notochord of zebrafish embryos cannot be reached by phagocytes , yet are killed in one day [21] . We confirmed the physical separation of freshly injected K12 from phagocytes by the notochord collagen matrix ( S1A and S1B Fig ) . To verify that this is not a quirk of this laboratory strain , we first compared enteric adherent invasive E . coli strains , E . coli AIEC LF82 and its mutant , LF82-ΔlpfA , E . coli JM83-ΔmsbB strain and laboratory K12 strain in our notochord infection model . We observed that they behaved similarly ( S1C and S1D Fig ) . We therefore went on using the laboratory K12 strain . To investigate the role of macrophages in the observed bacterial clearance , we injected liposome-encapsulated clodronate ( Lipo-clodronate ) that kills phagocytic macrophages [22 , 23] . At 1 day post-fertilization ( dpf ) , macrophage/neutrophil dual reporter embryos , tg ( mpeg1:mCherry-F ) /tg ( mpx:GFP ) , or macrophage reporter embryos , tg ( mpeg1:mCherry-F ) , were injected with 10 nl of Lipo-Clodronate in the posterior caudal vein ( intravenous , i . v . ) . As previously described [22] 24 h after Lipo-Clodronate injection , macrophages were efficiently eliminated without affecting the neutrophil population , nor inducing unspecific toxicity ( Fig 1A and 1B ) . This was correlated with the decrease of mpeg1 mRNA expression in Lipo-Clodronate treated larvae compared to Lipo-PBS controls , as shown by RT-qPCR ( Fig 1C ) . To further confirm the efficiency of lipo-clodronate to suppress macrophage population , we generated another macrophage reporter line with microfibrillar-associated protein 4 ( mfap4 ) promoter whose expression is strong and stable in zebrafish macrophages [24] , i . e . the tg ( mfap4:mCherry-F ) line . Injection of Lipo-clodronate in tg ( mfap4:mCherry-F ) induced a dramatic reduction in the number of mfap4+ cells ( Fig 1D and 1E ) , showing the suitability of this approach to deplete macrophages . Macrophage depleted larvae were selected and injected in the notochord with fluorescent E . coli . We observed that bacteria were cleared within the first 24 hours post infection ( hpi ) in both , macrophage-depleted larvae , as well as in control Lipo-PBS injected larvae , as revealed by fluorescence microscopy and CFU counts ( Fig 1F and 1G ) . Importantly , upon notochord infection , neutrophils were normally recruited around the infected notochord regardless of the presence or absence of macrophages ( Fig 1H ) . To confirm , that macrophages are not fundamental for bacterial clearance in notochord infection model , we ablate macrophages using tg ( mpeg1:Gal4 / UAS:nfsB-mCherry ) embryos in which macrophage express gene 1 promoter indirectly drives the expression of E . coli nitroreductase enzyme in macrophages . Treatment of tg ( mpeg1:Gal4/UAS:nfsB-mCherry ) embryos with the pro-drug metronidazole ( MTZ ) at 30 hpf ( hours post-fertlilization ) specifically decreased macrophage number at 1 and 2 days post-treatment ( dpT ) ( S2A and S2B Fig ) . Tg ( mpeg1:Gal4/UAS:nfsB-mCherry ) were then infected with E . coli-GFP at 2 dpf in the notochord . MTZ-mediated macrophage depletion did not impact the bacterial burden at 1 dpi ( day post-infection ) as shown by Fluorescent Pixel Counts ( FPC ) ( S2C and S2D Fig ) . Altogether , these data show that macrophages are not required for bacterial clearance in this model . To investigate the role of neutrophils in bacterial clearance , we ablated neutrophils by two independent approaches . First , we specifically inhibited neutrophil development and function by knocking down the G-CSF/GCSFR pathway using a morpholino oligonucleotide ( MO ) specifically blocking gcsfr/csf3r translation ( MO csf3r ) [25 , 26] . Injection of MO csf3r in the neutrophil reporter embryos , tg ( mpx:GFP ) , led to approximately 70% reduction in the total number of neutrophils as compared to larvae injected with a control morpholino ( MO CTRL ) at 3 dpf ( Fig 2A , 2C and 2D ) . We infected these morphants with 2500 CFUs fluorescent E . coli . Bacteria disappeared in the control larvae ( Fig 2B and 2E ) while they proliferated in neutrophil-depleted embryos ( Fig 2B and 2F ) . The bacterial proliferation correlated with a further dramatic reduction in neutrophil number at 1 and 2 dpi ( days post infection ) , suggesting neutrophil death ( Fig 2D ) . Subsequently , infected csf3r morphants died between 2 and 3 dpi ( Fig 2G ) with overwhelming bacterial proliferation and neutropenia ( S3B Fig ) . We also ablated neutrophils , using tg ( mpx:Gal4/UAS:nfsB-mCherry ) embryos in which the myeloperoxidase promoter ( mpx ) indirectly drives the expression of nitroreductase in neutrophils . Treatment of tg ( mpx:Gal4/UAS:nfsB-mCherry ) embryos with metronidazole at 40 hpf specifically depleted neutrophils at 1 and 2 days post-treatment ( Fig 3A ) . Since macrophages are required to clear apoptotic cells , we asked whether neutrophil death in MTZ treatment alters macrophage number or distribution in the triple transgenic line tg ( mpx:Gal4/UAS:nfsB-mCherry/ mpeg1:GFPcaax ) . At 1 dpT , MTZ treatment did not affect the number of macrophages and they were similarly distributed throughout the larva to the control ( Fig 3B and 3C ) . Larvae were then infected with E . coli-crimson and 4 hours after E . coli injection , macrophages were recruited to the infected notochord in both MTZ and DMSO conditions , showing that ablation of neutrophil using nfsB/MTZ system does not impair macrophage response ( Fig 3D ) . Infection outcome was then analysed in tg ( mpx:Gal4/UAS:nfsB-mCherry ) larvae infected with fluorescent E . coli-GFP . Similarly to csf3r morphants , bacteria were cleared in control larvae ( nfsB+ DMSO and nfsB- MTZ ) , while bacteria proliferated in embryos with low neutrophil density ( nfsB+ MTZ ) , as shown by fluorescent microscopy and by quantification of bacterial burden ( Fig 3E and 3F ) . These experiments demonstrate that neutrophils are essential for the control of notochord infection by E . coli . We further investigated the relationship between neutrophil supply and bacterial disappearance in the notochord . Normal neutrophil levels were able to eliminate small amounts of bacteria ( S3A Fig ) , but embryos with depressed neutrophil populations did not survive low bacterial loads ( S3B Fig ) , while a higher bacterial inoculum overcame larvae with a normal neutrophil population ( S3C Fig ) . However , by artificially increasing neutrophil density in the developing embryo through overexpression of gcsfa , we observed that increasing neutrophil density allow the embryo to cope with even higher amounts of injected bacteria ( S3D Fig and S4A and S4C Fig ) . Similar results were observed by overexpressing gcsfb ( S4 Fig ) . Our data reveals that the balance of neutrophils versus bacteria is instrumental for the outcome of the infection and that neutrophil populations are limiting in fighting the infection . To evaluate cell death , Sytox Green , a vital dye which labels DNA of dying cells , was injected into the vein of infected tg ( lyz:DsRed ) larvae . While PBS and low dose E . coli induced few cell death around the notochord , embryos experiencing neutropenia ( i . e . infected with high dose E . coli ) displayed increased cell death including dead neutrophils ( S5 Fig ) . This suggests that when the neutrophil versus bacteria balance is not correct , neutrophils die by apoptosis . Of note , by contrast to neutrophil , macrophage number did not decrease , but instead increased 2 days after high dose infection ( S6 Fig ) . These results are reminiscent to what happen in mammals in which neutrophil/bacteria ratio is fundamental for host defence [27] . Our previous study revealed that approximately one-third of recruited neutrophils degranulate around infected notochords [21] . We therefore investigated the role of the neutrophil-specific myeloperoxidase ( Mpx ) that is present in the azurophilic granules , in bacterial clearance . We introduced the mpx:GFP transgene in the mpx-null mutant ‘spotless’ [28] to generate tg ( mpx:GFP ) /mpx-/- offspring in which neutrophils express the eGFP but lack Mpx activity . Active MPX in neutrophil granules can be visualized in zebrafish embryos using Sudan black staining [29] . Sudan Black staining confirmed that neutrophils did not carry Mpx activity in tg ( mpx:GFP ) /mpx-/- while in tg ( mpx:GFP ) /mpx+/- siblings , neutrophils contained active Mpx in their granules ( Fig 4A ) . A low dose of fluorescent E . coli was injected in the notochord of 2 dpf tg ( mpx:GFP ) /mpx-/- embryos; neutrophils were normally recruited along the notochord , and the injected E . coli were cleared at 1 dpi as in the wild type ( Fig 4B ) . Mpx is therefore not required for the clearance of E . coli in the notochord . Neutrophils use different diffusible molecules to fight infections , including NO and ROS . We investigated NO production by neutrophils during the course of notochord infections using the NO reporter fluorescent probe DAF-FM-DA . We used Salmonella infected embryos as positive controls to detect NO production in neutrophils within the Aorta-Gonad-Mesonephros ( AGM ) ( S7A Fig ) [30] . As described [31] , the notochord itself was labelled by DAF-FM-DA in uninfected embryos , but we could not observe any evidence of NO production by neutrophils in our notochord infection model ( S7B Fig ) . L-NAME was previously shown to specifically inhibit NO synthases in zebrafish larvae [30] . To block NO production in our system , we thus treated larvae with L-NAME and injected E . coli into the notochord . We did not observe any difference in the outcome of the infection between L-NAME-treated larvae and controls ( DMSO ) ( S7C Fig ) . The phagocyte NADPH oxidase and ROS production play a key role in the elimination of engulfed bacteria [4] . To detect intracellular ROS accumulation in the form of superoxide anions in tg ( mpx:GFP ) embryos infected with E . coli , we used Dihydroethidium ( DHE ) , a cell permeable probe that fluoresces in red after reacting with superoxide within the cell [32 , 33] . First , we imaged the injection site , where some bacteria initially leaked from the pierced notochord and got engulfed by neutrophils and observed that these phagocytosing leukocytes , abundantly produced superoxide in intracellular compartments harboring bacteria , which are most probably phagosomes ( Fig 5A and 5B ) . Green fluorescent E . coli were rapidly lysed within 20 minutes in the putative phagosome ( Fig 5B and 5C and S1 Video ) . We then imaged the upstream region , where bacteria are separated from the recruited neutrophils by the notochord collagen sheath . Interestingly , these recruited neutrophils also produced large amounts of superoxide , even though they had not phagocytosed bacteria ( Fig 5D ) . DHE was also detected at a basal level in notochord surrounding tissues ( Fig 5E ) . To test the specificity of DHE staining in detecting superoxide anions we treated infected embryos with N-acetyl-cysteine ( NAC ) , a broad-specificity ROS scavenger . We observed a general decrease of DHE staining within cells of the trunk and more particularly a decrease of DHE+ recruited cells ( Fig 5E and 5F ) and of DHE+ recruited neutrophils ( Fig 5E and 5G ) around the infected notochord while the number of recruited neutrophils was unchanged by the treatment ( Fig 5E and 5H ) , confirming that DHE probe specifically detects ROS in this model . To investigate whether this superoxide production could be involved in bacterial killing , we used Apocynin , a NADPH oxidase ( NOX ) inhibitor [34 , 35] . Upon notochord infection , Apocynin-treated embryos had reduced number of superoxide producing cells , including recruited DHE+ neutrophils at the inflammation site , as compared to DMSO-treated larvae ( Fig 6A and 6B ) , showing the efficiency of Apocynin as a NOX inhibitor in zebrafish . To test whether Apocynin alters the steady state of neutrophils , tg ( mpx:GFP ) larvae were treated with this drug at 2 dpf . Apocynin treatment decreased the total number of neutrophils after 6 or 24 h of treatment , but by less than 15% ( Fig 6C and 6D ) , showing that this approach is suitable to test the role of NOX in zebrafish neutrophils . Therefore , we infected tg ( lyz:DsRed ) embryos with a very low dose of E . coli ( <1000 CFUs ) in the notochord . Even with the very low dose infection , 80% of Apocynin-treated embryos failed to clear the bacteria , while all bacteria were efficiently killed in DMSO-control embryos ( Fig 6E ) . Apocynin-treated embryos displayed unrestricted bacterial growth in the notochord at 1 dpi , as demonstrated with fluorescence microscopy ( Fig 6E and 6F ) . This was correlated with neutropenia and eventually death at 2–3 dpi ( Fig 6F and 6G ) . The effect was specific to the clearance of bacteria in this notochord infection model since Apocynin treatment did not interfere with the clearance of bacteria injected in the muscle , where phagocytosis occurs ( S8 Fig ) . Similar results were obtained using another NOX inhibitor [36] , VAS2870 ( VAS ) ( S9 Fig ) . Interestingly , in mammals , Apocynin activity requires that target cells do express an active Mpx [35] . Therefore , we compared the results of Apocynin treatment in mpx-/- and mpx+/+ infected embryos , and observed that Apocynin increased susceptibility to notochord infection only in the presence of Mpx ( Fig 6H and 6I ) . Thus , Apocynin action is also dependent on Mpx in zebrafish , and thus specifically acts on neutrophils . Overall , these data thus strongly suggest that inhibition of superoxide production in neutrophils increases susceptibility to notochord infection . To further examine the role of phagocyte NOX , morpholino-mediated gene knockdown was used . Injection of p47phox MO in tg ( mpx:GFP ) did not induce noticeable morphological defects , but , as expected , decreased superoxide production in neutrophils following infection compared to control morpholino ( CTRL MO ) ( S10 Fig ) . To address the effect p47phox MO on the development and the recruitment of neutrophil , we analyzed tg ( mpx:GFP ) p47phox morphants before and after E . coli infection in the notochord at 2 dpf . Although p47phox morphants displayed 20% less neutrophils than in control morphants , ( Fig 7A and 7B ) these leukocytes were recruited in normal numbers to the notochord at 4 hpi and 1 dpi ( Fig 7C ) , showing that p47phox morphants can mobilize neutrophils properly during the infection . Then , p47phox morphants were infected in the notochord with E . coli-GFP . P47phox MO induced higher bacterial burden as evidenced by fluorescence microscopy ( Fig 7D ) and CFUs counts ( Fig 7E ) . This was correlated with an increase in the severity of infection ( Fig 7F ) . As neutrophils are instrumental for larva survival and bacterial clearance during notochord infection and as pharmacological ( apocynin and VAS2870 ) and genetic ( p47phox morpholino ) inhibition caused a slight decrease of neutrophil numbers , we tested whether inducing high neutrophil number in the context of NADPH incompetence could restore survival of the infected larvae . One-cell stage tg ( lyz:DsRed ) embryos were thus injected with gcsfa expressing plasmid and 2 days later were treated either with DMSO or VAS2870 ( Fig 8A ) . Beside the fact that gcsfa forced expression increased the number of neutrophils compared to controls ( Fig 8B ) , it did not restore a better survival of the infected larvae in the presence of Nox inhibitor VAS2870 ( Fig 8C ) . Altogether these data show that NOX-induced superoxide is necessary for bacteria elimination at a distance by neutrophils .
Many studies have used the zebrafish embryo model to address the respective roles of neutrophils and macrophages in eliminating invading bacteria , but in all instances , at least one of these two cellular populations had direct access to the bacteria . In our model neither neutrophils nor macrophages could reach the bacteria . We first observed an active recruitment of both macrophages and neutrophils around the infected notochord that is correlated with the elimination of the bacteria in the notochord within 24 hours . Specifically depleting individual myeloid populations , we have investigated their contribution in the clearance of E . coli at a distance and describe molecular pathways involved in bacterial elimination by neutrophils . Using chemical and genetic ablation of macrophages , we revealed that despite being massively recruited to the notochord , macrophages are not required for the bacterial killing . By contrast , whichever the strategy to lower the amount of neutrophils within the developing zebrafish , the embryo becomes unable to cope even with low-dose infection , leading to bacterial proliferation and death of the embryo , showing that neutrophils are essential to control notochord infection . Further analysis should reveal whether other mechanisms are also involved in the death of E . coli within the notochord , such as complement-mediated killing or killing by the notochordal cells . Furthermore , we highlight the importance of the numerical balance between neutrophils and bacteria to the outcome of notochord infection in which phagocytosis is not feasible . This observation suggests that the bactericidal molecules produced by the neutrophils to fight the bacteria are produced in limiting quantities . During Salmonella infections , the correct population of neutrophils is maintained through a mechanism of demand-driven granulopoiesis in the main site of hematopoietic stem cells emergence , i . e . , the AGM [30] . Similarly , we observed here , that in low dose E . coli infections , the host is able to increase the neutrophil pool to control notochord infection . However , too low a neutrophil/bacteria ratio ( either by increasing bacterial load or decreasing the number of neutrophils ) results in bacterial proliferation , onset of neutropenia , and death within 2 to 3 dpi . Conversely , the neutrophil-enriched embryos can cope with a very high dose of bacteria . These data are reminiscent of results in human where the maintenance of a proper pool of neutrophil is critical for effective bacterial killing [27 , 37 , 38] , emphasizing thus the relevance of the tractable zebrafish larvae system for the study of dynamic interactions between neutrophil bactericidal activity and bacteria in vivo . To capture and kill microbes they cannot phagocytize , neutrophils have been described to expel their chromatin to form Neutrophil Extracellular Traps ( NETs ) , but this may lead to neutrophil death ( Netosis ) [39 , 40] . NET formation relies on complex intracellular processes involving the activity , among others , of myeloperoxidase [41] . We report here that myeloperoxidase activity is not necessary to fight the infection in our experimental system . This shows that MPX dependent-NET formation is not responsible for bacterial killing at a distance . However Myeloperoxidase may not be required with all stimuli , since MPO was shown to be dispensable for NET induction in infections with Pseudomonas aeruginosa or Staphylococcus aureus . Therefore , we cannot exclude the involvement of MPO-independent NETs in our system [42] . We report here that NOX activity and the production of superoxide by neutrophils are essential to cope with notochord infection by E . coli . Indeed , using fluorescent probes , we showed that neutrophils swarm around the notochord and produce large amounts of superoxide . Treatments of the embryos with inhibitors of NOX assembly , VAS2870 and Apocynin , or the specific knock down of Nox subunit p47phox using morpholinos , lead to bacterial proliferation and increased severity of the infection . This is accompanied with the decrease of superoxide production in neutrophils , consistent with an essential role of superoxide in the clearance of E . coli without direct phagocytosis ( Fig 9 ) . Apocynin activity was shown to be dependent on the presence of myeloperoxidase in neutrophils [35] . In our model , Apocynin has almost no activity in mpx-/- mutant , reinforcing the specificity of its effect . This demonstrates that Nox activity in neutrophils is required for bacterial clearance in the notochord . The present work raises different questions related to the death of the different actors , the bacteria , the neutrophils , and the embryo . Foremost is the question as to how bacteria are killed at a distance by neutrophils . Neutrophils massively degranulate around the infected notochord [21] and we show here that an oxidative burst is necessary for bacterial elimination . Superoxide is known to be weakly bactericidal [4 , 43] , but is rapidly converted to hydrogen peroxide by dismutation . Although products of NADPH oxidase are soluble , they are rapidly consumed by reactions with other targets within a limited diffusion distance [44]; however we cannot exclude the possibility that these ROS diffuse through the very thin ( <1 μm ) collagen sheath . A more possible scenario , would be that superoxide is not involved in a direct killing mechanism but instead is interacting with a host- or microbe- derived species , triggering a superoxide-dependent process ( Fig 9 ) . Indeed , besides inducing oxidative stress , ROS also serve as signalling molecules to regulate biological processes . One of the best-understood mechanism of redox signalling involves H2O2-mediated oxidation of cysteine residues within proteins , altering thus their function [45] . These reversible modifications could trigger activation of signalling cascade and the release of bactericidal agents . Another important target of ROS is the transcription factor NF-κB which is known to control many aspects of the immune response [46] . Therefore neutrophil superoxide may act as a second messenger of a killing strategy at a distance . Why do neutrophils die when the bacteria/neutrophil ratio is too high in favor of the invaders ? If bacteria proliferate within the infected notochord , then neutrophils massively die , and the embryo becomes neutropenic . This could be due to a factor released by the densely packed bacteria within the notochord . However , there may be no reason why this virulence factor would specifically kill neutrophils while sparing the highly endocytic macrophages that are also massively recruited to the notochord but not affected by bacterial proliferation . For this reason , we propose that death of neutrophils could rather be a consequence of the excessive concentration of bacteria-derived molecules , similarly to a quorum sensing mechanism , triggering hyper activation of the neutrophils and leading to their death [47] . This hyper activation , akin to a local cytokine storm is likely also responsible for the death of the embryo in cases where E . coli proliferates within the notochord . Importantly , we have no indication that the bacteria used in this study could kill the embryo by themselves . We consider that in cases where the embryos die , it is the consequence of their heavy inflammatory status mimicking a cytokine storm . This hypothesis is consistent with the similar outcome observed with pathogenic and non-pathogenic E . coli strains , as well as with our experiments with mycobacteria . We have demonstrated that mycobacteria can replicate within the notochord ultimately leading to notochord break down , without triggering the heavy inflammation described here with E . coli . The subsequent fate of the embryo depends on the virulence of the mycobacteria . The non-virulent Mycobacterium smegmatis is eliminated by phagocytosis , leading to the host survival while M . marinum resists destruction by phagocytosis and keeps proliferating until the host dies [48] . Conversely , E . coli only effectively kills infected embryos when injected alive in excessive amounts in the notochord where this triggers a heavy inflammation that kills the neutrophils and ultimately the embryo . To overcome killing by neutrophils , some pathogenic bacteria developed strategies to avoid contact with phagocytes . Some pathogens invade tissues that are inaccessible to phagocytes , while other employ strategies to prevent engulfment [3] . They harbor on their surfaces molecules preventing recognition by phagocytes , such as capsular antigens O75 and K5 of uropathogenic Escherichia coli ( Burns and Hull , 1999 ) and polysaccharide capsules of Streptococcus pneumoniae that increase the resistance to phagocytosis . Staphylococcus aureus secretes the 16 kD Extracellular fibrinogen binding protein that blocks its phagocytosis by human neutrophils by forming a “capsule”-like shield [49] . By contrast , Yersinia pestis ( the agent of bubonic and pneumonic plaque ) , Yersinia pseudotuberculosis ( gastroenteritis ) and Yersinia enterocolitica ( gastroenteritis and mesenteric adenitis ) are able to inhibit the actin cytoskeleton required for engulfment , through the secretion of effector proteins into the cytoplasm of the immune cell , leading to decreased phagocytosis by neutrophils and increased virulence [3] . Oxidative burst at a distance might be an alternative mechanism employed by neutrophils to prevent such escape mechanisms . Further investigations should determine whether host targeted therapeutic strategies may be beneficial against medically relevant infections , especially in patients suffering from Chronic Granulomatous Disease whose neutrophil function is deficient for NADPH activity .
Animal experimentation procedures were carried out according to the European Union guidelines for handling of laboratory animals ( http://ec . europa . eu/environment/chemicals/lab_animals/home_en . htm ) and were approved by the Comité d'Ethique pour l'Expérimentation Animale under reference CEEA-LR-13007 and APAFIS#5737–2016061511212601 v3 . Fish husbandry and experiments were performed at the University of Montpellier . Embryos were obtained from the University of Montpellier and the Institut Pasteur . Experiments were performed on 0 hour to 5 days past fertilization stages when the embryos were used . Fish maintenance , staging and husbandry were performed as described [21] with golden strain and transgenic lines . Tg ( mpeg1:mCherry-F ) ump2 , referred as tg ( mpeg1:mCherry-F ) [50] , tg ( mpeg1:GFPcaax ) [51] and tg ( mfap4:mCherry-F ) ( ump6tg , present study ) were used to visualize macrophages . Tg ( mpx:GFP ) i114 and tg ( lyz:DsRed ) nz50 used to label neutrophils and the mpxt30963/t30963 null ‘spotless’ mutant , are referred here as tg ( mpx:GFP ) [52] , tg ( lyz:DsRed ) [53] and mpx-/- [28] , respectively . Tg ( rcn3:gal4 ) ( PD1023 ) crossed with tg ( UAS:mCherry ) ( PD1112 ) were used to visualize notochordal cells [54] . Tg ( mpx:Gal4/UAS:nfsB-mCherry ) was used to ablate neutrophils [55] . Tg ( mpeg1:Gal4/UAS:nfsB-mCherry ) was used to ablate macrophages [26] . Embryos were obtained from pairs of adult fishes by natural spawning and raised at 28 . 5°C in tank water . Embryos and larvae were staged according to [56] . The Mfap4 promoter used to drive the specific expression of membrane-targeted mCherry in macrophages was amplified using the upstream primer zMfap4_3P1 ( 5’ ATC CAT GCC CTT CGA CTG TT 3’ ) and the zMfap4_123E2N primer matching the start of the second exon of the Mfap4 gene ( 5’ TAT AGC GGC CGC ACA GCA CGA TCT AAA GTC ATG AA 3’ ) . The 2 . 4 kb amplified fragment was digested by NotI , and ligated to the coding phase of the farnesylated mCherry protein so that the Mfap4 AUG is in phase with the downstream mCherry-F ORF on a I-SceI meganuclease and Tol2-derived vector ( GenBank accession no . GU394080 ) . The resulting plasmid was injected , together with I-SceI meganuclease , into embryos at the one-cell stage . E . coli K12 or Salmonella enterica serovar Typhimurium ( here called Salmonella ) carrying plasmids encoding GFP or DsRed fluorescent proteins were injected in the notochord of 2 dpf embryos as described [21] . Four different doses of E . coli were used: very low ( 1000 CFU ) , low ( <3000 CFU ) , high ( 3000<n<6000 CFU ) and very high ( >7000 CFU ) . 3000 CFU of Salmonella were injected in the hindbrain or in the notochord . Enteroinvasive E . coli AIEC bacteria strain LF82 [57] and its mutant , LF82-ΔlpfA [58] and JM83ΔmsbB [59] were injected at a low dose ( CFU<3000 ) in the notochord . CFU counts were performed as previously described [21] . For quantification of bacterial load by Fluorescent Pixel Counts ( FPC ) , fluorescent bacteria were injected in the larvae and imaged using MVX10 Olympus microscope . Fluorescence was quantified by computation using Fiji ( ImageJ software ) as following: 1/ Background was measured in images of PBS injected larvae and then was subtracted in the fluorescence images , 2/ “make binary” function was run , and 3/ “measure area” function was used to determine the number of fluorescent pixels of the image . To induce macrophage depletion , 10 nl of Lipo-Clodronate or Lipo-PBS ( clodronateliposomes . com ) were injected intravenously ( i . v . ) in larvae at 1 dpf . Macrophage-depleted larvae were selected for infection based on the reduction of red-labeled macrophages tg ( mpeg1:mCherry-F ) 24 h after the treatment . For neutrophil depletion , 3 nl of antisense translational morpholino csf3r 0 . 7 mM ( 5’GAAGCACAAGCGAGACGGATGCCAT3’ , Gene Tools ) was microinjected in the one-cell stage tg ( mpx:GFP ) embryos . Standard control from Gene Tools ( see Morpholino injection section ) was used as a control . Neutrophils or macrophages were alternatively depleted using metronidazole treatment of tg ( mpx:gal4/UAS:nfsb-mCherry ) larvae or tg ( mpeg1:gal4/UAS:nfsb-mCherry ) , respectively ( see below ) . Microinjection of 3 nl of 10 ng/μl of gcsf3a or gcsf3b over-expressing plasmids [60] at 1-cell stage was used to increase neutrophil supply in embryos . For neutrophil depletion , tg ( mpx:Gal4/UAS:nfsB-mCherry ) and tg ( mpx:Gal4 /UAS:nfsB-mCherry/mpeg1:GFPcaax ) embryos expressing a Nitroreductase-mCherry fusion protein specifically in neutrophils , were placed in fish water containing 5 or 10 mM Metronidazole/0 . 1% DMSO ( MTZ , Sigma-Aldrich ) ( freshly prepared ) , at 40 hpf . Treatment with 0 . 1% DMSO and not transgenic siblings treated with MTZ were used as controls . Higher neutrophil depletion was observed using 10 mM MTZ . Therefore , 10 mM concentration of MTZ was used for further analysis , excepted in Fig 3C where a representative larva with 50% neutrophil depletion using 5 mM MTZ is shown . For macrophage depletion , tg ( mpeg1:Gal4/UAS:nfsB-mCherry ) were treated with 10 mM Metronidazole/0 . 1% DMSO at 30 hpf . tg ( mpeg1:Gal4/UAS:nfsB-mCherry ) treated with 0 . 1% DMSO and not transgenic siblings treated with MTZ were used as controls . VAS2870 ( Sigma-Aldrich SML0273 ) stock was prepared in DMSO at 15 mM . Two dpf tg ( lyz:DsRed ) embryos were injected in the yolk with 5 nl of 20 μM VAS2870 diluted in miliQ water or with 5 nl of water-diluted DMSO . Apocynin ( Santa Cruz , CAS498-02-2 ) was dissolved at 100 mM in DMSO . E . coli-infected larvae were placed in fish water containing 250 μM Apocynin for 1 day . Decrease of superoxide production was detected using DHE ( Dihydroethidium , Santa Cruz CAS104821-25-2 ) staining ( see below ) . Nitric Oxide inhibition was performed with the pan-NOS inhibitor NG-Nitro-L-Arginine Methyl Ester ( L-NAME ) ( Sigma-Aldrich , CAS 51298-62-5 ) . After notochord infection , embryos were placed immediately in 1 mM L-NAME fish water for the whole time course of the experiments . To knock down translation of P47phox , the antisense oligonucleotide morpholino ( 5’ CGGCGAGATGAAGTGTGTGAGCGAG 3’ ) , overlapping the AUG start codon [61] was used . 2 . 1 ng of P47phox or Control ( standard control from Gene Tools , 5' CCTCTTACCTCAGTTACAATTTATA 3' ) morpholinos were injected at 1-cell stage . Mpx activity and neutrophils were detected in tg ( mpx:GFP ) /mpx+/- and tg ( mpx:GFP ) /mpx-/- larvae at 1 day post E . coli injection ( dpi ) using Sudan black staining and anti-GFP antibody ( molecular probe A11122 , dilution 1/500 ) , respectively [21] . For superoxide detection within the cells , DHE was added to the fish medium at 3 μM at 1 dpi for one hour and larvae were washed 2 times before imaging using confocal microscopy ( excitation/emission 532/605 nm ) [32] . To detect nitric oxide , infected tg ( lyz:DsRed ) embryos were stained with 4-Amino-5-methylamino-2’ , 7’-difluorofluorescein diacetate , Diaminofluorescein-FM diacetate ( DAF-FM-DA ) ( Sigma , CAS 254109-22-3 ) [31] at 5 μM in fish medium for 2 hours at 6 , 10 hpi and 1 dpi ( for E . coli infection ) or 2 hpi ( for Salmonella infection ) . Larvae were rinsed three times in fish water before imaging using epi-fluorescence and confocal microscopy ( excitation/emission: 488/515 nm ) . Dead cells were detected using Sytox Green staining . Larvae were injected with 3 nL of 50 μM Sytox Green ( Molecular Probes ) in the vein at 1 dpi and placed at 28 . 5°C . One hour after Sytox Green injection , larvae were mounted in 1% low-melting-point agarose and imaged using epi-fluorescence and spinning disk confocal microscopy ( excitation/emission: 488/526 nm ) . Tricaine-anesthetized reporter larvae were imaged using MVX10 Olympus microscope . In Figs 2 , S3 and S6 total numbers of fluorescent neutrophils or macrophages were quantified as Leukocyte Units ( LUs ) by computation using Fiji ( ImageJ software ) as described in [62] . In Figs 1 , 3 , 5 , 6 , 7 , 8 , S2 and S9 the total number of fluorescent leukocytes were quantified by computation using Fiji ( ImageJ software ) as following: 1/ leukocytes were detected using “Find Maxima” function , 2/ Maxima were automatically counted using run ( "ROI Manager …" ) , roiManager ( "Add" ) and 3/ roiManager ( "Measure" ) functions . For quantification of recruited fluorescent neutrophils , tricaine-anesthetized reporter larvae were imaged using MVX10 Olympus microscope or confocal microscope . Neutrophils were directly quantified on the images , in a defined region of interest ( the Notochord or muscle region as indicated in the figure diagrams ) . Dead cells were directly quantified on confocal images , in a defined region of interest . Graph Pad Prism 4 . 0 Software ( San Diego , CA , USA ) was used to construct graphs and analyze data in all figures , except Fig 6F , 6H and S9F , which were performed in Excel 2010 ( Microsoft ) . Specific statistical tests were used to evaluate the significance of differences between groups ( the test and p value are indicated in the figure legend ) . Outliers were determined using Grubbs' test ( Graph Pad Prism 4 . 0 Software ) . The sample size is indicated in the figure legend and the sample size estimation and the power of the statistical test were computed using GPower software . Samples were allocated into experimental groups by randomization . The number of independent experiments ( biological replicates ) is indicated in the figure legends when applicable . The survival rate of treated embryos was compared with that of the control embryos using the log-rank ( Mantel-Cox ) test . Larvae were anesthetized and mounted as previously described [21] . Epi-fluorescence microscopy was performed using a MVX10 Olympus microscope ( MVPLAPO 1X objective; XC50 camera ) . Confocal microscopy was performed using a confocal Leica SPE upright microscope ( 40x HCX APO L 0 . 80 W and 20x CHX APO L 0 . 5 W objectives ) and an ANDOR CSU-W1 confocal spinning disk on an inverted NIKON microscope ( Ti Eclipse ) with ANDOR Neo sCMOS camera ( 20x air/NA 0 . 75 objective ) . Image stacks for time-lapse movies were acquired at 23–26°C every 4 min , typically spanning 50 μm at 2 μm intervals , at 1024x512 or 512x512 pixel resolution . The 4D files generated from time-lapse acquisitions were processed using Image J , compressed into maximum intensity projections and cropped . Brightness , contrast , and colour levels were adjusted for maximal visibility . For gcsf over-expression , larvae were injected with gcsf3a or gcsf3b over-expressing plasmids or no plasmid as described above . At 2 dpf , larvae were either uninfected or infected with E . coli in the notochord . To determine the relative expression of gcsf3a , gcsf3b and lyz , total RNA from infected larvae and controls ( pools of 6 larvae each ) was prepared at 1–2 dpi . For mpeg1 mRNA expression analysis , total RNA was extracted from 3 dpf Lipo-PBS and Lipo-clodronate treated larvae ( 10 larvae per pool , 3 pools per conditions ) . RNA preparation , reverse transcription and Q-PCR were performed as described in [63] , using ef1a as a reference gene . Q-RT-PCR analyses were performed using LC480 software . The primers used were the following: zcsf3a . 32 ( 5’gac tgc tct tct gat gtc tg 3’ ) , zcsf3a . 52 ( 5’aac tac atc tga acc tcc tg 3’ ) , zcsf3b . 31 ( 5’ggc agg gct cca gca gct tc 3’ ) , zcsf3b . 51 ( 5’gga gct ctg cgc acc caa ca 3’ ) , LyzA ( 5’ccg tta cag taa gaa tcc cag g 3’ ) and lyzS ( 5’ aga att tgt gca aag tgg cc 3’ ) , zef1a . 5 ( 5’ ttc tgt tac ctg gca aag gg 3’ ) , zef1a . 3 ( 5’ ttc agt ttg tcc aac acc ca 3’ ) , mpeg1 . FW1 ( 5’ ttt cac ctg ctg atg ctc tg 3’ ) and mpeg1 . RV1 ( 5’ atg aca tgg gtg ccg taa tc 3’ ) .
|
Deciphering the defence mechanisms of leukocytes remains a challenge for public health . Although phagocytic killing has been extensively studied , little is known about how bacteria are eliminated extracellularly in live vertebrates . Herein we use the notochord infection model in the zebrafish embryo to describe how leukocytes eliminate distant bacteria that are inaccessible for phagocytosis . In this context neutrophils but not macrophages are instrumental for bacterial clearance and larva survival . We then found that neutrophil bactericidal action relies on the NADPH oxidase dependent production of superoxide and is independent of NO synthase or myeloperoxidase .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"blood",
"cells",
"fluorescence",
"imaging",
"medicine",
"and",
"health",
"sciences",
"immune",
"cells",
"immunology",
"light",
"microscopy",
"bacterial",
"diseases",
"developmental",
"biology",
"microscopy",
"embryos",
"neutrophils",
"research",
"and",
"analysis",
"methods",
"embryology",
"infectious",
"diseases",
"white",
"blood",
"cells",
"imaging",
"techniques",
"animal",
"cells",
"escherichia",
"coli",
"infections",
"fluorescence",
"microscopy",
"life",
"cycles",
"notochords",
"cell",
"biology",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"macrophages",
"larvae"
] |
2018
|
Neutrophils use superoxide to control bacterial infection at a distance
|
The Qinghai-Tibet plateau is a natural plague focus and is the largest such focus in China . In this area , while Marmota himalayana is the primary host , a total of 18 human plague outbreaks associated with Tibetan sheep ( 78 cases with 47 deaths ) have been reported on the Qinghai-Tibet plateau since 1956 . All of the index infectious cases had an exposure history of slaughtering or skinning diseased or dead Tibetan sheep . In this study , we sequenced and compared 38 strains of Yersinia pestis isolated from different hosts , including humans , Tibetan sheep , and M . himalayana . Phylogenetic relationships were reconstructed based on genome-wide single-nucleotide polymorphisms identified from our isolates and reference strains . The phylogenetic relationships illustrated in our study , together with the finding that the Tibetan sheep plague clearly lagged behind the M . himalayana plague , and a previous study that identified the Tibetan sheep as a plague reservoir with high susceptibility and moderate sensitivity , indicated that the human plague was transmitted from Tibetan sheep , while the Tibetan sheep plague originated from marmots . Tibetan sheep may encounter this infection by contact with dead rodents or through being bitten by fleas originating from M . himalayana during local epizootics .
Plague is an acute infectious disease caused by Yersinia pestis that killed millions of people in Europe in the 14th century and tens of thousands in China in the 19th century [1] . Plague is mainly a disease of wild rodents , and their parasitic fleas are considered the transmitting vectors . So far , four subspecies of Y . pestis have been recognized on the basis of their biochemical properties: Y . pestis antiqua , mediaevalis , orientalis , and pestoides ( microtus ) [2 , 3] . To date , at least 12 plague foci covering >1 . 4 million km2 have been identified in China [4]; the largest focus is the Marmota himalayana focus on the Qinghai-Tibet plateau in northwestern China . The overwhelming majority of Y . pestis pathogens on the Qinghai-Tibet plateau are biovar antiqua , with the exception of biovar microtus ( qinghaiensis ) in the Microtus fuscus focus , which is located in Chengduo county in Qinghai Province and in Shiqu county in Sichuan Province [4] . The Qinghai-Tibet plateau is the highest risk area for human plague in China and M . himalayana is the primary host in this area . The pathogen Y . pestis ( biovar antiqua ) in the Qinghai-Tibet plateau M . himalayana natural plague focus frequently causes pneumonic and septicemic plague with high mortality . Other rodents ( Allactaga sibirica , Mus musculus , Cricetulus migratorius , Microtus oeconomus , and Ochotona daurica ) , some wild animals ( foxes , lynxes , and badgers ) , and domestic animals ( sheep , cats , and dogs ) have been found to be infected by Y . pestis [5] . Human plague originating from Ovis aries ( Tibetan sheep ) was first reported in 1956 in Qinghai Province [5] , though no bacterial evidence was obtained at that time . Tibetan sheep account for ~1/3 of the total number of sheep in China [6] . And the distribution areas of Tibetan sheep plague broadly overlap with the habitat of marmots in the Qinghai-Tibet plateau M . himalayana plague focus [6 , 7] . In August 1975 , a patient suffered from plague after butchering a dead Tibetan sheep in Yushu Prefecture , Qinghai Province . The meat of the sheep was eaten by 10 people; two individuals suffered intestinal plague that then developed into pneumonic plague , and one died [5] . Three Y . pestis strains were isolated from the dead individual , Tibetan sheep , and Capra aegagrus hircus ( Tibetan goat ) . This incident was the first time that human plague associated with Tibetan sheep or Tibetan goats was confirmed with bacteriological evidence in China [5] . In this study , we report human plague cases associated with Tibetan sheep on the Qinghai-Tibet plateau since the 1950s . Meanwhile , to further determine the ecological function of Tibetan sheep in Y . pestis endemic epidemics , we performed a genome-wide single nucleotide polymorphism ( SNP ) analysis of Tibetan sheep-related plague events , including pathogens isolated from humans , Tibetan sheep , and marmots . The genome-wide SNP analysis confirmed that the human plague strains were transmitted from Tibetan sheep , while the Tibetan sheep plague strains originated from marmots .
This study was approved by the Ethics Committee of the Qinghai Institute for Endemic Disease Control and Prevention ( FLW2013-001 ) and the Institute for Communicable Disease Control and Prevention ( ACUC2013-002 ) . All animal plague surveillance procedures were performed in accordance with the National Regulations for the Administration of Affairs Concerning Experimental Animals approved by the State Council . All procedures were in accordance with the ethical standards of the National Research Committee . Y . pestis was isolated and identified by Gram staining , the reverse indirect hemagglutination assay , and the bacteriophage lysis test . All Y . pestis strains isolated from Tibetan sheep ( 15 ) or humans ( 7 ) associated with Tibetan sheep on the Qinghai-Tibet plateau were included ( S1 Fig and S2 Table ) . The 18 outbreaks of human infection were designated from A to R ( Fig 1 and S1 Table ) . The Tibetan sheep involved in human plague outbreaks based on epidemiological investigations were designated using the same alphabetic code ( see S1 Table ) . In addition , 14 Y . pestis strains isolated from M . himalayana were selected; whenever possible , they were from the same region as the Tibetan sheep and in the same year in order to match the isolates from Tibetan sheep plague and human plague . Furthermore , two Y . pestis strains isolated from patients infected by M . himalayana in Nangqian County ( 2004 ) were also included [7] . All the strains were collected from the Qinghai Institute for Endemic Disease Control and Prevention , Xining , China . In addition , we plotted the geographical distribution of human plague , Tibetan sheep plague , and the isolates involved on a satellite map sourced from the Institute of Geographical Sciences and Natural Resources Research , Chinese Academy of Sciences , and we have received permission to publish it under a CC BY license from the institute . A total of 38 Y . pestis strains isolated from Tibetan sheep or humans or M . himalayana were included in this study . Genomic DNA from each bacterium was extracted using the following method in a Biosafety Level 3 Laboratory of the Qinghai Institute for Endemic Disease Control and Prevention . Y . pestis strains were cultivated in Luria–Bertani broth at 28°C for 48 h , and the collected strains were suspended in 0 . 5 ml of TE buffer ( 10 . 0 mM Tris-HCl [pH 8] , 1 . 0 mM EDTA ) and incubated at 28°C for 20 min , Then 80 μl of 10% SDS was added to the preparation ( 10 μg in 1 ml PBS ) , and maintained at 65°C for 10 min . Next , 20 μl RNase ( 10 mg/ml ) was added , and the solution incubated at 37°C for 2 h . Following the addition of 10 μl of proteinase K , the preparation was incubated at 37°C for 2 h . The DNA was extracted twice with equal volumes of phenol and once with an equal volume of chloroform . The DNA was precipitated by adding two volumes of absolute ethanol . The precipitated DNA was washed with 70% ethanol and re-suspended in TE buffer ( pH 8 . 0 ) . The 38 isolates were sequenced using the Illumina HiSeq 2000 platform ( Illumina , San Diego , CA ) . Two paired-end libraries were constructed with average insertion lengths of 500 bp and 3 , 000 bp . The raw data were filtered by FastQC , and then the clean data were assembled into contigs using SPAdes v3 . 9 . 1 . Gene prediction was performed using Glimmer with the default parameters . The whole-genome raw SNPs were detected through pairwise comparisons of Y . pestis genomes to the reference genome of the Angola strain ( NC_010159 ) [8] using Bowtie 2 software [9] and MUMmer [10] with the default parameters . Twenty-one completed genomes or draft genomes obtained from the NCBI database were also included in the analysis [1 , 11–17] ( S2 Table ) . Then the SNPs were combined , and those of low quality ( read depth <5 ) and those located within 5 bp on the same chromosome were removed to avoid the effect of recombination . A phylogenetic tree of Y . pestis was established based on these SNPs with the Bayesian evolutionary method in BEAST software [18] using the 38 Y . pestis genomes from our study and the 21 genome sequences of Y . pestis from GenBank and rooted with Y . pseudotuberculosis ( IP32953 ) [1 , 13] . The sequencing data of the Y . pestis strains are available in GenBank under accession numbers SRP131404 , and the genome sequences of 38 Y . pestis strains sequenced in our study have been deposited in GenBank with accession Nos SRR6512812 to SRR6512849 .
According to the epidemic history of plague on the Qinghai-Tibet plateau and the annual national plague surveillance data in China , a total of 18 human outbreaks ( events , designated A to R ) associated with Tibetan sheep have occurred since 1956 ( S1 Table and S1 Fig ) . Among these events , a total of 78 human cases associated with Tibetan sheep ( cases of original infection and successive secondary generation ) and 47 deaths were reported , of which 70 cases and 42 deaths occurred in Qinghai . In addition , 8 human cases ( 5 deaths ) associated with Tibetan sheep occurred in Tibet . All index infectious cases had an exposure history of butchering or skinning diseased or dead Tibetan sheep . Massive deaths or larger numbers of infection cases mainly occurred in four events before 1975; for example in 1956 , the index case ( Tianjun county ) suffered pneumonic plague and died after skinning a dead Tibetan sheep , and this individual infected a total of 13 individuals of whom 11 died . Eating meat from infected sheep that is not fully cooked is another cause of human plague infection , such as those in 1961 ( Dulan county ) , 1963 ( Yushu county ) , and 1965 ( Zhaduo county ) , that caused 26 cases of infection due to eating the meat; only the index individual in each outbreak slaughtered or skinned a diseased or dead Tibetan sheep . Considering the months in which Tibetan sheep plague , M . himalayana plague , and human plague events have occurred on the plateau since 1956 , 14 of the 27 Tibetan sheep plague events occurred during October and November . In contrast , the peak occurrence of M . himalayana plague was during June and July and usually ended in October ( National Plague Surveillance data and reference [5] ) . The plague in Tibetan sheep clearly lagged that in M . himalayana ( Wilcoxon signed rank test , P <0 . 05 ) . In addition , 9 of the 18 human plague events in which the index case ( s ) originated from Tibetan sheep occurred during October and November , while the peak months of human plague originating from M . himalayana were during August and September [5] ( Fig 1 ) . From 1997 to 2016 , no human plague cases were caused by Tibetan sheep due to active prevention and intervention measures in Qinghai , even though Y . pestis was still isolated from local Tibetan sheep and Tibetan goats on the Qinghai-Tibet plateau . The genomic sequences of the 38 isolates of Y . pestis were assembled de novo , producing 52 contigs and 70 scaffolds on average . The number of genes per strain ranged from 2 , 623 to 2 , 990 . The phylogenetic tree of Y . pestis was established using all isolates in our study as well as 21 complete genomes or draft genome sequences from the NCBI GenBank database ( S2 Table ) . We identified 1663 high-quality SNPs compared with the reference genomes and 216 within our isolates , with 39–63 SNPs per genome . Among these SNPs , 149 were located in 143 genes , including 41 synonymous SNPs and 108 nonsynonymous SNPs , with 1–2 in each gene , whereas the remaining 67 SNPs were located in intergenic regions . The 108 nonsynonymous sites were distributed among 106 genes . The phylogenetic relationships we constructed ( Fig 2A ) were very similar to the genomic maximum parsimony tree reported previously [1] . The nomenclature of the lineages in the phylogenetic tree are according to the literature [1 , 19] . The pathogens associated with Tibetan sheep plague were clustered into the 1 . IN2 lineage in the phylogenetic tree . These strains were comparatively closer to Y . pestis Z176003 , which was isolated from M . himalayana in Naqu County , Tibet , in 1976 [11] . In addition , strains H21 and H22 ( human plague isolates originating from M . himalayana in Bangqian Village , Nangqian County in 2004 ) were clustered in 2 . ANT1 . Y . pestis isolated from Tibetan sheep or local M . himalayana all fermented glycerin and reduced nitrate to nitrite , i . e . , they belonged to biovar antique , the same as human plague in this focus . Combining the epidemiological information ( S1 Table and S2 Table ) and the population structure based on the genome-wide SNP analysis , we divided the 36 Y . pestis in the 1 . IN2 lineage ( including those originating from Tibetan sheep ( 15 ) and humans ( 7 ) associated with Tibetan sheep , as well as 14 Y . pestis strains isolated from M . himalayana ) into four clusters ( I–IV ) , corresponding to eight clades ( 1–8 ) ( Fig 2B ) . Generally , the clade-based classification agreed well with the geographical area , i . e . , the strains isolated from the same area were found in the same clade ( Fig 2B ) . In fact , where no geographic barrier existed between adjacent areas , the pathogens isolated from adjacent areas also grouped together; for example , Juela Village in Nangqian County and Xialaxiu Village in Yushu are adjacent , and the strains isolated from the two villages grouped into Clade-1; Shanglaxiu , Batang , and Guoqing Villages are neighbors , and the lineages were grouped in Clade-4 . This shows that the genomic phylogenetic analysis of the Tibetan sheep-related strains have territory-specific characteristics . In addition , in Clade-1 and Clade-4 , the strains isolated in different years were grouped together . For example , the human plague cases and those corresponding to Tibetan sheep plague occurring in 1979 ( in Xialaxiu Village , Yushu County ) and in 1997 ( in Juela Village , Nangqian County ) were grouped together into Clade-1 . In 1975 , in Yushu County , the first human plague associated with Tibetan sheep was confirmed by bacteriological evidence . However , in 2005 in Yushu County , a larger-scale Tibetan sheep plague occurred , in which a total of 13 Tibetan sheep and 1 Tibetan goat in the same flock died . The isolates from these two events were grouped into Clade-4 . This indicated that the same strains of Y . pestis successively caused Tibetan sheep or human plague outbreaks in these areas . Of course , some isolates could not be grouped together by event although the strains were isolated in the same area , such as lineages M8 and S20 . In fact , finding any clear epidemiological connection between these two isolates and the rest was difficult . One possible explanation is the genomic diversity of the strains in these foci . In Clade-1 , the Y . pestis isolated from patients ( H19 ) and Tibetan sheep ( S30 and S24 ) in Juela Village , Nangqian County in 1997 , as well as isolates from M . himalayana ( M39 ) , were grouped together . According to the epidemiological information , the diseased herdsman ( H19 ) suffered pneumonic plague after processing a dead Tibetan sheep ( S24 ) , and isolate S30 was obtained from a sheep in the same breeding herd as the dead sheep ( S24 ) . In addition , one strain ( M39 ) from a dead M . himalayana in a sheep grazing area had been isolated four months earlier . In fact , a raging animal plague epidemic had occurred one year previously ( in 1996 ) in Juela Village , and a total of three strains ( M18 , M32 , and M38 ) were collected in the area ( National Plague Surveillance data ) . The above isolates were grouped together in Clade-1 . Two strains ( S31 and S23 ) isolated from Tibetan sheep in Xialaxiu Village , Yushu County ( adjacent to Nangqian County ) also grouped into Clade-1 . Furthermore , the strain ( H5 ) from the human plague in 1979 , the corresponding Tibetan sheep strains ( S6 and S7 ) , and some strains isolated from M . himalayana also clustered into Clade-1 . As noted above , in 2005 , three Y . pestis isolates ( S11 , G12 , and S13 ) were identified in two Tibetan sheep and one Tibetan goat from an outbreak of Tibetan sheep plague in Guoqing Village , Yushu . The Y . pestis isolated from the dead M . himalayana found in the same village and in the same year ( 2005 ) were clustered into the same clade ( Clade-4 ) . In fact , it was in Shanglaxiu Village , Yushu County , that the first human plague case associated with Tibetan sheep was confirmed in 1975 . In addition , three Y . pestis strains isolated from dead patients and Tibetan sheep and Tibetan goats in the same herd were also clustered in Clade-4 . Similar clustering of Tibetan sheep and M . himalayana was also found in Zongwulong Village , Delingha County , in 1996 ( Clade-5 ) . The above findings , together with the epidemiological connections , support the conclusion that human plague came from Tibetan sheep and Tibetan sheep plague originated from marmots .
The Qinghai M . himalayana natural plague focus was first identified in 1954 as a result of the isolation of Y . pestis from a dead marmot in Qinghai Province [20] . M . himalayana is the primary plague host in this area [5] . According to plague surveillance data in Qinhai Province , a total of 468 human plague cases with 240 deaths were reported , of which 162 cases originated from M . himalayana ( 34 . 62% ) , 39 from Tibetan sheep ( 8 . 33% ) , 16 from carnivorous animals ( 3 . 42% ) , and 216 from successive infection of pneumonic plague by person-to-person transmission ( 46 . 15% ) [5] . Tibetan sheep plague was sporadic on the Qinghai-Tibet plateau , and was restricted to areas that had M . himalayana plague epidemics . One previous investigation in Yushu Prefecture in 2005 found that the infection rate of Y . pestis in Tibetan sheep was 6 . 08% ( 64/1051 ) with serum titers in the range of 1:20 to 1:1280 [7] . Tibetan sheep-related human plague infection is always associated with slaughtering or skinning diseased or dead Tibetan sheep . Eating incompletely cooked meat from infected sheep or goats is another cause of human infection [5] . In previously research , the incidence of Tibetan sheep-related human plague outbreaks occurring in Qinghai Province between 1975–2009 were counted , and a total of 10 Tibetan sheep-related human plague outbreaks occurred during this period , resulting in 25 cases and 10 deaths , including bubonic plague ( 9 ) , primary pneumonic plague ( 6 ) , secondary pneumonic plague ( 6 ) , septicemic plague ( 3 ) , and intestinal plague ( 1 ) [2 , 3] . The even-toed ungulates ( Artiodactyla ) , including camels and goats [21–24] , donkeys , and cows [25] , can be naturally infected by Y . pestis . Previous studies have shown that the sheep is a plague reservoir with high susceptibility and moderate sensitivity [26 , 27] . And , under natural circumstances , only individual Tibetan sheep in a flock are infected , and they do not become infected directly by sheep-to-sheep contact , even when the same flock contains a mixture of sick and healthy sheep [26] . These findings indicate that the ecological function of the Tibetan sheep in associated human plague should be considered as an intermediate or accidental host . Another piece of supporting evidence is the fact that the occurrence of Tibetan sheep plague during the year lags behind the occurrence of M . himalayana plague . October and November were the high incidence months for the Tibetan sheep plague and human plague originated from Tibetan sheep . On the Qinghai-Tibet plateau , marmots begin hibernation from October to early November . One possible reason is that the fleas living in the caves escape after the marmots enter hibernation in October and attack other animals , such as Tibetan sheep . A minor peak for the human plague associated with Tibetan sheep occurs in June to July and presumably is caused by the massive death of marmots . Such an ecological change could also result in more fleas escaping from dead hosts and colonizing Tibetan sheep or human beings . Several possible scenarios may explain how Tibetan sheep become infected by marmots . First , they could be infected by contact with the bodies of dead marmots . Our field observations showed that Tibetan sheep have a habit of licking the bodies of dead rodents such as marmots , which may be a means of ingesting micronutrients in the plateau environment . Previously , a study successfully induced plague infection by feeding or smearing Y . pestis in the mouths of Tibetan sheep [27] . Another possible cause is that Tibetan sheep could be infected by fleas such as Callopsylla dolabris or Oropsylla silantiewi . These are the main parasitic fleas in M . himalayana . Even though they have comparatively specific host selection , they have been found to attack human beings or other animals after the death of their preferred host [6] . Previous research has shown that C . dolabris and O . silantiewi bite and can suck the blood of Tibetan sheep in the laboratory , and the sheep can become infected and die after being challenged for 10 days [27] . The above evidence shows that fleas play an important role in Y . pestis transmission from marmots to Tibetan sheep . Through genomic analysis , we confirmed that human plague came from Tibetan sheep , and Tibetan sheep plague originated from marmots . To the best of our knowledge , natural infection of sheep with Y . pestis is rare elsewhere in the world . The Tibetan sheep plague epizootic has some novel features , such as a complex transmission route , an extended epizootic period , and the possibility of transmission across long distances . Therefore , the hazards of Tibetan sheep plague should not be underestimated .
|
Plague is mainly a disease of wild rodents , and their parasitic fleas are considered the transmitting vectors . However , human plague originating from Ovis aries ( Tibetan sheep ) is found in the Qinghai-Tibet plateau in China , where Marmota . himalayana is the primary plague host . Tibetan sheep-related human plague infection is always associated with slaughtering or skinning diseased or dead Tibetan sheep . The plague in Tibetan sheep clearly lags that in M . himalayana . In this study , we performed a genome-wide single nucleotide polymorphism analysis of Tibetan sheep-related plague events , including pathogens isolated from humans , Tibetan sheep , and marmots . Through genomic analysis , together with the epidemiological connections , we confirmed that human plague came from Tibetan sheep , and the Tibetan sheep plague originated from marmots . Tibetan sheep account for about 1/3 of the total number of sheep in China . Tibetan sheep and goats are important domestic livestock on the Qinghai-Tibet plateau . Therefore , the hazards of Tibetan sheep plague should not be underestimated .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"methods",
"Results",
"Discussion"
] |
[
"invertebrates",
"livestock",
"medicine",
"and",
"health",
"sciences",
"plagues",
"pathology",
"and",
"laboratory",
"medicine",
"ruminants",
"pathogens",
"microbiology",
"vertebrates",
"genomic",
"library",
"construction",
"animals",
"mammals",
"ethnicities",
"bacterial",
"diseases",
"yersinia",
"fleas",
"dna",
"construction",
"molecular",
"biology",
"techniques",
"bacteria",
"bacterial",
"pathogens",
"tibetan",
"people",
"research",
"and",
"analysis",
"methods",
"infectious",
"diseases",
"sheep",
"yersinia",
"pestis",
"medical",
"microbiology",
"microbial",
"pathogens",
"molecular",
"biology",
"goats",
"insects",
"agriculture",
"arthropoda",
"people",
"and",
"places",
"eukaryota",
"dna",
"library",
"construction",
"biology",
"and",
"life",
"sciences",
"population",
"groupings",
"amniotes",
"organisms"
] |
2018
|
Human plague associated with Tibetan sheep originates in marmots
|
In Amazonian tropical forests , recent studies have reported increases in aboveground biomass and in primary productivity , as well as shifts in plant species composition favouring fast-growing species over slow-growing ones . This pervasive alteration of mature tropical forests was attributed to global environmental change , such as an increase in atmospheric CO2 concentration , nutrient deposition , temperature , drought frequency , and/or irradiance . We used standardized , repeated measurements of over 2 million trees in ten large ( 16–52 ha each ) forest plots on three continents to evaluate the generality of these findings across tropical forests . Aboveground biomass increased at seven of our ten plots , significantly so at four plots , and showed a large decrease at a single plot . Carbon accumulation pooled across sites was significant ( +0 . 24 MgC ha−1 y−1 , 95% confidence intervals [0 . 07 , 0 . 39] MgC ha−1 y−1 ) , but lower than reported previously for Amazonia . At three sites for which we had data for multiple census intervals , we found no concerted increase in biomass gain , in conflict with the increased productivity hypothesis . Over all ten plots , the fastest-growing quartile of species gained biomass ( +0 . 33 [0 . 09 , 0 . 55] % y−1 ) compared with the tree community as a whole ( +0 . 15 % y−1 ) ; however , this significant trend was due to a single plot . Biomass of slow-growing species increased significantly when calculated over all plots ( +0 . 21 [0 . 02 , 0 . 37] % y−1 ) , and in half of our plots when calculated individually . Our results do not support the hypothesis that fast-growing species are consistently increasing in dominance in tropical tree communities . Instead , they suggest that our plots may be simultaneously recovering from past disturbances and affected by changes in resource availability . More long-term studies are necessary to clarify the contribution of global change to the functioning of tropical forests .
Tropical rain forests play a major role in the global carbon cycle: they encompass over a third of terrestrial carbon stocks [1] , and they contribute approximately 30% of terrestrial net primary productivity [2] . Not only are many tropical forests under direct threat from land-use changes and logging [3–5] , but it has also been suggested that pristine , apparently undisturbed rainforests may also be undergoing widespread shifts in carbon stocks and floristic composition as a result of large-scale anthropogenic environmental changes . Models suggest that plants in general , and tropical forest plants in particular , are sensitive to environmental changes such as increased atmospheric CO2 concentration , nitrogen deposition , temperature , drought frequency , and irradiance [6–11] . Such sensitivity could have profound implications for the future of one of earth's most critical ecosystems [12] . Field studies have reported several patterns consistent with hypothesized responses to global change [13]: increases in aboveground biomass stocks [14 , 15] , in aboveground net primary productivity ( ANPP ) [16–18] , in tree turnover [19] , and in the dominance of fast-growing species [20 , 21] . These patterns of change in tropical forest and the mechanisms proposed to explain them have , however , been much debated [22–25] . An alternative explanation is that the observed changes in forest structure may be a response to natural disturbances alone . Under this second hypothesis , most forested areas in the tropics would be increasing in aboveground biomass because they were slowly recovering from past disturbances [26 , 27] . If this were true , this effect would be exactly offset by the large carbon losses in areas currently undergoing natural disturbances , and net ecosystem production would equal zero at the landscape scale . Neither increased ANPP nor an increase in dominance of fast-growing species would be expected under this hypothesis . Instead , the disturbance hypothesis predicts that slow-growing species would increase in dominance following a disturbance [28 , 29] . Here , we address these important predictions using a long-term dataset on tropical forest trees across a broad range of environmental conditions , examining both stand-level changes in biomass and changes in dominance for different guilds of tropical trees for up to 20 y of observations . Repeated forest inventories , including detailed taxonomic identification , combined with information on species traits , enable a direct evaluation of the relationship between changes in tree species composition and aboveground carbon stores of tropical forest ecosystems . In order to evaluate the long-term changes in the dynamics and composition of tropical forests , such inventories must encompass large samples of forest , including treefall gaps [21 , 30] . This can be accomplished through the use of large-scale plots [31] . We used datasets from ten large ( 16 to 52 ha ) undisturbed tropical forest dynamics plots in America ( n = 3 ) , Africa ( n = 2 ) , and Asia ( n = 5 ) [32] . The dataset included over 5 million stem-diameter measurements taken between 1985 and 2005 according to a standard census protocol ( see Methods , Table S1 , and Text S1A ) . Aboveground biomass was calculated for each free-standing stem ≥ 1 cm diameter at breast height ( dbh; i . e . , 130 cm from the ground for most trees ) , and wood density . We calculated a demographic index to reflect the relative position of each taxon on a slow-growth/low-mortality to fast-growth/high-mortality axis . For each site , species were ranked using this demographic index and were partitioned into four groups with an equal number of species and roughly the same biomass . Change in biomass was defined as the annual percent change ( % y−1 ) in aboveground biomass , and this was calculated in the whole plot and separately for each quartile group . For the top and the bottom quartiles , henceforth referred to as fast-growing and slow-growing groups , respectively , we assessed the statistical significance of biomass changes by bootstrapping over spatial heterogeneity .
We found that four of our plots increased significantly in aboveground biomass , three plots showed a nonsignificant trend of increasing biomass , and three showed a trend of decreasing biomass . A single plot , Sinharaja ( Sri Lanka ) , showed a large , but not statistically significant , decline in aboveground biomass ( −0 . 98 Mg ha−1 y−1; bootstrapped 95% confidence intervals: [−2 . 48 , 0 . 40] Mg ha−1 y−1 ) . This decline at Sinharaja was caused by the high mortality of a single shade-tolerant canopy species , Mesua nagassarium ( Clusiaceae ) , which dominates the topographic ridges in the plot [33]: in this species , 42% of the trees ≥ 70 cm dbh ( n = 86 ) and 22% of the trees ≥ 30 cm dbh ( n = 877 ) died between the two censuses . Averaging over all plots , we found a significant mean aboveground biomass increase of +0 . 47 [0 . 14 , 0 . 79] Mg ha−1 y−1 ( or about +0 . 24 MgC ha−1 y−1 ) . Excluding the Sinharaja plot , the other nine plots showed an average increase of ( +0 . 63 [0 . 30 , 0 . 96] Mg ha−1 y−1 ) . These patterns were the same when we restricted our analysis to trees ≥ 10 cm dbh , as has been done in most previous studies [14 , 15 , 22 , 34] ( Table S5 and Text S1B ) . In the three plots with two or more intercensus intervals , we found that aboveground biomass did not accumulate consistently over the study period . At BCI ( Panama ) , both significant increases and significant decreases in aboveground biomass were observed over the 20 y of study , consistent with the response of the forest to short-term disturbances , such as droughts ( Figure 1 ) . The two plots in Malaysia , Pasoh and Lambir , showed significant biomass increases between 1990 and 1995 , followed by decreases between 1995 and 2000 . This latter interval included a strong El Niño and regional droughts . Aboveground biomass growth rate did not consistently increase over the survey period , although it did marginally increase at BCI ( Figure 1 ) . Aboveground biomass mortality rate consistently increased at BCI , Pasoh , and Lambir across the study period . Next , we explored whether biomass changes in fast-growing and slow-growing species diverged from those observed for forest stands as a whole . The fast-growing species group ( species in the top quartile of the demographic index ) increased significantly in biomass at only one of our plots ( Table 1 ) , Sinharaja . The slow-growing group ( bottom quartile in demography ) increased significantly in biomass in five of our plots . Again , Sinharaja stood out: the slow-growing group declined dramatically and significantly ( −2 . 47 % y−1 ) , reflecting the aforementioned die-off of a dominant slow-growing species . Averaging across the plots , both the fastest-growing quartile of species ( 0 . 33 [0 . 09 , 0 . 55] % y−1 ) , and the slowest-growing quartile ( 0 . 21 [0 . 02 , 0 . 37] % y−1 ) increased significantly in biomass . Both groups increased more than the stand-level mean ( +0 . 15 % y−1 ) . As for the stand-level biomass trends , Sinharaja alone had a strong effect on the mean across plots . When this site was excluded , fast-growing species increased in biomass by only 0 . 17 [−0 . 08 , 0 . 40] % y−1 , not significantly different from the stand mean , whereas slow-growing species increased significantly by 0 . 50 [0 . 32 , 0 . 65] % y−1 . This result implies that at all sites except Sinharaja , slow-growing species increased at the expense of species growing at an intermediate rate rather than at the expense of fast-growing species . Exploring the temporal trend at the long-term plots , we found that the change in slow-growing species was consistently above the stand-level mean at Lambir and Pasoh ( Figure 2 ) . A different pattern was observed at BCI , however , where fast-growing species increased consistently more than the stand as a whole between 1985 and 2000 , and then declined at the expense of slow-growing species between 2000 and 2005 . We next explored trends by groups based on functional traits rather than on demographic rates ( Table 2 ) . Immediately after disturbance , species increasing in abundance are generally hypothesized to have a low wood density and small seed size . As ecological succession proceeds , these species would decline at the expense of species with , on average , high wood density and large seed size [28 , 29] . Our results were not consistent with either scenario . Species with high wood density tended to increase in biomass at all sites except Sinharaja . Here again , this plot alone led to an overall trend for a decrease in the high wood density group ( −0 . 12 [−0 . 30 , 0 . 04] % y−1 ) . At the other nine plots ( excluding Sinharaja ) , this group increased significantly in biomass ( +0 . 27 [0 . 12 , 0 . 41] % y−1 ) . In contrast , species with low wood density showed a nonsignificant change in biomass , both for all sites combined and when Sinharaja was excluded ( +0 . 04 or −0 . 06 % y−1 , respectively ) . Small-seeded species increased significantly in biomass ( +0 . 37 [0 . 19 , 0 . 54] % y−1 ) , across all plots , and this change remained significant when Sinharaja was excluded ( +0 . 31 [0 . 13 , 0 . 48] % y−1 ) . Overall , large-seeded species did not increase ( +0 . 01 [−0 . 17 , 0 . 16] % y−1 ) , but they did increase significantly when Sinharaja was excluded ( +0 . 23 [0 . 07 , 0 . 37] % y−1 ) . A previous study of a tropical forest near Manaus found that all the genera that increased in basal area were canopy species , whereas all declining genera were confined to the forest understory [20] . When we assigned functional groups based on maximal plant size , we found no significant changes in their dominance . To assess the hypothesis that disturbances lead first to decreasing biomass and increasing abundances of fast-growing species , and then to increasing biomass and increasing abundances of slow-growing species as succession occurs , we performed the same analyses with plots that experienced significant disturbances shortly before their first censuses , Luquillo ( Puerto Rico ) and Mudumalai ( India; Text S1A ) . Luquillo experienced farming , selective logging , and finally , a major hurricane ( Hugo ) immediately before its first census [35 , 36] . Mudumalai underwent selective logging years prior to plot establishment [37] . The Luquillo plot decreased significantly in biomass ( −1 . 43 [−2 . 03 , −1 . 06] % y−1 ) as trees died back , damaged by Hurricane Hugo , and it also exhibited a significant increase in the abundance of fast-growing species; both patterns are consistent with predicted initial responses to disturbance . The Mudumalai plot increased significantly in biomass throughout the study period ( +0 . 72 [0 . 55 , 0 . 86] % y−1 ) , and there was a consistent decline in the abundance of fast-growing species , consistent with longer-term succession following disturbance .
Our results from old-growth forest plots are consistent with an overall increase in aboveground biomass in tropical forests . Such an increase was previously observed in a large number of plots ( n = 59 ) , totalling 78 ha in size in the Amazon [15] . Here , we find the same pattern in fewer larger plots totalling 400 ha in size and spread over a broader geographical area and diversity of forest types , including a monodominant forest stand ( Lenda , Democratic Republic of Congo ) . However , the significant mean aboveground biomass increase of +0 . 47 Mg ha−1 y−1 dry mass ( or +0 . 24 MgC ha−1 y−1 , assuming that 50% of dry biomass is carbon ) in our plots was half as large as the +0 . 98 Mg ha−1 y−1 dry mass previously reported for Amazonian forests [15] . The mechanism underlying the observed increase in tropical forest biomass is still unclear . The inconsistencies of biomass growth rate over time in long-term sites do not argue strongly for a widespread increase in primary productivity in tropical forests . The increase in rate of biomass loss at these sites would instead suggest that tropical trees are growing in an increasingly unfavourable environment . There has been some controversy in the literature about the relative merits of long-term monitoring of tropical forests based on many small plots versus a few large plots [38] . Although our study suggests that the two approaches yield similar results , these approaches should be seen as complementary , rather than competing . Our large , permanent plots are big enough to subsume the fine-scale variation created by treefall gap formation , and by site selection bias . However , they may not always appropriately sample the landscape-scale variability of the forest [22] . In contrast , existing networks of small plots cover a larger range of environmental conditions , but they currently gather sites created for other purposes than environmental monitoring . A large amount of effort has been devoted to test the possible bias related to such heterogeneous datasets [16 , 19 , 38] , but complementing these tests with an independent network of large plots is important to move the debate forward . Working with large plots also is advantageous because in species-rich tropical forests , it is far easier to develop intensive botanical programs at a few sites than across networks of small scattered plots , despite recent progress in documenting spatial patterns of floristic tree diversity in the tropics [22 , 39] . Of our ten undisturbed plots , nine followed a dynamic consistent with the hypothesis that tropical forests are recovering from a past disturbance , with a significant increase in aboveground biomass , and a faster increase in dominance of slow-growing species relative to fast-growing species . The only exception was the Sinharaja ( Sri Lanka ) plot , in which an abundant canopy species , Mesua nagassarium , experienced a massive die-off during the study period . The cause of this decline is as of yet not known , although the presence of fruiting bodies of a particular fungus on the dead trees suggests a role for a pathogen . Although it has seldom been reported in large canopy trees , the massive decline of a single locally abundant species is consistent with theories of density-dependent regulation in tropical forests plants [40] . If this pattern is general across tropical forests , this would explain why many tropical forests plots are locally increasing in biomass , despite the fact that signs of large-scale past disturbances are difficult to detect . Our results fail to support the hypothesis that fast-growing and canopy species are increasing in dominance across tropical forests [20] . We found evidence for an increase in the biomass of fast-growing species at a single site , while slow-growing species increased significantly in dominance at half of our sites . Although alternative scenarios cannot be ruled out [13] , one plausible explanation is that our plots are indeed recovering from undocumented past disturbances . Successional changes in community composition are slower than changes in stand structure [41] , and past meso-scale disturbances are difficult to detect in temperate and tropical forests alike [24 , 41 , 42] . For instance , careful scrutiny of the history of temperate landscapes has revealed the complexity of the interplay between natural and human disturbances [43–45] , and this leads to serious uncertainties about the contribution of these environments to the global carbon cycle [46] . Even if this explanation is correct , recovery from disturbance alone is unlikely to be the only explanation for our observations of the increase in biomass and compositional shifts . It is likely that some physiological mechanism that is responding to the changing environment may also contribute . Wide-spread and long-term floristic monitoring programs in the tropics , in combination with better large-scale efforts to assess the stand structure of forests within tropical landscapes [47] , are thus crucial to understanding the past , present , and future of species composition and carbon stores in tropical forests .
In each plot , all trees ≥ 1 cm in dbh were mapped , tagged , identified botanically , and had their diameters measured to the nearest millimetre in each census . There is no evidence that any of the ten main plots have been disturbed by past human activities Over 80% of the taxa , encompassing 94% of total aboveground biomass , were reliably identified to the species level . We assumed that trees that increased in diameter by more than 45 mm/y or shrank more than −5 mm/y were inaccurately measured in the field [31] . For these individuals , we corrected the diameter by assuming a mean growth rate for the individuals in the same diametric class ( in millimetres , diametric size class limits were set to 10 , 15 , 20 , 25 , 30 , 40 , 50 , 100 , 200 , 300 , 400 , 500 , 600 , 700 , and 10 , 000 ) . The same correction was applied to recruits of anomalously large diameters . The default point of measurement was at 130 cm above ground following standard forestry techniques [31] . Measurements made at different heights due to an irregularity of the bole were marked with paint . Changes in the point of measurement were recorded in the database , and they were ignored in the computation of the average dbh growth rate . However , ignoring these stems in stand-level biomass estimation would have resulted in serious underestimates , since the dbh of many of the large trees had to be measured at different heights . We then filtered the dataset as in the case of inaccurate measurements described above . Simple data corrections were performed using a computer routine , but most of these corrections were carried out manually . The dbh of all the large trees in our plots ( dbh ≥ 70 cm dbh , n = 3 , 811 ) were manually checked . For the plots with more than two censuses , we were able to correct the anomalous dbh values more precisely , by comparing the stem dbh growth rates across census intervals . If a tree showed a dramatic change in dbh growth rate , we changed the one outside of the range ( −5 mm/y , +45 mm/y ) with the likely value , and updated the dbh value accordingly . This filter was applied using a computer routine , and then checked manually . In a recent work , aboveground biomass results were reported for the HKK and Pasoh sites that differ only slightly with the present figures [48] . These differences are explained by slight differences in the dataset corrections used in reference [48] and the present work . In the present work , all corrections in the raw data were performed by the lead author . All analyses were performed using the R project software , version 2 . 5 . 1 ( http://www . R-project . org/ ) . Aboveground biomass was calculated using a regression model that converts diameter and wood density into an estimate of total oven-dry aboveground biomass [49] . We evaluated the contribution to the aboveground carbon cycle of trees ≥ 1 cm in dbh , excluding seedlings , lianas , rattans , non-woody monocots , and rapid aboveground carbon pools ( coarse woody debris , twigs , leaves , and reproductive organs ) . Each plot was classified into one of the following three tropical forest types: dry , moist , and wet [50 , 51] . The majority of the tropical forested area consists of moist forests [52] . We used the following allometric regression models for individual trees to convert the inventory data into aboveground biomass [49]: Dry forest stands: Moist forest stands: Wet forest stands: where AGB is in Mg , D ( in cm ) is the trunk diameter at breast height ( 130 cm above the ground , or 50 cm above any buttresses or deformities ) , and r is the corresponding wood specific gravity ( oven-dry weight at 0% moisture over green volume , in g/cm3 ) . In the case of multiple-stemmed trees , the allometric model was applied to each stem and summed , to provide a tree-level aboveground biomass estimate . Palm species were often poorly estimated using the above method . We excluded climbing palms ( rattans ) from our analysis ( abundant in all the wet Asian plots ) . In addition , at Pasoh and at Lambir , arborescent palms were excluded from the sampling protocol ( they constitute a very small fraction of the total biomass in these forests , with the exception of Licuala in Lambir ) . In palms that showed diameter increases throughout ontogeny ( genera Socratea , Iriartea , Oenocarpus , and Attalea ) , aboveground biomass was relatively well estimated . For the sake of consistency , we also used this model for understory palms ( e . g . , genera Geonoma , Bactris , and Prestoea ) , acknowledging that gain in biomass was probably largely underestimated in these genera . This is not a serious issue in most of the plots , except in the Luquillo plot , where one palm species ( Prestoea acuminata ) constitutes a large fraction of the estimated biomass ( ca . 10% of the total ) . Existing allometric models for this palm [53] are based on trunk height , data that are currently unavailable for the palms in the plot . Tree fern aboveground biomass was also poorly estimated . For these reasons , we did not consider palms and tree ferns in interspecific comparisons . Note that the decrease reported in the Luquillo plot ( see below ) might have been exaggerated by our inaccurate estimate of the biomass dynamics in Prestoea acuminata . The missed gain might be on the order of 0 . 5–1 . 0 t/ha/y , and is insufficient to balance the observed loss . Statistical tests within a plot were based on the computation of annual aboveground biomass change ( in Mg ha−1 y−1 ) for each 0 . 25-ha subplot . Bootstrap samples of these quarter-hectare subplots were drawn 1 , 000 times to generate estimates of 95% confidence intervals [30] . The quadrat bootstraps were also applied to estimates of biomass change in species and species groups . Mean net changes across groups were computed by assuming the independence of the plots , and the normality of errors , described by the mean confidence interval 〈CI〉 . If n samples are available , the estimated confidence interval on the mean is 〈CI〉 / . Demographic species groups were defined from the demographic parameters of species with at least 20 saplings ( stems <5 cm and ≥1 cm dbh ) . Individuals of these species represented 70 . 5% to 95 . 2% of the total standing biomass ( Table S3 ) . Log-transformed sapling relative growth rate ( lsRGR ) and log-transformed sapling mortality rate ( lsM ) were positively and significantly correlated across species ( R2 = 0 . 208; Figure S1 ) . Our demographic index is defined as the first principal component analysis score between lsRGR and lsM . Species were divided into quartiles based on this index; these quartiles varied in biomass because species varied in abundance ( Table S3 ) . Because some individual trees were excluded from the analyses because they were not identified to species or belonged to species that had fewer than 20 saplings , net biomass changes summed over groups did not exactly match the total stand-level net biomass change . We assumed that unclassified stems were evenly spread across the groups . Under this assumption , the net biomass change summed over groups should be equal to the net biomass change of the plot . More precisely , if a group has a biomass stock of Bi ( in Mg ha−1 ) , a net biomass change of ΔBi ( in Mg ha−1 y−1 ) , and the entire stand has a biomass of B and a net biomass change of ΔB , the sum is generally smaller than B , because a number of species could not be classified . We then corrected ΔBi by the following formula: . With this correction , the sum of net biomass changes across groups is equal to ΔB . In addition to grouping species according to their demographic rates at the sapling stage , we also defined groupings using functional traits ( see summary statistics in Tables S2 and S4 ) . Wood specific gravity ( referred to as wood density in the main text , for simplicity ) is an important correlate of maximal growth rate and plant longevity , since species with dense wood must invest more in construction costs and are less vulnerable to stem breaks and microbial attack [54] . Wood density was estimated for the taxa based on surveys of the forestry literature [55 , 56] . If species-level information was unavailable , a genus-level or a family-level mean was taken . For 57 taxa , mostly in lesser-known taxonomic groups , no wood specific gravity information could be found , and a plot-level average was assumed . The contribution of these rare taxa to carbon pools and fluxes is negligible . Seed size is an important correlate of seed production and establishment strategies [57–59] . We matched the tree taxa in the plot to the Seed Information Database ( release 6 . 0 , Oct . 2004; http://www . kew . org/data/sid ) . Seed mass ( in grams ) was based on species-level information when available ( 22% ) , otherwise on information at the genus ( 68% ) or family ( 10% ) level . Seed mass was log-transformed prior to statistical analyses [58] . We also used potential tree size as a final predictor of demographic success . Free-standing woody plant species vary greatly in their life history strategies , especially in comparisons among those species that complete their entire life cycles in the understory and those that do not reproduce until they emerge above the canopy . Following the precedent of recent studies [60] , we estimated potential maximum tree height as 95 percentile in dbh . Understory species were defined as species whose upper 95 percentile of dbh ( maxdbh95 ) does not reach 10 cm ( for species with at least 20 individuals in a plot ) . Canopy species were species whose maxdbh95 ≥ 30 cm dbh .
|
Recent studies have reported major changes in mature tropical forests , with increases in both forest biomass and net primary productivity , as well as shifts in plant species composition that favour fast-growing species over slow-growing ones . These pervasive alterations were attributed to global environmental change , and may result in dramatic shifts in the functioning of tropical forest ecosystems . We reassessed these findings using a dataset of large permanent forest plots on three continents . We found that tree biomass increased at seven of our ten plots , and showed a large decrease at a single plot . Overall , this increase was significant , albeit lower than reported previously for Amazonian forests . At three sites for which we had data for multiple census intervals , we found no concerted increase in biomass gain , in conflict with the increased productivity hypothesis . With the exception of one plot , slow-growing species gained more biomass than either fast-growing species or the tree community as a whole . Hence , our results do not support the hypothesis that fast-growing species are consistently increasing in dominance in tropical tree communities . Overall , our results suggest that our plots may be simultaneously recovering from past disturbances and affected by changes in resource availability .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"ecology"
] |
2008
|
Assessing Evidence for a Pervasive Alteration in Tropical Tree Communities
|
Persistent production of type I interferon ( IFN ) by activated plasmacytoid dendritic cells ( pDC ) is a leading model to explain chronic immune activation in human immunodeficiency virus ( HIV ) infection but direct evidence for this is lacking . We used a dual antagonist of Toll-like receptor ( TLR ) 7 and TLR9 to selectively inhibit responses of pDC but not other mononuclear phagocytes to viral RNA prior to and for 8 weeks following pathogenic simian immunodeficiency virus ( SIV ) infection of rhesus macaques . We show that pDC are major but not exclusive producers of IFN-α that rapidly become unresponsive to virus stimulation following SIV infection , whereas myeloid DC gain the capacity to produce IFN-α , albeit at low levels . pDC mediate a marked but transient IFN-α response in lymph nodes during the acute phase that is blocked by administration of TLR7 and TLR9 antagonist without impacting pDC recruitment . TLR7 and TLR9 blockade did not impact virus load or the acute IFN-α response in plasma and had minimal effect on expression of IFN-stimulated genes in both blood and lymph node . TLR7 and TLR9 blockade did not prevent activation of memory CD4+ and CD8+ T cells in blood or lymph node but led to significant increases in proliferation of both subsets in blood following SIV infection . Our findings reveal that virus-mediated activation of pDC through TLR7 and TLR9 contributes to substantial but transient IFN-α production following pathogenic SIV infection . However , the data indicate that pDC activation and IFN-α production are unlikely to be major factors in driving immune activation in early infection . Based on these findings therapeutic strategies aimed at blocking pDC function and IFN-α production may not reduce HIV-associated immunopathology .
Chronic immune activation is a driving factor in CD4+ T cell loss and disease progression in HIV-infected individuals , yet the mechanisms responsible for this process are not completely understood [1] . Recent comparative studies in nonhuman primate models have shed light on the etiology of chronic immune activation [2] . Pathogenic simian immunodeficiency virus ( SIV ) infection in non-natural hosts including the Asian macaque species is characterized by sustained depletion of peripheral and mucosal CD4+ T cells , microbial translocation across the gut mucosa and persistently high levels of proinflammatory cytokines and lymphocyte activation that culminate in disease progression and AIDS [3]–[7] . In contrast , SIV infection of natural hosts such as the African green monkey and sooty mangabey results in preserved T cell homeostasis , low levels of chronic immune activation and a benign clinical course despite high levels of circulating virus [8]–[11] . A key distinction between the two models is that the innate immune response is rapidly resolved in SIV-infected natural hosts , whereas upregulation of the type I interferon ( IFN ) response and expression of IFN-stimulated genes ( ISG ) persists in SIV-infected macaques [12]–[17] . This dichotomy suggests that the innate immune response and persistent type I IFN production in particular may play a key role in chronic immune activation and disease progression [18] , [19] . Plasmacytoid dendritic cells ( pDC ) produce copious amounts of type I IFN in response to virus exposure but their role in HIV infection appears to be complex [20] . pDC are activated in HIV and SIV infection and are rapidly lost from blood , coincident with their recruitment to lymph nodes and mucosal tissues [21]–[27] , and within acutely infected lymph nodes IFN-α is produced largely by pDC [16] , [17] . In addition , pDC may be chronically stimulated in HIV infection and be a continuing source of IFN-α that leads to CD4 T cell death [28]–[32] . These findings have led to a model in which activated pDC that are recruited to lymphoid tissues chronically produce IFN-α that drives sustained expression of ISG and mediates T cell dysfunction and loss [18] , [33] . However , to date a direct link between the pDC response and chronic immune activation has not been made as reagents that selectively deplete pDC or interfere with their function in nonhuman primates have not been available . Resolving this issue has important clinical implications as therapeutic strategies aimed at disrupting pDC function are being considered as a means of controlling HIV-associated immunopathology [34] , [35] . pDC are activated by HIV and SIV nucleic acids to produce IFN-α and TNF-α through engagement of endosomal receptors TLR7 and TLR9 [36] , [37] , and dual antagonists comprised of nonstimulatory DNA sequences have been developed that block pDC stimulation through these receptors [38] . These compounds have been used to demonstrate the critical contribution of pDC to systemic lupus erythematosis and skin autoimmune disease in mouse models [39] , [40] . In the present study , we used a dual TLR7 and TLR9 antagonist to selectively block the response of pDC but not other mononuclear phagocytes to viral RNA in SIV-infected rhesus macaques and directly determine the role of pDC in immune activation .
To determine if TLR7 and TLR9 blockade was effective at selectively inhibiting pDC responses to SIV in rhesus macaques , we first tested the specificity and efficacy of a TLR7 and TLR9 antagonist in vitro . Peripheral blood mononuclear cells ( PBMC ) and peripheral lymph node cell suspensions from healthy SIV-naïve macaques were incubated with aldrithiol-2-inactivated SIVmac239 particles ( iSIV ) or influenza virus in the presence or absence of DV056 , a 25-based single-stranded phosphorothioate oligodeoxynucleotide antagonist of TLR7 and TLR9 that contains the necessary inhibitory motifs with optimization for activity in nonhuman primates and humans [37] . After incubation cytokine production was determined by flow cytometry using standard approaches to define mononuclear phagocytes ( Figure 1A ) [23] , [41] . Blood and lymph node pDC produced abundant IFN-α and TNF-α in response to iSIV and influenza virus , which was potently blocked by DV056 ( Figure 1B ) . pDC were the exclusive producers of IFN-α in SIV-naive blood in response to stimulation with iSIV , live SIV , influenza virus and the TLR9 agonist CpG-C , as shown by the lack of IFN-α secretion from stimulated PBMC that were depleted of pDC ( Figure 1C ) . Moreover , DV056 completely blocked secretion of IFN-α from unseparated PBMC in response to these varied stimuli , demonstrating the activity of the drug in antagonizing both TLR7 and TLR9 ( Figure 1C ) . In contrast , DV056 did not impair TNF-α production from blood and lymph node myeloid DC ( mDC ) and blood monocytes or TNF-α and IFN-α production from lymph node macrophages in response to stimulation with iSIV , consistent with the fact that these cells recognize viral RNA through receptors other than TLR7 and TLR9 ( Figure 1D ) [42] . We next determined the capacity of DV056 to block the response to viral RNA in rhesus macaques in vivo . We treated 4 SIV-naive macaques with DV056 at a dose of 2 mg/kg by weekly subcutaneous injections and collected PBMC 3 days after each injection for functional analyses . This drug regimen was based on preliminary studies showing that a similar dose and course of a related TLR7 and TLR9 antagonist in macaques blocked the ex vivo induction of ISG in PBMC in response to influenza virus stimulation ( F . J . B . , unpublished data ) . After a single treatment the proportion of blood pDC producing cytokines in response to virus stimulation dropped by 80 to 90% . In addition , the amount of each cytokine produced by pDC as judged by mean fluorescence intensity was substantially reduced in DV056-treated macaques ( Figure 2A , B ) . To assess the level of blockade in lymphoid tissues we harvested inguinal lymph nodes 3 days after the third dose of DV056 to allow time for accumulation of drug in tissues . The mean proportion of lymph node pDC producing IFN-α in DV056 treated animals dropped by 66% and 85% following stimulation with iSIV and influenza virus , respectively , relative to untreated lymph nodes , accompanied by marked reductions in the amount of cytokine produced . Reductions in lymph node pDC production of TNF-α ranged from 54% to 75% for iSIV and influenza virus , respectively ( Figure 2A , B ) . The responsiveness of mDC and monocytes/macrophages in each compartment was unaffected by DV056 treatment ( data not shown ) . After a total of 4 weekly doses of DV056 we infected the 4 rhesus macaques with SIVmac251 intravenously and continued weekly drug dosing up to day 53 post infection , encompassing the critical acute-to-chronic phase when immune activation is established . In parallel we infected 5 untreated macaques with the same dose of virus as controls . The kinetics of viremia were highly similar in both groups of animals , with nearly identical peak virus loads at day 11 and virus set-points at around day 70 post infection ( Figure 3A ) . Infection was associated with a marked increase and then decline in the number of blood pDC which then steadily returned to pre-infection levels by around day 30 post infection , and these kinetics were unaffected by DV056 treatment ( Figure 3B ) . Acute infection was also associated with a significant increase in the proportion of pDC expressing the chemokine receptor CCR7 regardless of DV056 treatment , consistent with the potential for blood pDC to migrate to lymph nodes ( Figure 3C ) . We next harvested PBMC from DV056-treated and control macaques and measured proinflammatory cytokine production after virus exposure ex vivo . Blood pDC from macaques that did not receive DV056 treatment had robust IFN-α and TNF-α production when stimulated with iSIV and influenza virus at the time of SIV infection . However , pDC from these animals rapidly lost responsiveness to virus stimulation ex vivo , and pDC hyporesponsiveness persisted into chronic infection , consistent with previous studies ( Figure 4A ) [43] . pDC from monkeys in the DV056-treated group had suppressed responses to virus stimulation at the time of SIVmac251 infection , revealing that TLR7 and TLR9 blockade remained effective even after multiple drug administrations , and this suppression was sustained after infection and persisted for the length of the study ( Figure 4A ) . In addition , expression of IRF-7 , a key IFN-α transcription factor that is induced in pDC upon HIV stimulation [32] , [44] , was markedly upregulated in iSIV-stimulated blood pDC from untreated macaques at day 0 but not at day 14 post infection , and was not upregulated in pDC from DV056-treated macaques at either time , consistent with both virus- and drug-induced inhibition of pDC responses ( Figure 4B ) . In contrast to pDC , blood mDC gained significant capacity to produce IFN-α in acute SIV infection when stimulated with iSIV ex vivo , although the intensity of cytokine production based on mean fluorescence intensity was relatively modest and there was considerable variation between animals ( Figure 5A , B ) . Notably , the enhancement of mDC function was not blocked by DV056 treatment , consistent with the inability of TLR7 and TLR9 antagonist to impact mDC responses to viral RNA ( Figure 5A , B ) . We next determined if DV056 treatment impacted pDC recruitment and cytokine production in lymph nodes following SIV infection . The frequency of CD123+ pDC within the Lineage– HLA-DR+ fraction of cells did not change at day 14 or 56 post infection in either DV056-treated or control animals when compared to preinfection samples ( Fig . 6A ) . However , the proportion of lymph node pDC that was BrdU+ increased significantly at day 14 post infection in both groups , reflecting the influx of recently divided pDC to lymph nodes [25] that occurred regardless of TLR7 and TLR9 blockade ( Fig . 6A ) . We next analyzed serial sections of lymph nodes for in situ expression of IFN-α and TNF-α , quantifying the frequency of cytokine-expressing cells using imaging software ( Figure S1 ) . In lymph nodes from untreated macaques at day 14 post infection IFN-α-expressing cells were frequent in lymph nodes , with an average of 0 . 5% of all cells in the paracortex and parafollicular cortex producing IFN-α . This expression was transient , as by day 56 post infection IFN-α-producing cells were rarely identified and their frequencies approximated those of naive lymph nodes ( Fig . 6B , C ) . In contrast , animals receiving DV056 treatment did not experience this transient increase in IFN-α production in acute infection; IFN-α-producing cells were rare in the paracortex and parafollicular cortex at both day 14 and 56 post infection with frequencies similar to that seen in naive macaques ( Fig . 6B , C ) . A majority but not all of the IFN-α expressing cells in untreated lymph nodes co-expressed CD123 , indicating that pDC were major but not exclusive producers of IFN-α in the absence of DV056 treatment . Additional double labeling experiments revealed IFN-α production by CD163+ macrophages and by mDC , identified using antibody to CD1a ( Fig . 6D ) [23] . IFN-α in lymph nodes from DV056-treated macaques did not co-stain with antibody to CD123 but was largely restricted to macrophages and mDC ( Fig . 6D and data not shown ) . In contrast to IFN-α , massive numbers of TNF-α-producing cells were present in lymph nodes at both day 14 and 56 post infection , regardless of TLR7 and TLR9 blockade , with 25 to 30% of all cells in the paracortex and parafollicular cortex expressing this cytokine ( Fig . 6B , C ) . TNF-α producing cells were primarily CD163+ macrophages and CD3+ T cells ( Fig . 6D ) . CD123+ pDC that co-labeled with antibody to TNF-α were rarely identified , even in the untreated animals ( data not shown ) . We next evaluated the impact that DV056 treatment had on the systemic type I IFN response following infection . In macaques that were SIV-infected without concurrent TLR7 and TLR9 blockade we detected a robust and transient increase in plasma levels of IFN-α which peaked at day 11 post infection . DV056-treated macaques had nearly identical plasma IFN-α responses ( Figure 7A ) . To determine the impact that TLR7 and TLR9 blockade had on ISG expression , we analyzed the relative expression of 6 genes ( IRF-7 , Mx-b , ISG-20 , 2 . 5 OAS , GBP-1 and ISG-54 ) that are induced in blood and lymph node CD4+ T cells in SIV-infected macaques [14] , [15] . In PBMC from untreated animals ISG responses paralleled those of plasma IFN-α , peaking at day 11 post infection and then dropping to near pre-infection levels ( Figure 7B ) . Induction of expression of all ISG examined was impacted minimally or not at all by DV056 treatment ( Figure 7B ) . In lymph node cell suspensions of untreated macaques , expression of ISG increased progressively from day 14 to day 56 , in contrast to blood ( Figure 7C ) . Treatment with DV056 blocked the induction of IRF-7 but otherwise did not diminish expression of ISG in lymph nodes subsequent to SIV infection ( Figure 7C ) . To determine if T cell activation and proliferation were impacted by TLR7 and TLR9 blockade , we analyzed CD4+ and CD8+ T cell subsets in blood and lymph node cell suspensions in DV056-treated and untreated macaques by flow cytometry ( Figure 8A ) . We focused on the memory subsets characterized by high expression of CD95 with and without expression of CD28 , as dysregulation is most prominent in these cells [45] . The number of CD4+ T cells in peripheral blood rapidly and transiently declined at day 11 post infection before rebounding and then steadily declining in untreated macaques . In macaques treated with DV056 blood CD4 T cell counts were transiently increased at day 4 and day 21 post infection , although the pattern of changes in CD4+ T cell counts over time was not significantly different between groups ( Figure 8B ) . In untreated macaques SIV infection resulted in the continual increase in frequency of memory CD4+ and memory CD8+ T cells in blood expressing Ki67 , an indicator of recent proliferation , and the activation marker CD38 [46] . The frequency of memory T cells expressing CD38 was not altered in DV056 treated macaques , while the frequency of both memory T cell subsets expressing Ki67 actually increased over time as a result of DV056 treatment ( Figure 8C ) . In inguinal lymph node cell suspensions the increase in proliferating memory T cell subsets following SIV infection , as determined by BrdU incorporation , was almost identical in control and DV056-treated macaques ( Figure 8D ) . The frequency of memory T cell subsets in lymph nodes expressing the activation marker CD38 approached 90% to 100% in both groups indicating profound activation regardless of DV056 treatment ( Figure 8D ) .
This study is the first to directly dissect the role of innate immunity in driving immune activation in pathogenic SIV infection of rhesus macaques , a model that produces AIDS-like disease very similar to HIV infection in humans but with an accelerated time frame [47] . We have demonstrated that stimulation of pDC by viral RNA through engagement of TLR7 and TLR9 induces a robust but transient IFN-α response in the lymph nodes of SIV-infected rhesus macaques , and provide evidence that this response in itself is insufficient to drive persistent ISG expression and immune activation that distinguishes pathogenic from nonpathogenic models [13]–[15] . While pDC were found to be major producers of IFN-α in response to SIV infection , they were not exclusive producers of this cytokine . Macrophages from naïve macaques produced IFN-α when stimulated in vitro with iSIV and were found to contribute to IFN-α production in lymph nodes during the acute phase of infection , although as shown here for pDC , macrophages and monocytes also become refractory to stimulation following SIV infection [43] . Conversely , while mDC from SIV naïve animals are largely incapable of producing IFN-α upon stimulation with SIV or virus-encoded oligonucleotides [43] , we found that mDC taken after infection gained a modest capacity to produce IFN-α upon virus stimulation and also contributed to in situ-IFN-α production in acutely infected lymph nodes . Similarly , in humans with chronic HIV infection , macrophages , mDC and lymphocytes produce IFN-α in the spleen and appear to be greater overall contributors to IFN-α production than pDC [48] . These findings reflect the fact that multiple cell types including hematopoietic and non-hematopoietic cells produce type I IFN through both TLR-dependent and independent pathways , depending on the type of stimulus [49] . Indeed , in murine models of virus infection ablation of pDC impacts IFN-α production only in the very early stages after infection , up to 36 hours; non-pDC are responsible for IFN-α at later time points [50] . The redundancy in IFN-α production may explain why blocking TLR7 and TLR9 function did not substantially diminish the peak plasma IFN-α response or alter the kinetics of ISG expression in infected monkeys . In addition , it is possible that the dramatically reduced levels of IFN-α in lymph nodes of DV056-treated animals are nevertheless sufficient to induce the high levels of ISG observed in these tissues , or that other viral IFN including IFN-β and IFN-λ that were not measured in this study contribute to a robust ISG response in the face of TLR7 and TLR9 blockade [51] . Notably , IFN-α production in lymph nodes was transient despite sustained high levels of virus and was temporally unrelated to chronic immune activation . In other studies , deliberately increasing levels of IFN-α in chronically infected sooty mangabeys through exogenous injection of an IFN-α agonist does not induce lymphocyte activation but reduces virus load [52] . In addition , in individuals with chronic HIV infection systemic IFN-α administration reduces virus load even as it enhances CD8+ T cell activation [53] , [54] . Collectively , these data do not support a direct pathologic role for IFN-α in disease progression in HIV infection . It is possible that chronic immune activation that is the hallmark of pathogenic SIV and HIV infection is more a consequence of sustained exposure to microbial products that have been translocated across the gut lumen , resulting in persistent activation of mononuclear phagocytes and other cells through the actions of lipopolysaccharide [7] , [55] . TNF-α overproduction by a subset of blood monocytes in response to microbial products has been demonstrated in HIV viremic individuals , providing a mechanistic link between microbial translocation and systemic inflammation [56] . In our study copious in situ TNF-α production was temporally associated with ISG expression in lymph nodes , and other cytokines including TGF-β , IL-1β and IL-15 that are increased in lymph nodes in acute SIV and HIV infection may also play a role either directly or indirectly in immune activation and disease progression [57]–[59] . We were unable to sample gut mucosal tissues for pDC activity and TLR7 and TLR9 blockade and therefore cannot rule out the possibility that pDC recruited to these tissues maintained continuous production of IFN-α that drove ISG expression and immune activation [26] , [27] . Phosphorothioate oligonucleotides have very predictable pharmacokinetics across species , with rapid clearance from plasma and accumulation primarily in lymphoid tissues , liver and kidney with a half-life of 2–6 days [60]–[63] , and studies with a related TLR7 and TLR9 antagonist to DV056 indicate a similar profile ( C . G . and F . J . B . , unpublished data ) . These compounds also accumulate in intestine at levels 3–6 times that of blood [61] , [64] . We would therefore predict that DV056 blocked pDC responses in mucosal tissues as it did in lymphoid tissues . Evidence also suggests that mucosal pDC become hyporesponsive to stimulation and IFN-α production subsequent to SIV infection [27] , as they do in lymphoid tissues [43] . Further studies are needed to directly investigate the response of pDC and other mononuclear phagocytes in mucosal tissues and its relationship to disease progression . TLR7 and TLR9 antagonist did not suppress the upregulation of CCR7 by pDC or their recruitment to lymph nodes in acutely infected macaques , despite the fact that HIV-mediated activation of TLR7 drives CCR7 upregulation in these cells [65] . This could be explained by the fact that the viral matrix protein alone induces functional CCR7 expression on pDC without inducing IFN-α [66] , and this pathway would be intact in our study animals as TLR7 and TLR9 blockade did not reduce the amount of circulating virus . These findings reinforce those in the mouse model of autoimmune skin disease [40] and highlight the fact that pDC recruitment to inflamed tissues is not linked to their capacity to secrete proinflammatory cytokines . An unexpected finding of our study was the significant increase in DV056-treated macaques of proliferating memory CD4+ and CD8+ T cells in blood . These increases were not associated with increased in virus load ( data not shown ) which is known to impact T cell proliferation [67] , [68] and may therefore be a consequence of TLR7 and TLR9 blockade itself . The mechanism for this enhanced proliferative response and its biologic relevance is unclear at present . TLR-mediated activation of DC is known to impact regulatory T cell development and influence CD4+ T cell proliferation [69] , [70] , but the impact of selectively blocking TLR7 and TLR9 on T cell proliferation in the context of SIV or HIV infection has not been studied and deserves further attention . Our findings support the conclusion that TLR7- and TLR9-mediated activation of pDC leads to significant and transient IFN-α production in lymph nodes but may not by itself drive the pathologic process of immune activation . The data reinforce the findings of others [16] , [17] showing that production of IFN-α from activated pDC that have been recruited to lymph nodes in early infection does not define pathogenic SIV infection , contrary to earlier findings [37] . Our findings suggest that specifically targeting pDC as a therapeutic strategy to reduce immunopathology in HIV infection may have limited benefit [34] , at least in the early stages of infection . It will be important to determine whether TLR7- and TLR9-driven activation of pDC plays any role in immune activation in the chronic stages of infection when disease is manifest .
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health . All animal procedures were performed according to a protocol approved by the Institutional Animal Care and Use Committee of the University of Pittsburgh ( protocol number: 1002392 ) . Appropriate sedatives , anesthetics and analgesics were used during handling and surgical manipulations to ensure minimal pain , suffering and distress to animals . Nine Indian-origin rhesus macaques ( Macaca mulatta ) were used in the study . Four macaques were treated with the dual TLR7 and TLR9 antagonist DV056 ( Dynavax Technologies ) , by weekly subcutaneous injections at 2 mg/kg , with 4 doses prior to and 8 doses after SIV infection . All macaques were infected by intravenous inoculation with 300 TCID50 SIVmac251 ( provided by Preston Marx , Tulane National Primate Research Center ) . Bromodeoxyuridine ( BrdU; Sigma ) was administered by intravenous injection at 30 mg/kg at 24-h intervals for 3 days prior to collection of peripheral lymph nodes at day −7 , 14 and 56 post infection , which were processed as previously described [23] . The quantitative real-time RT-PCR assay to determine SIV viral load was performed as previously described [25] . The following monoclonal antibodies were used to label PBMC and lymph node cell suspensions and were purchased from BD Biosciences unless otherwise noted: CD3 ( clone SP34-2 ) , CD20 ( eBiosciences , 2H7 ) , CD14 ( MøP9 ) , CD123 ( 7G3 ) , CD11c ( S-HCL-3 ) , HLA-DR ( L243 or G46-6 ) , CD163 ( GHI/61 ) , CCR7 ( 150503 , R&D Systems ) , CD4 ( L200 ) , CD8 ( RPA-T8 ) , CD28 ( CD28 . 2 ) , CD38 ( AT-1 , Stem cell Technologies ) , CD95 ( DX2 ) , IRF-7 ( H-246 , Santa Cruz Biotechnology ) and Ki67 ( B56 ) . Flow cytometric analysis and determination of blood pDC and CD4+ T cell counts were done as described [25] , [41] . For detection of intracellular cytokines , cells were first labeled with surface-binding antibody and then fixed and permeabilized prior to incubation with antibody to TNF-α ( MAb11 ) and IFN-α ( 225 . C ) . In vivo incorporation of BrdU was detected using a BrdU-FITC staining kit ( BD Biosciences ) . Dead cells were excluded using a Live/Dead viability kit ( Invitrogen ) . Cells were run on a BD LSR-II flow cytometer system , collected with BD FACS Diva 6 . 0 software , and analyzed with FlowJo 8 . 8 . 7 ( TreeStar ) . In general 1 million live cells were analyzed resulting in between 600 and 2 , 000 events of the respective mononuclear phagocyte subset after gating . PBMC and lymph node cell suspensions were stimulated at 2 . 5×106 cells/well for 7 h with iSIV at a capsid concentration of 200 ng/ml , virus-free microvesicles at 200 ng/ml , live SIVmac239 at a capsid concentration of 400 ng/ml ( SIV preparations provided by Jeffrey D . Lifson , AIDS and Cancer Virus Program , SAIC-Frederick ) , live H7N3 influenza virus ( provided by Ted M . Ross , University of Pittsburgh ) at a multiplicity of infection of 5 , or CpG-C oligodeoxyribonucleotide C274 ( Integrated DNA Technologies ) at a concentration of 5 ug/ml with and without prior incubation of cells with 1 µM DV056 for 1 h . Data generated using microvesicles as controls showed background cytokine production similar to unstimulated cells stained with isotype control antibody . In some experiments pDC were depleted from PBMC by labeling cells with PE-conjugated antibody to CD123 followed by anti-PE microbeads ( Miltenyi ) and then passing cells through a magnetic column ( Miltenyi ) . For flow cytometric analysis , cytokine secretion was blocked by addition of 5 µg/ml brefeldin-A ( BD Biosciences ) after 2 h . IFN-α in plasma and culture supernatants was measured using a commercial ELISA ( multi-subtype IFN-α ELISA kit , PBL Biomedical ) according to the manufacturer's instructions . Lymph nodes were prepared as described [23] , cut to 7 µm –thick sections and stained overnight with mAb to IFN-α ( MMHA2 , PBL Interferon Source ) , TNF-α or isotype-matched antibody ( BD Bioscience ) . Staining was developed using goat anti-mouse-HRP AlexaFluor 546 with tyramide signal amplification ( Invitrogen ) . Sections were co-stained with rabbit antibody to CD3 ( Dako ) or biotinylated mouse antibody to CD123 , CD163 , CD1a ( SK9 , BioLegend ) , or isotype controls followed by donkey anti-rabbit Alexa488 secondary antibody or mouse HRP-streptavidin AlexaFluor 488 . Nuclei were stained with Hoechst dye . Sections were visualized on an Olympus Fluoview 1000 confocal microscope and analyzed using Olympus FluoView Software . To quantify cytokine-expressing cells at least 10 non-overlapping regions were randomly imaged and cytokine-producing cells enumerated using image analysis software ( Figure S1 ) . Composite images were made by scanning entire lymph node sections using a Nikon 90i motorized epifluorescence microscope followed by digital reassembly using NIS-Elements . Gene expression analysis was done as previously described [40] . Briefly , total RNA was extracted from previously cryopreserved PBMC and lymph node cells using an RNA micro kit ( QIAGEN ) and cDNA was generated with SuperScript First-Strand Synthesis System ( Invitrogen ) . Quantification of ISG transcripts was performed by real-time RT-PCR in triplicate using TaqMan gene expression assays ( Applied Biosystems ) . RCT threshold cycle ( CT ) values for each gene were normalized to the housekeeping gene ubiquitin using the formula gene expression = 1 . 8 ( Avg CT Ubi – CT Gene ) ×100 , 000 , where Ubi is the mean CT of triplicate housekeeping gene runs , Gene is the mean CT of duplicate runs of the gene of interest , and 100 , 000 is an arbitrary factor to raise values above 1 . Primer sequences used were as follows: IRF7 , CTGTTTCCGCGTGCCCT ( forward ) , GCCACAGCCCAGGCCTT ( reverse ) ; Mx-b , GAGACATCGGACTGCAGAT ( forward ) , GTGGTGGCAATGTCCACGTTA ( reverse ) ; 2 . 5 OAS , AGGGAGCATGAAAACACATTTCA ( forward ) , TTGCTGGTAGTTTATGACTAATTCCAAG ( reverse ) ; GBP-1 , TGGAACGTGTGAAAGCTGAGTCT ( forward ) , CATCTGCTCATTCTTTCTTTGCA ( reverse ) ; and ISG-54 , CTGGACTGGCAATAGCAAGCT ( forward ) , AGAGGGTCAATGGCGTTCTG ( reverse ) . Primers and probes for ISG20 ( Hs00158122_m1 ) were provided by Applied Biosystems . Comparisons between treatment groups at a single time were performed by two- or three-way ANOVA . Profiles of measurements on animals assessed at multiple times were compared using mixed effects ANOVA , with a random effect for animal and fixed effects for time and other control variables . Where appropriate , tests were performed against baseline ( Day 0 or −7 ) within treatment . For analysis of T cell subsets trend was tested by treating time as a discrete variable [71] . Tests of profiles across treatments were required to be significant before comparisons between treatments were made at and single time . All statistical analyses were performed using SAS v9 . 3 ( SAS Institute ) .
|
A persistent type I interferon ( IFN ) response is thought to be important in driving immune activation and progression to AIDS in human immunodeficiency virus ( HIV ) -infected individuals . Plasmacytoid dendritic cells ( pDC ) produce copious amounts of type I IFN upon virus exposure through engagement of Toll-like receptor ( TLR ) 7 and TLR9 and thus may be central players in the etiology of immune activation . We used a dual antagonist of TLR7 and TLR9 to selectively block the response of pDC but not other mononuclear phagocytes prior to and for 8 weeks following simian immunodeficiency virus ( SIV ) infection of rhesus macaques . We show that pDC are major , but not exclusive , producers of IFN-α that mediate a marked but transient IFN-α response in lymph nodes in the acute phase of infection . TLR7 and TLR9 antagonist prevented this IFN-α production without suppressing pDC recruitment . Nevertheless , TLR7 and TLR9 blockade did not impact expression of IFN-stimulated genes or decrease the activation of T cells , the hallmarks of immune activation . The findings indicate that TLR7 and TLR9-driven activation of pDC is unlikely to be a major contributor to immune activation in the early stages of immunodeficiency virus infections and suggest that therapeutic strategies aimed at targeting pDC and IFN-α production may not reduce HIV-associated immunopathology .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"hiv",
"medicine",
"infectious",
"diseases",
"viral",
"diseases"
] |
2013
|
Blocking TLR7- and TLR9-mediated IFN-α Production by Plasmacytoid Dendritic Cells Does Not Diminish Immune Activation in Early SIV Infection
|
Optimal behavior relies on the combination of inputs from multiple senses through complex interactions within neocortical networks . The ontogeny of this multisensory interplay is still unknown . Here , we identify critical factors that control the development of visual-tactile processing by combining in vivo electrophysiology with anatomical/functional assessment of cortico-cortical communication and behavioral investigation of pigmented rats . We demonstrate that the transient reduction of unimodal ( tactile ) inputs during a short period of neonatal development prior to the first cross-modal experience affects feed-forward subcortico-cortical interactions by attenuating the cross-modal enhancement of evoked responses in the adult primary somatosensory cortex . Moreover , the neonatal manipulation alters cortico-cortical interactions by decreasing the cross-modal synchrony and directionality in line with the sparsification of direct projections between primary somatosensory and visual cortices . At the behavioral level , these functional and structural deficits resulted in lower cross-modal matching abilities . Thus , neonatal unimodal experience during defined developmental stages is necessary for setting up the neuronal networks of multisensory processing .
Most environmental events provide inputs to multiple senses that need to be integrated into a unified percept . Multisensory interactions have been observed both in higher-order brain regions and in primary sensory cortices [1–4] . Exposure to cross-modal stimuli has been shown to modulate the neuronal firing of cortical neurons and to shape the power and phase of oscillatory network activity [5–7] . Direct cortico-cortical connectivity [8] and feed-forward projections from thalamic nuclei [9] represent possible anatomical substrates of efficient multisensory processing at the cortical level . Furthermore , this multisensory processing in primary sensory areas supports the decoding of behaviorally relevant information and thereby improves task performance [10] . While the multisensory abilities are indispensable for daily life , their emergence is a progressive and protracted process that continues well after the development of individual senses [11 , 12] . For example , one of the earliest forms of multisensory integration in humans has been described at two months of age , when vowel information in faces and voices are matched [13] . In kittens , the first multisensory neurons were detected in the superior colliculus ( SC ) during the second postnatal week [14] . However , other multisensory processes do not reach adult level until adolescence [15 , 16] . The crucial requirements and processes controlling the development of multisensory interactions at the level of primary sensory cortices remain mechanistically unsolved . This knowledge gap is particularly striking in rodents , which have been shown to express highly multisensory abilities at the level of primary sensory cortices [8] . In contrast to multisensory maturation , the mechanisms underlying the unisensory development are better understood . The processing of unimodal stimuli is considered to rely on a highly precise wiring of the corresponding neuronal networks in the primary sensory cortices . The maturation of the neural circuitry is initiated under the control of molecular cues and subsequently refined by both experience-independent and experience-dependent electrical activity [17–20] . The experience-dependent refinement peaks during defined developmental stages , which have been termed critical or sensitive periods [21 , 22] . Unimodal inputs during these periods are crucial for shaping neuronal networks according to the environment and individual features . Hence , deprivation or reduction of unimodal input during this time results in abnormal patterns of neuronal activity in the corresponding primary sensory cortex and permanent behavioral impairment , whereas manipulation of the input before or after these periods is overall less detrimental [23 , 24] . While the development of multisensory processes in the SC neurons seems to depend on genuine cross-modal experience [25] , it remains unclear which type of inputs controls the development of multisensory processing within neocortical networks . Here , we aimed at elucidating the role of early unisensory experience for the development of multisensory interactions in the primary visual ( V1 ) and somatosensory ( S1 ) cortices as well as of multisensory object recognition in rats . We provide electrophysiological , anatomical , and behavioral evidence that neonatal unimodal inputs prior to the cross-modal experience are necessary for the correct maturation of multisensory processing by contributing to the setting up of structural and functional coupling within cortical networks .
Since previous studies showed that unimodal experience during early development is crucial for the refinement of corresponding cortical networks and topographic maps , we assessed the impact of transient whisker trimming on the anatomy and tactile processing in S1 . We firstly analyzed the structural changes in the barrel field of NWT rats post-mortem ( Fig 2A ) . Similar to previous findings [27] , the overall cytoarchitecture ( i . e . , number and position of barrels ) of the posteromedial barrel subfield corresponding to the major whiskers was not affected by deprivation . The size of individual cytochrome oxidase-stained ( COX ) barrels was slightly increased in the NWT rats ( n = 7 ) when compared to the corresponding barrels in CON rats ( n = 7 ) , but none of the differences reached significance level ( Fig 2B , S1B Table ) . Second , we characterized the effects of neonatal tactile restriction on unisensory evoked potentials ( EPs ) elicited by a mechanical deflection of the principal whiskers . Extracellular local field potentials ( LFPs ) were recorded simultaneously in S1 and V1 of lightly urethane-anesthetized NWT ( n = 9 ) and CON ( n = 10 ) rats using electrodes with 16 recording sites that covered the entire cortical depth . All recordings were conducted under urethane anesthesia to avoid the impact of varying alert states on multisensory processing [28] . Whisker deflection elicited in all cortical layers [supragranular ( S ) , granular ( G ) and infragranular ( I ) ] of the contralateral S1 of all investigated rats EPs with a depth profile characterized by layer specific polarity ( Fig 2C ) . EPs consisted of a first prominent peak with positive surface polarity ( P1 ) followed by additional negative and positive peaks ( N1 , P2 ) . The absolute amplitude of tactile EPs in G and I layers was larger in NWT rats when compared to CON rats ( Fig 2C and 2F ) . In contrast , the absolute amplitude of tactile EPs in the S layer was decreased in NWT compared to CON animals . Thus , the neonatal restriction caused specific changes of unimodal processing in the S1 . In line with our previous findings [8] , the S1 responded to visual stimulation via light flashes , yet the evoked responses were significantly smaller when compared to tactile EPs and had the same polarity over all cortical layers , suggesting that they are at least partially volume conducted ( e . g . , non-specifically electrically spread at a distance from their source generator ) . However , these low-amplitude visually evoked responses in the S1 differed between CON and NWT rats ( Fig 2D and 2G ) . Additionally , such plasticity was also induced for a non-deprived modality , since the EPs evoked by light flashes in the contralateral V1 differed between CON and NWT rats ( Fig 2E and 2H ) . The S1 seems to integrate simultaneously presented spatially congruent ( i . e . , in the same hemifield ) whisker deflections and light flashes as reflected by the supra-additive enhancement of the EPs . Their magnitude was larger than the arithmetic sum of responses evoked by unimodal tactile and visual stimulations [8] . To assess the effects of neonatal tactile restriction on these multisensory interactions , we compared the bimodal EPs in NWT and CON rats . In CON rats , a supra-additive enhancement of the P1 peak after congruent visual-tactile stimulation was detected in all cortical layers when compared to unimodal stimulation ( Fig 3A , S1C Table ) . Most prominently , in the G layer the absolute amplitude of multisensory P1 peak ( 581 . 8 ± 6 . 9 μV ) was larger when compared to the unisensory response ( 552 . 9 ± 7 . 1 μV , p = 0 . 003 ) and the arithmetic sum of EPs resulting from tactile and visual stimulation ( 536 . 9 ± 8 . 0 μV , p < 0 . 001 ) . Here , also the amplitude of the N1 peak was significantly augmented by bimodal stimulation ( 133 . 1 ± 2 . 5 μV versus unimodal: 98 . 6 ± 2 . 7 μV , p = 0 . 001 and arithmetic sum: 123 . 5 ± 3 . 8 μV , p = 0 . 016 ) , exceeding the enhancing effect of unimodal visual stimulation on the tactile response . In contrast , this multisensory EP enhancement was significantly reduced in NWT rats . The absolute P1 amplitude in the G layer was similar after uni- and cross-modal stimulation ( tactile , 621 . 6 ± 13 . 5 μV; unimodal arithmetic sum , 641 . 0 ± 14 . 8 μV; bimodal , 644 . 9 ± 13 . 4 μV; p = 0 . 73 , p = 0 . 95 ) . Moreover , the multisensory enhancement of the N1 peak ( 167 . 2 ± 4 . 4 μV ) reached significance level only when compared to tactile stimulation ( 124 . 4 ± 4 . 4 μV , p < 0 . 001 ) , but not when compared to the arithmetic sum of unimodal stimulations ( 166 . 0 ± 7 . 9 μV , p = 0 . 38 ) . However , we cannot fully exclude the possibility that this small multisensory effect in NWT rats partially resulted from the volume-conducted visual responses in S1 ( s . above ) . In any case , it can be concluded that neonatal whisker trimming diminished the multisensory enhancement of evoked activity in S1 . Besides affecting the evoked activity , cross-modal stimulation has been reported to influence the cortical networks by modulating the stimulus-induced oscillatory activity ( i . e . , related but not locked to the stimulus ) and the spontaneous ongoing oscillations ( i . e . , not causally related to the stimulus ) [8 , 29] . To get first insights into the ontogeny of these modulatory processes , we quantified the multisensory power change of induced oscillatory activity at different time points after the stimulus onset for both NWT ( n = 9 ) and CON rats ( n = 10 ) . Baseline-normalized wavelet spectra were used to calculate the power of oscillatory activity . For time windows of significant power change between tactile and bimodal stimulation ( 150–230 ms , 440–550 ms , and 850–900 ms post-stimulus ) the differences between the two conditions were calculated and averaged ( S1 Fig ) . As previously reported [8] , the strongest modulatory effect of congruent bimodal stimuli on the power of induced activity of CON rats has been observed for the G layer ( 150–230 ms: 1 . 20 ± 0 . 41 times higher unisensory versus multisensory , p = 0 . 002; 440–550 ms: 0 . 57 ± 0 . 15 times higher multisensory versus unisensory , p = 0 . 002; 850–900 ms: 0 . 47 ± 0 . 15 times higher multisensory versus unisensory , p = 0 . 015 ) . Power modulation after congruent bimodal stimulation was equally detected in NWT rats ( 150–230 ms: 2 . 36 ± 0 . 70 times higher unisensory versus multisensory , p = 0 . 002; 440–550 ms: 0 . 15 ± 0 . 43 times higher multisensory versus unisensory , p = 0 . 6; 850–900 ms: 0 . 11 ± 0 . 13 times higher multisensory versus unisensory , p = 0 . 43 ) ( S1C Table ) . However , its magnitude , especially at later time windows post-stimulus , was lower . These changes were most prominent in the G layer ( 440–550 ms: p = 0 . 006; 850–900 ms: p = 0 . 025 ) ( Fig 3B ) . The modulation of the power of oscillatory activity in CON and NWT rats was not accompanied by changes in population firing of S1 neurons . Analysis of the multi-unit activity ( MUA ) pre- and post-stimulus revealed the absence of multisensory effects in both groups of rats ( n = 10 CON rats , n = 9 NWT rats ) ( S2 Fig ) . Next , we assessed the effects of neonatal tactile restriction on the ongoing spontaneous activity ( i . e . , network activity non-related to the stimulus ) in the S1 . Cross-modal phase reset of network oscillations has been identified as a major cortical mechanism of multisensory interactions [5 , 6] . It has been hypothesized that timing of the neuronal activity by such a phase reset increases the processing efficiency of the stimulus [30] . In CON rats ( n = 9 ) , contralateral visual stimuli induced a prominent phase reset in all S1 layers that was detected as a concentration of a specific oscillatory phase in the histograms of phase resultant vector length ( S3 Fig ) . Among the different frequency bands of oscillatory activity , the strongest phase reset was confined to 8–12 Hz ( Fig 3C ) . For NWT rats ( n = 10 ) , the phase reset persisted , yet switched to a lower frequency ( 4–8 Hz ) ( Fig 3C ) . Thus , neonatal tactile restriction altered multisensory processing in the adult S1 by perturbing the power modulation and the phase reset of network oscillations . To identify the substrate of altered multisensory processing after neonatal tactile restriction , we investigated the structural and functional communication between V1 and S1 of CON and NWT rats . We firstly examined the direct visual-somatosensory intrahemispheric projections in NWT ( n = 5 ) and CON rats ( n = 7 ) by injecting small amounts of the retrograde tracer Fluorogold ( FG ) into their S1 ( Fig 4 ) . Bright fluorescent back-labeled parental cell bodies in V1 feed-forwardly projecting to the ipsilateral barrel field were detected both in NWT and CON rats , indicating that intrahemispheric visual-somatosensory projections emerge even in the absence of unimodal stimuli during neonatal development . However , the maximal density of retrogradely labeled neurons was significantly ( p = 0 . 023 ) lower ( 3–26 cells/0 . 16 mm2; mean: 8 . 6 ± 4 . 5 cells/0 . 16 mm2 ) in the V1 of NWT rats when compared to CON rats ( 11–56 cells/0 . 16 mm2; mean: 28 . 6 ± 6 . 4 cells/0 . 16 mm2 ) ( S1D Table , S4 Fig ) . Thus , unimodal restriction during neonatal development led to a “sparsification” of the direct connectivity between S1 and V1 . To test whether this reduced cortico-cortical connectivity after neonatal tactile restriction is accompanied by an impairment of the functional communication between primary sensory cortices , we assessed the impact of bimodal versus unimodal stimulation on the network synchrony between V1 and S1 of NWT ( n = 9 ) and CON rats ( n = 10 ) ( Fig 5 ) . The mean values of relative coherence were separately calculated for bimodal and unimodal stimulation conditions and their difference was averaged over 12–80 Hz ( S5 Fig ) . In CON rats , bimodal stimulation significantly increased the intrahemispheric coherence between all layers of S1 and V1 by 0 . 058 ± 0 . 033 , p = 0 . 008 ( S ) , 0 . 088 ± 0 . 034 , p = 0 . 017 ( G ) , and 0 . 067 ± 0 . 03 , p = 0 . 03 ( I ) when compared to unimodal stimulation ( Fig 5A ) . The interhemispheric S1-V1 synchrony after bimodal versus unimodal stimulation was similarly affected in NWT rats ( Fig 5B , S1E Table ) , even if their augmented coherence was below the significance threshold ( S: 0 . 043 ± 0 . 040 , p = 0 . 469; G: 0 . 074 ± 0 . 040 , p = 0 . 12; I: 0 . 073 ± 0 . 035 , p = 0 . 095 ) . The symmetric interdependence of coherence precluded reliable insights into the directionality of precisely timed interactions between S1 and V1 . To test for causal information flow between primary sensory cortices in adult rats with and without neonatal tactile restriction , we performed Granger analysis for the V1 and S1 activity and plotted the directionality changes after bimodal versus unimodal stimulation ( blue color in time-frequency plot: weaker directionality; red color in time-frequency plot: stronger directionality ) ( Figs 6 and S6 ) . In CON rats ( n = 10 ) , the bimodal stimulation modulated the directed interactions between the primary sensory cortices by decreasing the drive from V1 to S1 during the first 200 ms post-stimulus and increasing it afterwards ( Fig 6A–6C ) . In these rats , the drive from S1 to V1 was equally modulated by bimodal stimulation , yet the communication increased after stimulus , the strongest effect being detected within the time-window 0–200 ms post-stimulus ( Fig 6D–6F , S1F Table ) . This cross-modal modulation of directed interactions V1 → S1 and S1 → V1 was decreased in NWT rats . The observed changes were particularly strong 200 to 1 , 000 ms after the stimulus in S and G layers ( Fig 6B , 6C , 6E and 6F ) . These results indicate that transient tactile restriction during neonatal development decreased the functional communication between S1 and V1 in line with the sparsification of monosynaptic cortico-cortical projections . To assess the behavioral correlate of impaired multisensory processing at the neural level after transient neonatal tactile restriction , we tested CON ( n = 51 ) and NWT rats ( n = 51 ) in a modified cross-modal novel object recognition task ( NOR ) [31] . This task uses the intrinsic preference of rodents for novel objects [32] . Rats were allowed to explore two identical objects during the sample phase . Subsequently , one of these two identical objects was replaced by a novel object with different features during the choice phase . The paradigm was conducted under four sensory conditions ( S7 Fig ) . We allowed ( i ) simultaneous visual and tactile exploration ( bimodal condition ) , ( ii ) only tactile exploration ( tactile condition , i . e . , the experiment was performed under red light to prevent visual exploration ) or ( iii ) only visual exploration ( visual condition , i . e . , objects were placed under familiar glass containers to prevent tactile interactions ) . The fourth condition was used to test whether rats could match tactile with visual information ( cross-modal condition ) . For this , the sensory modality used for object exploration differed between the sample and the choice phase . Rats needed to transfer the cross-modal information between modalities to recognize the novel object . These four conditions were tested in two experimental settings . In the first setting , termed as continuous setting , each rat was consecutively assigned to the bimodal , one unimodal and the cross-modal conditions . In contrast , in the second setting , termed as discrete setting , each rat was assigned to only one condition ( S7 Fig , S2 Table ) . The familiarization phase was identical for both settings . Since in an open field pre-test the travelled distance , speed of locomotion , the duration and occurrence of rearing , wall-rearing , and grooming , as well as the time spent in the different areas of the open field was similar for CON and NWT rats , differences in locomotor and anxiety behavior between groups were excluded ( Fig 7 , S1G Table ) . In both settings , CON and NWT rats spent more time with the exploration of the new object compared to the time spent with the exploration of the familiar object when they relied on both senses ( continuous setting , CON: 65 . 5 ± 2 . 8% of the total time , p < 0 . 001; NWT: 61 . 8 ± 4 . 2% , p < 0 . 001; discrete setting , CON: 65 . 1 ± 5 . 7% of the total time , p = 0 . 002; NWT: 63 . 5 ± 4 . 8% , p = 0 . 001 ) ( Fig 8 ) . Similarly , the rats relying on tactile perception explored the novel object longer ( continuous setting , CON: 64 . 8 ± 2 . 8% of the total time , p < 0 . 001; NWT: 71 . 1 ± 7 . 9% , p < 0 . 001; discrete setting , CON: 65 . 7 ± 5 . 2% of the total time , p < 0 . 001; NWT: 67 . 1 ± 5 . 3% , p = 0 . 009 ) . In both groups of rats the visual recognition of the novel object was less precise than the tactile recognition and this was independent of the experimental settings ( continuous setting , CON: 62 . 7 ± 7 . 0% of the total time , p = 0 . 025; NWT: 55 . 6 ± 5 . 6% , p = 0 . 18; discrete setting , CON: 64 . 94 ± 7 . 0% of the total time , p = 0 . 009; NWT: 53 . 39 ± 3 . 84% , p = 0 . 241 ) . For bimodal , tactile and visual conditions , the discrimination ratio did not differ between groups and experimental settings . Under cross-modal conditions , CON rats spent significantly longer time exploring the novel object ( continuous setting , 59 . 7 ± 3 . 3% of the total time , p = 0 . 002; discrete setting , 57 . 4 ± 4 . 3% of the total time , p = 0 . 027 ) , whereas NWT rats failed recognizing the novel object ( continuous setting , 47 . 7 ± 3 . 5% , p = 0 . 37; discrete setting , 44 . 97 ± 5 . 03% of the total time , p = 0 . 179 ) . In only this matching test , the recognition performance of NWT rats was lower when compared to CON rats ( Figs 8 and S1H ) . These results indicate that transient tactile restriction during neonatal development permanently impairs cross-modal matching abilities , while leaving the unimodal object recognition largely unaffected .
Cross-modal modulation of neuronal assemblies in primary sensory cortices is an important mechanism of multisensory processing in adult rodents [8] and across much of the brain in more complex species . To which degree the development of this mechanism depends on uni- or cross-modal experience is still unresolved . The present study provides first insights into the essential requirements of the neural basis of multisensory development . We demonstrate that unimodal experience during neonatal development ( i ) refines the somatosensory topography and tactile processing in S1 ( Figs 1 and 2 ) , ( ii ) shapes visual-tactile interactions in S1 and along the subcortico-cortical sensory tract ( Fig 3 ) , ( iii ) ensures the correct maturation of direct cortico-cortical projections ( Fig 4 ) , ( iv ) enables multisensory communication by synchrony and modulation of directed interactions within adult cortico-cortical networks ( Figs 5 and 6 ) , and ( v ) is necessary for the development of cross-modal matching abilities ( Figs 7 and 8 ) . These findings point towards a sensitive/critical period in the multisensory maturation of both neuronal networks and behavior . Unimodal input and thus , experience-driven electrical activity , has been shown to shape and refine the structure and function of cortical circuits [23 , 33] . For example , monocular deprivation during development dramatically alters the V1 topography , visually induced responses , and visual acuity of the deprived eye [34 , 35] . Similarly , tactile deprivation of the vibrissal sensory system by permanent damage of the sensory periphery ( cauterization or lesion of follicle sinuses ) causes extensive structural changes in S1 [36] . Besides this irreversible and complete tactile deprivation , even transient restriction of tactile inputs in neonatal rats by daily whisker trimming seems to affect the morphology of barrels . In line with the increased dendritic span and spine density of spiny stellate neurons in layer IV after transient vibrissal deprivation [27 , 37] , the size of the barrels in NWT rats was slightly , yet not significantly , increased ( Fig 2 ) . During neonatal development , the transient absence of unimodal inputs may decrease the global level of activation within neonatal S1 , since it diminishes non-whisking related ( so-called spontaneous ) patterns of oscillatory activity and the individual neuronal firing [18 , 38 , 39] . The changed timing of individual spiking , which most likely resulted from the reduced spontaneous activity , perturbed the developmental Hebb-like processes that control the refinement of projections [40] . Thus , the absence of unisensory experience during a defined developmental time window may have long-lasting structural and functional effects on unisensory processing . At adulthood , tactile EPs were enhanced in NWT rats ( Fig 2 ) . This increase in amplitude correlated with the enlarged excitatory receptive fields and augmented cellular responses in layer IV [41] . Since the responses of neurons in layer II/III were normal [42] , this altered evoked activity most likely reflects an impairment of thalamo-cortical responses . Moreover , the neonatal tactile restriction has been reported to cause permanent behavioral deficits in texture discrimination or gap crossing tests [27 , 43] . In contrast , we did not observe significant deficits in tactile object recognition skills ( Fig 8 ) . This discrepancy may be explained by the lower sensorimotor complexity of the present matching task . In addition to the deficits in unisensory processing , the transient tactile restriction during neonatal development caused reduced multisensory processing in S1 ( Fig 3 ) . In contrast to the effects observed in CON rats , neither the EPs were supra-additively enhanced nor the induced activity was modulated by a cross-modal stimulus in NWT rats . It might be argued that the absence of a multisensory enhancement resulted from the effects of neonatal tactile restriction on unisensory processing . As a result , an additional visual stimulus was not able to further augment the enlarged tactile responses in NWT rats . However , a multisensory EP enhancement was not only absent in the G layers but additionally in the S layers , where the unisensory processing was not augmented by neonatal restriction . Thus , we alternatively suggest that the absence of a multisensory enhancement in NWT rats was due to the fact that the transient unisensory deprivation affected not only the cortico-cortical processing , but also the feed-forward thalamo-cortical interactions . We have previously shown that the cross-modal augmentation of early ( ~30 ms ) evoked responses in S1 results from the integration of sensory inputs along the sensory tract preceding primary cortices , most likely within thalamic nuclei [8] . The absence of a cross-modal EP enhancement suggests that unisensory experience controls the emergence of multisensory integration at the subcortical level . The mechanisms by which neonatal unimodal inputs shape the physiology of individual thalamic neurons [44] and the axonal patterning along the feed-forward sensory tract remain to be elucidated . Most research of the past has investigated the emergence of multisensory processing in the SC and association cortices . Multisensory interaction in these areas has been demonstrated to depend on sensory experience [25] . When animals were reared in the dark [45 , 46] or in omnidirectional sound conditions [47] , SC neurons did not integrate any visual-nonvisual or auditory-nonauditory cue pairs , respectively . Critically , these deprivation paradigms precluded not only unisensory but additionally concurrent experience with multiple modalities . In contrast , in the present experimental paradigm we partially deprived the animals of tactile experience during an early neonatal time period when visual inputs activate neither the sensory periphery nor subcortical and cortical circuits ( Fig 1 ) . The pigmented rats opened their eyes at P16–17 and the first visual evoked potentials in V1 emerged 2–3 d before eye opening , reflecting the onset of retinal light sensitivity [48] . By this time , the length and function of neonatally-trimmed whiskers were already fully restored in NWT rats . Thus , the reduction of tactile input during early development caused the multisensory deficits in NWT rats . Though the alternative explanation that impairment of unisensory processing may contribute to multisensory deficits cannot be fully excluded , the intact tactile object recognition of NWT rats is inconsistent with this explanation . Whether distinct intracortical circuits differently involving the S , G , and I layers mechanistically disconnect uni- and multisensory processing remains to be investigated . The present study identified both structural and functional deficits within neocortical networks that might cause abnormal cross-modal processing after neonatal unisensory restriction . Deprivation of a sensory modality has been reported to cause the reorganization of sensory systems for another modality ( cross-modal plasticity ) [49 , 50] . For instance , permanent visual deprivation in neonatal hamsters or mice resulted in reciprocal connections between V1 and other brain regions , such as the major midbrain auditory nucleus , the inferior colliculus and S1 [51 , 52] . Consequently , V1 neurons responded to tactile inputs by increasing their firing rate and synchronizing their depolarization [51 , 53] . It has been hypothesized that these plastic effects are mediated by the stronger reliance on the remaining modality after deprivation [50 , 54] . However , the present findings showed that transient unisensory restriction during neonatal development led to reduced rather than an exuberant V1-S1 connectivity ( Fig 4 ) . In line with these results , visual inputs did not evoke an increased response in S1 in NWT rats . Fewer projections coupling the primary sensory cortices of NWT rats may have caused a phase reset in an atypical , presumably suboptimal frequency band ( Fig 3 ) . Setting of the oscillatory activity to a non-random instantaneous phase has been identified as a ubiquitous mechanism for increasing the processing efficiency of individual stimuli [55 , 56] . A cross-modal phase reset might improve the predictability of the associated stimulus or cause an enhanced salience of the cross-modal stimulus , as proposed for visual-auditory interactions [57] . A phase rest in a different frequency band , as observed in NWT rats , might either has prevented an efficient perceptual encoding or has caused a lower stimulus salience . Consequently , we observed a lower cross-modal transfer during object recognition task in NWT rats . Besides influencing the phase of ongoing oscillations , the fewer cortico-cortical projections seem to decrease the power , synchrony and drive between V1 and S1 in NWT rats ( Figs 5 and 6 ) . The present data provide evidence for the critical role of neonatal unimodal experience for the establishment of multisensory interplay both at the neural and behavioral level . Rats have previously been shown to own excellent skills for cross-modal object recognition [31] . Our data confirmed these findings and identified unimodal inputs during development as a critical factor for the emergence of cross-modal matching abilities of object features ( Fig 8 ) . Rats experiencing neonatal tactile restriction were not able to use previously acquired tactile information for recognizing visually explored objects and vice versa . Remarkably , the behavioral changes were very robust and replicated under different experimental settings . This suggests that learning effects did not influence the behavioral outcome and that the divergent features of objects used for testing did not allow unspecific familiarization . In contrast to the reduced multisensory abilities , the unisensory object recognition was intact after neonatal restriction . In both groups of rats visual discrimination of objects was worse than tactile object discrimination . One explanation for this finding might be the weak saliency or contrast of the used objects or a too small distance to the objects . The abilities for cross-modal object recognition seem to critically depend not only on unimodal sensory inputs during defined developmental stage but also on the integrity of the perirhinal and posterior parietal cortices important for memory retention and retrieval [31] . The precise neural mechanisms integrating sensory and mnemonic aspects of the task remain to be elucidated . In conclusion , the present findings demonstrate that the absence of adequate unisensory experience during a specific developmental period impairs the setting up of cortico-cortical and cortico-subcortical connectivity as well as functional network communication underlying multisensory behavior .
Extracellular recordings were performed under light urethane anesthesia ( 0 . 5 g/kg , i . p . , Sigma-Aldrich , St . Louis , MO ) as previously described [8] . One-shank 16-channel electrodes with 100 μm spacing between recording sites ( 0 . 5–3 MΩ , Silicon Michigan probes , NeuroNexus Technologies , Ann Arbor , MI ) were perpendicularly inserted into S1 ( 2 . 4–2 . 6 mm posterior and 5 . 5–5 . 8 mm lateral to bregma ) and V1 ( 6 . 9–7 . 1 mm posterior and 3 . 4–3 . 7 mm lateral to bregma ) bilaterally to a depth of 1 . 6 mm . Electrodes were labeled with DiI ( 1 , 1′-dioctadecyl-3 , 3 , 3′ , 3′-tetramethyl indocarbocyanine , Invitrogen , Carlsbad , CA ) to enable the post-mortem reconstruction of electrode tracks in histological sections . A custom-built stimulation device placed in front of the animal was used to achieve unimodal ( light-flash , whisker deflection ) or bimodal stimulation [8] . During bimodal stimulation , whisker deflection and light-flashes ( 50 ms , 300 lx , full eye field stimulation ) were presented simultaneously in the same ( congruent ) hemifields with respect to the tactile stimulus . During tactile stimulation , all whiskers were displaced by an upward moving stick activated through compressed air-controlled roundline cylinders gated via solenoid valves . The stimulation device produced almost silent , nonelectrical stimulation with precise timing ( 0 . 013 ± 0 . 810 ms ) that was constant over all trials/ conditions . The stimulation device randomized stimuli in the different stimulation conditions ( unimodal visual , unimodal tactile , bimodal ) . The non-stimulated eye was covered with an aluminum foil patch and ears were sealed with cotton to avoid auditory inputs . The solenoid valves were placed outside the setup and isolated with foamed material to lower their noise . For all stimulation conditions this noise was constant , running out a possible contribution of auditory modulation . Each type of stimulus was presented 100 ± 10 times with an inter-stimulus-interval of 6 . 5 ± 0 . 5 s . The transient restriction of neonatal vibrissal input was achieved by cutting the whiskers close to the intersection with the skin . All of the macrovibrissae were daily trimmed to <0 . 4 mm from 3–12 h after birth until P6 with an electric micro shaver ( ChroMini , Moser , Unterkirnach , Germany ) . Rats were daily weighed and whisker length was measured via a caliper gauge under the microscope ( 20–40x magnification ) ( SZX51 , Olympus , Tokyo , Japan ) . The onset of adult whisking behavior ( amplitude >45° , frequency >5 Hz , duration >2 s ) and time of eye opening were detected . For retrograde tracing , anesthetized rats were fixed in the stereotaxic apparatus on a custom-built mold and received unilateral injections of Fluorogold ( Fluorochrome , Denver , CO ) in S1 ( 2 . 4–2 . 6 mm posterior and 5 . 5–5 . 8 mm lateral to bregma ) . A total volume of 100 nl FG ( 5% in PBS , 30 nl/min ) was delivered via a 26G needle attached to a pump controller ( Micro4 , WPI , Sarasota , FL ) . Three to five days after FG injection , animals were deeply anesthetized with ketamine/xylazine and perfused transcardially with 4% paraformaldehyde ( PFA ) . Brains were removed and postfixed in 4% PFA for 24–72 h . Blocks of tissue containing S1 or V1 were sectioned in the coronal plane at 100 μm , air dried , and examined using ultraviolet excitation filter of the fluorescence microscope ( SZX16 , Digital camera DP72 , Olympus , Tokyo , Japan ) . For quantification , FG-stained cells were detected ( cellSens 1 . 4 . 1 , Olympus ) and counted by eye in an area of 0 . 16 mm2 in S1 ( S4 Fig ) . For COX staining , the rats were transcardially perfused with 4% PFA immediately after electrophysiological recordings . Their brains were removed and halved along the midline . Subcortical brain regions were removed , the cortex was flattened between two acrylic glass plates and postfixed in 4% PFA for 24–72 h . The flattened cortices were sectioned in the transverse plane at 100 μm and processed for COX histochemistry ( Fig 2A ) . The sections were incubated in a solution containing diaminobenzidine ( 0 . 5 mg/ml ) , cytochrome C ( 0 . 6 mg/ml ) , katalase ( 0 . 36 mg/ml ) , saccharose ( 44 . 4 mg/ml ) , and examined using light microscopy . Individual barrels were visually identified and their borders manually defined by monitoring the contrast changes ( cellSens 1 . 4 . 1 and Adobe Photoshop , 11 . 0 . 2 ) . The size of the first four barrels in the rows A–D was calculated and compared between CON and NWT rats ( S2B Table ) . The litters used for behavioral testing included both males and females to avoid sex-based maternal behavior biases [58] . All behavioral tests were conducted during the light phase and by one investigator . Rats were daily weighed from P0 to P21 . All experiments were conducted in a black acrylic glass arena with a white ground ( L: 52 cm , W: 30 cm , H: 32 cm ) . The seven objects used for testing of novelty recognition were seven differently shaped and textured , easy-to-clean items ( Fig 8A , S2A Table ) that were provided with magnets to fix them to the bottom of the arena ( 15 cm from the borders and 10 . 5 cm from the center of the arena ) . Object sizes ( H: 5–9 cm , diameter: 3–4 . 5 cm ) were smaller than twice the size of the rat and did not resemble living stimuli ( no eye spots , predator shape ) . Glass containers ( L: 6 cm , W: 5 . 5 cm , H: 10 . 5 cm ) were placed over the objects to prevent tactile object exploration during visual exploration . During tactile exploration , the light was switched off and a red-tinted bulb ( 18 W ) illuminated the arena to exclude any contribution of visual perception to the exploration of objects . After every experiment the objects and arena were cleaned with 70% ethanol to remove all odors . A black and white CCD camera ( Videor Technical E . Hartig GmbH , Roedermark , Germany ) was mounted 90 cm above the arena and connected to a PC via PCI interface serving as frame grabber for video tracking software ( Video Mot2 software , TSE Systems GmbH , Bad Homburg , Germany ) . First , exploratory behavior was assessed 1–3 d after eye opening in P17–19 rats . The animals were allowed to freely explore the empty arena for 15 min . During this time , which was defined as habituation phase ( open field ) , rearing , wall-rearing and grooming were quantified in their occurrence and duration . The floor of the arena was digitally subdivided into 16 zones of the same size ( four in the corners , eight at the borders , and four in the open field ) . The time spent in , the travel speed and distance were measured for each zone . Second , rats were familiarized with the arena and testing conditions . This phase was defined as familiarization phase ( S7 Fig ) . Five minutes after the open-field test , P17–19 rats were placed back into the arena and allowed to explore for 5 min the two glass containers that later were used in the recognition task exclusively relying on visual perception . The following day , rats were familiarized again for 5 min in the arena with the glass containers and for 10 min with the red light-illuminated arena and glass containers . Third , novel object recognition was investigated in a series of tests . All protocols for assessing bimodal , unimodal and cross-modal object recognition were tested in P19–23 rats . During the sample trials , rats were placed into an arena containing two identical objects and released against the center of the opposite wall with their back to the objects . After 10 min of free exploration of objects , the rat was returned to a temporary holding cage . During the test trials one familiar object was replaced by a novel object with a different texture and visual appearance ( e . g . , cube instead of cylinder ) . One day after familiarization with the empty arena and the two glass containers , two groups of rats were tested in different behavioral settings . In the first setting ( continuous setting ) , bimodal object recognition was tested in rats relying on both visual and tactile perception . Two days after familiarization , unimodal object recognition was tested in rats randomly assigned to either the group relying exclusively on visual perception ( light , objects covered by glass containers ) or the group relying exclusively on tactile perception ( red light , no glass containers ) . Three days after familiarization , cross-modal object recognition ( cross-modal matching ) was tested in rats relying in the choice phase on a different sensory modality than during the sample phase . In the second behavioral setting ( discrete setting ) , each rat was tested in only one sensory condition . The trials were video-tracked and the analysis was performed using the Video Mot2 analysis software . The object recognition module of the software was used and a three-point tracking method identified the head , the rear and the center of gravity of the rat . Rats that did not interact with objects in the sample phase were excluded from analysis . Digitally , a circular zone of 2 cm was created around each object and every entry of the head point into this area was considered as object interaction . For tactile and bimodal exploration , the first 5 min of interaction with the objects were analyzed , whereas for visual exploration , the analysis was confined to the first 2 min . Climbing or sitting on the object , mirrored by the presence of both head and center of gravity points within the circular zone , were not counted as interactions . Relative interaction time tRI was calculated by tRI= tNOtNO+tFO where tNO/tFO is the time that the rat spent with the novel/familiar object . Significance was tested between objects . Discrimination ratio DR was calculated as DR= tNO−tFOtNO+tFO Data were imported and analyzed offline using custom-written tools in Matlab software version 7 . 7 ( MathWorks , Natick , MA ) . Data were recorded at a sampling rate of 32 kHz . For anti-aliasing , the signal was band-pass filtered ( 0 . 1 Hz–5 kHz ) by the Neuralynx recording system . A third order Butterworth filter was applied . The subsequent downsampling of the data was analysis-dependent ( factor 5 for EP calculation , factor 30 for spectral and phase analysis , factor 10 for coherence analysis , factor 128 for Granger directionality analysis ) . Data in the text are presented as mean ± SEM . All values were tested for normal distribution with Lilliefors test ( α = 0 . 05 ) . For normally distributed values t tests , for not normally distributed values or if n<8 , Kruskal-Wallis test detecting significance levels of p < 0 . 05 ( * ) , p < 0 . 01 ( ** ) , and p < 0 . 001 ( *** ) were used ( S1 Table ) . For behavioral testing and anatomical investigations the given n corresponds to the number of investigated rats . For statistics on EPs , the given n corresponds to the number of stimulation trials that reliably evoked activity with consistent peak timing . This better mirrors the subtle modulatory effects on EPs , since cross-modal effects were previously observed for congruent ( i . e . , whisker deflection and light stimulus were presented in the same hemifield ) but not for incongruent ( stimuli applied on opposite hemifields ) stimulation conditions when tested in the same group of rats [8] . For the remaining electrophysiological data , we considered the values recorded from each hemisphere , because , according to our stimulation paradigm ( Fig 1A ) , they were independent ( different stimulation time-points , recording electrode , recording depth ) . Variable n values resulted from the exclusion of animals with insufficient behavioral performance during control conditions and of recordings outside the layers of interest or containing artifacts/errors after pre-processing . The high number of statistical tests requires corresponding correction ( i . e . , reduction of type I errors ) by using the Holm-Bonferroni method . However , in line with previous recommendations [65 , 66] a trade-off between type I and type II errors ( i . e . , the probability of accepting the null hypothesis when the alternative is true ) should be maintained . Therefore , we corrected only test series with more than 20 observations , meaning that for a threshold α = 5% one false positive would be expected .
|
Our senses , working together , enable us to interact with the environment . To obtain a unified percept of the world , diverse sensory inputs need to be bound together within distributed but strongly interconnected neuronal networks . Many multisensory abilities emerge or mature late in life , long after the maturation of the individual senses , yet the factors and mechanisms controlling their development are largely unknown . Here , we provide evidence for the critical role of unisensory experience during early postnatal life for the development of multisensory integration . Focusing on visual-tactile interactions in pigmented rats with good visual acuity , we show that a transient reduction of tactile inputs during neonatal development leads to sparser direct connections between adult primary visual and somatosensory cortices . As a result , these animals showed reduced neuronal activation following co-occurring tactile and visual stimuli , as well as impaired communication within visual-somatosensory networks . The structural and functional deficits resulting from an early manipulation of tactile experience had major behavioral consequences , impairing the rats’ ability to transfer information about encountered objects between senses . Thus , unisensory experience during early development shapes the neuronal networks of multisensory processing and the ability to transfer cross-modal information .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2015
|
Neonatal Restriction of Tactile Inputs Leads to Long-Lasting Impairments of Cross-Modal Processing
|
Because lymphatic filariasis ( LF ) elimination efforts are hampered by a dearth of economic information about the cost of mass drug administration ( MDA ) programs ( using either albendazole with diethylcarbamazine [DEC] or albendazole with ivermectin ) , a multicenter study was undertaken to determine the costs of MDA programs to interrupt transmission of infection with LF . Such results are particularly important because LF programs have the necessary diagnostic and treatment tools to eliminate the disease as a public health problem globally , and already by 2006 , the Global Programme to Eliminate LF had initiated treatment programs covering over 400 million of the 1 . 3 billion people at risk . To obtain annual costs to carry out the MDA strategy , researchers from seven countries developed and followed a common cost analysis protocol designed to estimate 1 ) the total annual cost of the LF program , 2 ) the average cost per person treated , and 3 ) the relative contributions of the endemic countries and the external partners . Costs per person treated ranged from $0 . 06 to $2 . 23 . Principal reasons for the variation were 1 ) the age ( newness ) of the MDA program , 2 ) the use of volunteers , and 3 ) the size of the population treated . Substantial contributions by governments were documented – generally 60%–90% of program operation costs , excluding costs of donated medications . MDA for LF elimination is comparatively inexpensive in relation to most other public health programs . Governments and communities make the predominant financial contributions to actual MDA implementation , not counting the cost of the drugs themselves . The results highlight the impact of the use of volunteers on program costs and provide specific cost data for 7 different countries that can be used as a basis both for modifying current programs and for developing new ones .
Lymphatic filariasis ( LF ) , commonly known as elephantiasis , is a profoundly disfiguring parasitic disease caused by thread-like nematode worms . The World Health Organization ( WHO ) places the number of people at risk in 83 countries at 1 . 307 billion . [1] Globally , the reduced productivity as a consequence of LF disability has been well recognized . The chronic and debilitating burden of LF maintains the cycle of poverty not only in infected individuals but also in entire endemic communities . [2 , [3] Indeed , as a disease of poverty , LF is endemic in 43 of the 50 countries classified as least developed nations . [4–6] The 1997 World Health Assembly resolved to eliminate LF as a public health problem after LF had been identified as one of only a small number of diseases classified as potentially eradicable . [7] The principal strategy for elimination relies on once-yearly concurrent administration of two drugs , albendazole with diethylcarbamazine ( DEC ) or albendazole with ivermectin – both regimens shown to be highly effective in removing microfilariae from the blood for a full year after treatment . [8] Administration of these once-yearly , single-dose regimens to people in at-risk communities for 4–6 years makes feasible the prospect of interrupting transmission and thereby eliminating LF , [9] largely because the reproductive life span of the adult worm is estimated to be 4–6 years . While the delivery of DEC fortified salt is another strategy that has been applied in some regions of the world to eliminate LF , this strategy was not in the purview of the current studies . Demonstrating that LF elimination is a cost-effective and affordable investment is essential for both Ministries of Health and potential donors as they choose among competing health needs . While several cost-effectiveness analyses have been conducted and one estimate for MDA in high prevalence areas gives a range of $4–8 per disability adjusted life year ( DALY ) averted , [2 , 10–13] more cost and cost-effectiveness data are lacking . The persistent need for detailed costing of mass drug administration ( MDA ) programs led to the present initiative to develop and implement a common cost-analysis protocol . Countries participating in this multi-country study were selected to represent the different stages and scope of national LF elimination efforts , different economic conditions of endemic countries , and different geographic regions affected by the disease . Health expenditures per capita in the participating countries for 2001–2002 ranged from $12 in Tanzania and Ghana , to $22 in Haiti , $27 in Burkina Faso , $30 in the Philippines , $46 in Egypt and $153 in the Dominican Republic . [14 , 15] The cost analysis objectives were to 1 ) estimate total annual costs of the LF program for a specific year or more ( depending on the availability of information ) , 2 ) calculate the average cost per person treated , 3 ) identify the relative contributions of the endemic country and external partners , 4 ) provide data that could be used for program evaluation and analysis , 5 ) understand how the results differed among countries , and 6 ) build a framework for the development and implementation of cost studies elsewhere . Cost analyses aggregated in this report are based on studies in Burkina Faso , the Dominican Republic , Egypt , Ghana , Haiti , the Philippines , and Tanzania . Detailed analyses from two of these countries have already been published elsewhere . [12 , 16]
Data collection and analysis were carried out by research teams in each of the participating countries . The cost analysis teams in Burkina Faso , Tanzania , and Ghana were part of the Ministry of Health . In Egypt the study was a joint effort between the Ministry of Health and Population and the Ain Shams University Research and Training Center on Vectors of Diseases . The Philippines' research team was based at the National Institutes of Health-Philippines . The Dominican Republic researchers at the Centro de Gerencia Social , Instituto Tecnologico de Santo Domingo worked with the Ministry of Health LF program . External consultants in Haiti worked with Ministry of Health LF program staff to collect and analyze cost study data . Researchers from the participating countries collaborated with the Emory University LF Support Center in the development of a protocol ( http://www . taskforce . org/lfsc/professionals . html ) which served as the tool to identify , organize , analyze , summarize and present the cost data . The protocol and accompanying data collection instrument created a systematic process for data collection and analysis so that the cost estimates would be comparable across a variety of country settings , programmatic approaches and program sizes . The cost analysis identified both economic costs ( i . e . , the value of all resources used in the program , including donated items , such as medications for the MDA ) and financial costs ( i . e . , the actual cash disbursements for a program including resources provided by the national government and local communities ) . Economic costs are useful to evaluate the allocation of program resources and their opportunity costs , e . g . , whether these resources could be used more productively elsewhere . Financial costs are helpful to program managers to measure actual program expenditures and assess affordability . [17] ( See Table 1 for the classification of costs . ) Financial costs include all costs with the exception of donated materials . Capital items were annualized to reflect costs incurred in one year for the project . They do not include bicycles because these were the property of the volunteer drug distributors , and thus considered a donation . Volunteer training per diems are classified as Financial and Economic costs , but the value of volunteer unpaid time is not included in either Financial or Economic costs . Because the capital costs were annualized , the cash expenditures described reflect those incurred in one year for the project . The protocol adopted a national program perspective because most resources dedicated to the LF MDAs were channeled through Ministry of Health ( MOH ) programs . Participants agreed to gather MDA costs , beginning with the year 2000 , the first year of the LF global elimination effort . Costs were calculated in local currencies and converted to US dollars for the final analysis , based on average exchange values for the years being analyzed ( base year 2002 ) . Most of the countries were still in the early stages of expanding their MDAs when the cost analysis studies began . The exception was Egypt where the MDA targeted almost 90% of the population at risk the first year and 100% the second year studied . Project inputs defined in the protocol were: personnel , supplies and drugs ( medications for the MDA and for treatment of adverse reactions ) , as well as capital and recurrent costs for equipment , transportation , and facilities . Each input was allocated to one or more categories of activities involved in accomplishing the MDA: training , mapping , mobilization and education , drug distribution , adverse reaction monitoring , surveillance/laboratory , and administration ( Table 2 ) . Capital costs were defined as one-time investments in physical goods that have a life longer than one year and generally cost more than $1000 . These costs were annualized , using a formula that accounted for years of useful life , scrap value , and a discount rate set at 3% . Recurrent costs included those items that were consumed on a regular basis; i . e . , personnel time , office and laboratory supplies , fuel , and drugs . Recurrent costs for maintenance of donated capital items such as bicycles were also included . In cases where facilities or equipment costs were not available , costs of similar facilities or rentals were used as a proxy . For the purpose of Economic costs , the value of the donated drugs was set as $0 . 19 plus $0 . 0019 for shipping per 400mg tablet of albendazole ( Personal Communication: GlaxoSmithKline . February 17 , 2004 ) and $1 . 50 plus $ . 0018 per 3 mg tablet of Mectizan ( Personal Communication: Merck & Co . , Inc . March 1 , 2004 ) DEC , used in four of the seven countries studied , was not donated by the private sector and had to be purchased for the national programs . Study sites were chosen purposively to be representative of the LF- endemic regions in the countries ( Table 3 ) . National estimates were developed using the data from these representative study sites . Tanzanian investigators studied four districts located on the eastern coast , Kilwa , Mafia , Mkuranga and Masasi for the years 2000–2003 . Data collection for the first three years was retrospective . Data collected retrospectively in Burkina Faso documented the first two years of the MDA in the region of Gaoua in the first year and expansion into Tenkodogo in the second year . The Philippines team chose to study seven municipalities and one city in the province of Sorsogon . These were selected based on filariasis endemicity using microfilaria rates , accessibility and population size . Half had high microfilaria rates ( MF ) , half were easily accessible , while one quarter had large populations . In each municipality , one sentinel barangay or village was selected based on high MF rates and one to three adjacent barangays were also studied . Cost data collection in Haiti was conducted in Leogane , where the first MDAs took place . Located 30 km west of Port-au-Prince in the south , Leogane is one of the highest risk areas in the country . Except for lower coverage in year-2 ( 2001 ) of the MDA program ( because of side reactions that occurred after the first MDA ) the program expanded not only in Leogane in 2002 but also to Milot , another high risk area just south of Cap Hatien in the north . Researchers in the Dominican Republic collected cost data for the first two MDA campaigns . These were launched in Barahona on the southwest coast of the country . The Ghana study costed three districts from the two epidemiological zones in the country , north and south . The districts were selected to reflect differing levels of program assistance . Builsa in the upper eastern region of the country received government support only , while Lawra in the upper west and Ahanta West on the west coast also received NGO support . The cost analysis in Egypt covered the eight governorates along the Nile affected by LF . Costs were obtained for one district in each governorate and applied to the rest of the affected areas in the governorate based upon information about numbers of persons at risk , participating government personnel and quantities of medication distributed . [16] Data collection was both retrospective and prospective . The number of rounds of MDA costed per country ranged from one to four , with most countries costing two rounds . Countries developed national estimates for the program using data collected from a sampling of sites representative of the program ( Table 3 ) . Coverage rates are defined as the number of individuals reported to have ingested the antifilarial drugs divided by the total at-risk population in the program area . Those excluded from treatment included pregnant women , lactating women in the 1st week post-partum , the very sick , children under two years of age in countries where DEC plus albendazole is the MDA regimen , and children under 90 cm in height ( generally under 5 years of age ) where albendazole is administered with Mectizan . The data were collected from national , regional and district levels of the health care system via pre-tested questionnaires and spreadsheets , sometimes with the assistance of other agencies such as Ministries of Agriculture and Information . As LF is but one of many population-based health programs , most inputs ( including personnel time , facilities , equipment , supplies , vehicles and fuel ) were often shared by more than one program , and costs were apportioned accordingly . To capture the actual costs and percentage of the resources dedicated to LF , the teams reviewed program records and interviewed LF program administrators and personnel about allocations of personnel and resource time per year . Government tax fees such as customs tax on the drugs and the road tax were excluded .
As indicated in Table 4 , the Financial costs per person treated ranged from $0 . 06 to $2 . 23 while Economic costs varied between $0 . 40 and $5 . 87 . MDA coverage rates in the study populations ranged from 53% to 91% . While cost per person at risk can be easily calculated , cost per person treated is the more useful summary of costs for the purposes of planning and operations . Of the several trends that can be seen in Table 4 , the most notable is that for those countries for which there is more than one year of data , cost per person treated decreased after the first year of the program , especially as the number of persons treated increased ( see Figure 1 ) . In addition , the Financial costs per person treated for Burkina Faso , Ghana , Tanzania and the Philippines , all of which used volunteers , were the lowest among the seven countries participating in the study . Identification of resource allocation by activity and input is a useful and informative outcome of this study . The use of resources for different activities varied among countries , and Table 5 identifies the proportion of the average national Financial costs expended for each activity in a ‘non-start-up’ MDA round ( since yearly costs tend to stabilize after the ‘start-up’ year ) . Drug distribution generally represented the largest proportion of financial expenditures ( average of 46% ) , with social mobilization/education and administration being next most prominent . Analysis of financial costs by input , particularly useful for projecting budgets and for gauging the need for additional program support , again gave results varying appreciatively by country ( Table 6 ) . In all but one country , Egypt , the input that consumed the largest proportion of financial resources was personnel , averaging 53% , followed by supplies , equipment/facilities and transportation . These cost analyses not only identified how program resources were allocated , but data also were used to identify sources and amounts of funding for 5 of the 7 study countries . Funding was categorized as coming from national governments ( excluding external donations for LF ) , international organizations ( IDAs , NGOs , WHO ) , pharmaceutical companies and local communities ( Table 7 ) . As expected , the drug donations represent a large proportion of contributions to MDA programs , so that when MOH and partner contributions are examined from the perspective of Economic costs , the drug donations can make up over 90% ( range 9%–99% ) of the costs . This was particularly true in countries like Burkina Faso and Tanzania where both drugs used in the MDA ( albendazole and Mectizan ) were donated . When Financial costs are analyzed , however , it is clear that contributions from national governments represent a significant portion of the resources used to implement the MDAs ( average = 56% , range 9%–99% ) . These relationships can be seen graphically in Figure 2 which presents the combined average of Financial and Economic funding sources for the Burkina Faso , the Dominican Republic , Egypt , the Philippines and Tanzania programs detailed in Table 7 . Sensitivity analyses were conducted on the personnel input ( data not shown ) to gauge how much an increase in personnel costs would impact overall costs . Personnel was selected for sensitivity analysis because , with the exception of Egypt , this input represented the largest source of costs in the participating countries . Sensitivity analyses conducted on personnel input for Ghana and Burkina Faso demonstrated that increasing the proportion of personnel time dedicated to the LF MDA did not raise financial costs beyond the currently-calculated ranges , even under conditions where personnel time was doubled ( Ghana - US$0 . 25 and Burkina - US$0 . 08 ) or tripled ( Ghana - US$0 . 34 and Burkina - US$0 . 11 ) . Additionally , a sensitivity analysis of personnel costs in the Philippines showed that personnel costs would remain well within the range of the current Financial costs ( US$ 0 . 69 ) even if personnel time were doubled . There were three principal study limitations . First , investigators frequently encountered problems in estimating the proportion of time and money allocated specifically to the LF MDA programs . Staff had not previously tracked time apportioned to LF-specific activities , and since LF programs are small in relation to other MOH initiatives ( consuming anywhere from 1% to 5% a year in most countries ) , recall was problematic at times . Similar situations were encountered with respect to other resources such as transportation , facilities and equipment dedicated to LF . Second , interpretations of the definitions of certain activity categories sometimes differed among researchers applying the cost analysis protocol and data collection instrument . At times questions arose about allocation to one or the other of categories mentioned . Third , while study teams attempted to identify the original source of funds , the original funder may not have been correctly identified in all cases , especially where money was channeled through one or more intermediary sources .
MDA program start-up years resulted in higher Financial and Economic costs per person treated than did subsequent years . Inevitably , there are more costs and reallocation of resources in the first year of a program . At the same time , new programs typically cover a limited geographic area and a relatively small population in their first year . As the population covered expands after the first year , the cost per person treated tends to drop . In addition to extra expenditures in a start-up year as compared to other years , costs can also decrease over time because new cost-savings strategies are identified and implemented . For example , in the first MDA year in the Dominican Republic , health workers from non-governmental organizations participated in the MDA; these individuals were paid per diem for all of the time they participated , in some cases higher than the wages of workers in the health system . MDA integration into the health system contributed to the 50% decrease in the second year Program cost per person treated . Similarly in Haiti , after the first year cost analysis revealed that 22% of resources were dedicated to adverse reactions , the adverse reaction protocol was revised to an equally effective but less costly strategy . In the second year only 4% of resources were used for adverse reactions . The use of volunteers had the greatest impact on costs . In Burkina Faso , Tanzania , Ghana , and the Philippines - where Program costs are lowest - health workers are employed down to the sub-district level , and volunteers , who are compensated very little in the LF program , work at the village level . Volunteers in these countries contribute a large proportion of the time dedicated to the MDA ( frequently receiving per diem only for days in training ) and at times provide their own transportation ( e . g . , Burkina Faso ) . However , while there is a strong relationship between the use of volunteers and lower cost per person treated , this does not suggest that any country choosing to use volunteers would see savings of 85% . Country-specific conditions that lead to the use of volunteers may also determine lower costs overall . The use of volunteers leads to the question of how to accurately value the time volunteers ‘donated’ to each program . One traditional approach to valuing volunteer time is to apply the wage from the volunteer's normal paid employment and value their volunteer time accordingly . The problem with this , as pointed out by McFarland et al . [18] in a report on the costs of onchocerciasis MDA , is that many volunteers are subsistence farmers who do not participate in the formal labor market . One alternative is to decide the fair market value of the time , i . e . the amount the volunteer would be paid if the program had to hire individuals for the work , [19] or another , by using an estimate from prior studies in a similar setting . [18] In the country programs included in this study , a diverse group of individuals , comprising students , teachers , farm laborers , and elderly retirees , served as volunteers . Owing to this fact and the economic conditions in the participating countries , applying traditional methodologies for costing volunteer time may not be appropriate . After several countries explored volunteer participation in MDAs and the earning capacity of volunteers in their regular pursuits , a decision was made by the investigators in all participating countries not to include these costs . Therefore , while the study included the per diem paid to volunteers during training , it did not account for the opportunity cost of the volunteer's time dedicated to the MDA itself . The evaluation of the community contribution from countries which used volunteers is very definitely underestimated . Program managers can control , to some extent , the use of volunteers and might explore this strategy in resource-constrained environments . Beyond the monetary savings , there is a benefit from connecting the program with volunteers who are opinion leaders from different community sectors . So while initially decisions to utilize volunteers might be financially based , countries ultimately can benefit not only from volunteers' labor but also from their connections to the communities in which they serve . However , program managers must bear in mind the competition for volunteers from other health programs which sometimes pay volunteers more for their efforts . The sensitivity analyses on personnel time devoted to the MDA point to the possibility that it may be worthwhile exploring opportunities to increase remuneration for volunteers . Indeed , both valuing and best utilizing volunteer time merit continued examination . The third principal source of variation in cost per person treated was the size of the population treated , an element that can be controlled by program managers during program expansion . When programs scale-up , the cost per person treated drops , primarily because most of the overhead costs are associated with start-up costs at the national and district levels . Once the system has been established , the majority of the new costs are in the new area ( district or governorate ) that is being covered . In Burkina Faso and Tanzania the Financial costs per person were halved as the MDA covered larger populations . Between 2001 and 2002 Program costs in Burkina declined 45% while the population treated increased by over 400% . The progression in Tanzania between 2000 and 2003 was similar , with the Financial cost per person treated decreasing by 47% while the treated population increased over 13-fold . Between 2000 and 2001 , the number of persons treated in Egypt increased 29% , while the Financial cost per person treated decreased 27% . Such findings emphasize the need to keep current programs adequately funded so that these programs can expand and increase the number of persons treated and thereby capture the savings resulting from the economies of scale . Once the principal determinants of the cost per person treated are identified , they can be used to manage program costs either by making internal changes within a program , such as scaling-up and increasing efficiency , or by taking advantage of existing external resources , as through integration . LF elimination programs can be integrated into the existing health system , as in the Dominican Republic , or with other preventive chemotherapy programs . For the LF elimination program , costs per person treated are within the range of those estimated for other similar disease control and elimination programs; namely soil-transmitted helminths ( $0 . 25 per treatment[20] ) , trachoma ( $0 . 50 per treatment[21] ) and onchocerciasis ( $0 . 58 per treatment[18] ) . The potential programmatic overlap of activities among these and other public health initiatives includes administration , drug distribution , monitoring , surveillance , and social mobilization , so attractive cost-saving opportunities can be envisioned through integration . While estimates place the cost savings from integrated delivery between 25% and 47% , [22 , 23] the LF cost analysis protocol utilized in the present cost study can provide a useful tool by which to document the potential savings from integrating some or all programmatic activities of these initiatives . Indeed , countries that have already completed the LF cost analysis are well placed to estimate costs of integrated programs , given that many of the costs common to all programs have already been identified . Also of particular interest from this study was the documentation of the substantial contributions ( i . e . , program ownership ) by national governments ( Figure 2 ) . On average 56% of Financial costs of the LF elimination programs were financed by governments . The Egyptian government contributed 80% of Program costs , including participation of the Ministries of Health and Population , of Agriculture , and of Information . This is consistent with similar programs for TB and malaria which estimate that 70% of Financial costs are paid for by national governments . [24 , 25] Furthermore , LF MDA programs do not require new or additional funding for all inputs . For example , the portion of the salary of a District Medical Officer ( DMO ) spent on the project would be included in the analysis as a program cost , but the DMO's salary would have been funded regardless of whether the time was spent on LF MDA activities or other programs and may not require additional financial outlays . Governments have a choice as to where they use their resources , and the percentages noted in this study emphasize the commitment these national governments have made to LF elimination . Surveillance and mapping activities represented a significantly higher proportion of resource requirements in Egypt than in other countries ( between 10% and 12% at both the national and governorate levels ) , since Egypt has selected small MDA implementation units , i . e . the village level , and hence focused on monitoring potential ‘at-risk’ areas closely in a large number of implementation units . Tanzania's national program chose to invest heavily in social mobilization and education to assist the districts in raising awareness in the general population for current and future MDAs . The districts placed more emphasis on funding drug distribution and personnel training . The principal aim of this study was to provide critical information regarding the cost of implementing MDA for the prevention of lymphatic filariasis . The study demonstrates that the costs of MDA programs for LF elimination are comparable to those estimated for other similar disease control and elimination programs . While some programs had costs per person treated of over a dollar , it was quite straightforward to identify those factors most affecting program costs . Such findings can be used on a national scale for program planning , development and fundraising , and on a global scale for calculating current global costs , predicting scale-up costs and calculating savings from integration with other programs . These results also will form the basis for guiding cost-effectiveness and cost-benefit analyses , as more information on the effectiveness of MDAs in the study countries becomes available . Additionally , the analytic tool used in this study will be valuable for further studies of the LF elimination program including costing of LF disability alleviation activities and the process certifying LF elimination . A further finding of particular importance was documentation of the impact that the use of volunteers has on program costs . Further research on how best to utilize volunteerism for such public health programs could contribute appreciably to ensuring success of the MDA-based programs and to anchoring them in the communities they seek to protect . Finally , implementation of this study produced a number of ancillary program benefits . The process of defining and reviewing costs allowed for a review of operations at all levels ( inputs , processes and outputs ) that was also frequently used to assess efficiency . The findings provide the opportunity for the development of cost-effective implementation models built on best practices from each country; hopefully , these can be adopted and adapted by new programs from the start , particularly in Africa where almost 30 more endemic countries still need to initiate LF elimination programs .
|
Lymphatic filariasis ( LF ) , commonly known as elephantiasis , is a profoundly disfiguring parasitic disease caused by thread-like nematode worms . This disease can often be disabling , thus reducing the potential productivity of the affected individuals . The WHO places the number of people at risk in 83 countries at 1 . 307 billion . This study was undertaken in seven countries—Burkina Faso , Ghana , Egypt , Tanzania , the Philippines , the Dominican Republic , and Haiti—using a common protocol to determine the costs of mass drug administration ( MDA ) programs to interrupt transmission of infection with LF , because there is lack of sufficient information about the costs of these programs . The results demonstrate that LF MDA is affordable and relatively inexpensive when compared to other public health programs . In the context of initiatives for integrating programs for the control and elimination of neglected tropical diseases , this study adds specifically to the relatively scarce body of information about the costs of MDA programs for LF . It also adds to the general knowledge about the application of methods that can be used to estimate the costs and cost-effectiveness of an integrated approach .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"public",
"health",
"and",
"epidemiology/global",
"health",
"public",
"health",
"and",
"epidemiology/infectious",
"diseases"
] |
2007
|
National Mass Drug Administration Costs for Lymphatic Filariasis Elimination
|
Parkinson disease ( PD ) is characterized by the preferential , but poorly understood , vulnerability to degeneration of midbrain dopaminergic ( mDA ) neurons in the ventral substantia nigra compacta ( vSNc ) . These sensitive mDA neurons express Pitx3 , a transcription factor that is critical for their survival during development . We used this dependence to identify , by flow cytometry and expression profiling , the negative regulator of G-protein signaling Rgs6 for its restricted expression in these neurons . In contrast to Pitx3−/− mDA neurons that die during fetal ( vSNc ) or post-natal ( VTA ) period , the vSNc mDA neurons of Rgs6−/− mutant mice begin to exhibit unilateral signs of degeneration at around 6 months of age , and by one year cell loss is observed in a fraction of mice . Unilateral cell loss is accompanied by contralateral degenerating neurons that exhibit smaller cell size , altered morphology and reduced dendritic network . The degenerating neurons have low levels of tyrosine hydroxylase ( TH ) and decreased nuclear Pitx3; accordingly , expression of many Pitx3 target gene products is altered , including Vmat2 , Bdnf , Aldh1a1 ( Adh2 ) and Fgf10 . These low TH neurons also express markers of increased dopamine signaling , namely increased DAT and phospho-Erk1/2 expression . The late onset degeneration may reflect the protective action of Rgs6 against excessive DA signaling throughout life . Rgs6-dependent protection is thus critical for adult survival and maintenance of the vSNc mDA neurons that are most affected in PD .
Parkinson disease ( PD ) is characterised by the progressive loss of midbrain dopaminergic ( mDA ) neurons [1] . Although the clinical manifestations of PD can be variable , the appearance of motor deficits is the hallmark of this neurodegenerative disease . Similarly , the etiology of PD appears to be multifactorial but one consistent feature of this disease is the greater sensitivity of ventral substantia nigra compacta ( vSNc ) mDA neurons to degenerate [2] as opposed to mDA neurons of the dorsal SNc ( dSNc ) and ventral tegmental area ( VTA ) . The molecular basis for this preferential sensitivity remains poorly understood although work in animal models has been useful [3] . Vertebrate animal models based on genetic causes of PD , which constitute 10–20% of PD cases , have not been overly successful in reproducing the selective neurodegeneration patterns of the mDA system [4] . For example , mice with transgenic expression of human autosomal dominant mutants of α-synuclein ( SNCA ) or leucine-rich repeat kinase 2 ( LRRK2 ) rarely produce mDA neurodegeneration [5] , [6] . Murine loss-of-function mutations in autosomal recessive gene products for PTEN induced putative kinase 1 ( Pink ) and Parkinson protein 2 ( Park2 ) , have not been enlightening either , with the recent exception of Parkinson protein 7 ( Park7 or DJ-1 ) . Indeed , DJ-1−/− mice show progressive adult degeneration of SNc mDA neurons upon backcrossing to an appropriate genetic background [7] , indicating that many factors are necessary in order to model polygenic diseases such as PD . However , it is noteworthy that rat models of Pink1 and DJ-1 loss-of-function showed progressive loss of mDA neurons [8] . Both early-onset and late-onset forms of PD bear a major histopathological hallmark , the presence of Lewy bodies , which are α-synuclein-rich protein inclusions that are also found in non-dopaminergic brain regions depending on the stage of disease progression . Many familial PD genes have widespread brain expression , without any preferential expression in mDA subpopulations [5] . In general , they all participate in similar inter-related cellular processes such as in mitochondrial function ( PINK1 , DJ-1 , SNCA ) , the secretory pathway ( PARK2 , LRRK2 ) and the ubiquitin-proteasome degradation pathway ( PINK1 , SNCA , PARK2 ) . Rodent animal models based on genes that participate in development and survival of mDA neurons ( Pitx3 , Nurr1 , Girk2 , En1/2 , Otx2 , etc ) have proved extremely useful , especially in defining the candidate cellular pathways underlying the respective differential vulnerability of SNc versus VTA mDA neuron subpopulations to toxin-induced neurodegeneration and in human pathology [9]–[11] . Notably , mouse mutants for the homeobox transcription factor Pitx3 ( Entrez gene ID: 5309 ) are unique in that they display a selective and stereotypic pattern of mDA cell loss that resembles typical PD [12]–[14] . In particular , Pitx3-deficient mice exhibit developmental loss of Pitx3-positive Calbindin ( Calb ) -negative mDA neurons of the vSNc ( Pitx3-dependent ) while Pitx3-negative Calb-positive mDA neurons of dSNc and VTA ( Pitx3-independent ) remain essentially unaffected by Pitx3 deficiency [15] . This cell loss is associated with a loss of spontaneous movement that can be partially rescued by L-dopa treatment [16] , [17] . Human PITX3 polymorphisms are associated with sporadic PD [18] . Pitx3 mediates its effects by regulating the expression of many genes ( Aldh1a1 , DAT , Drd2 , TH , Bdnf ) in a subset specific fashion [19] . A well-studied Pitx3 target gene in mDA neurons of the SNc is Aldh1a1 , which is important for retinoic acid production and subsequent neuronal maturation and protection through regulation of TH expression [19]–[21] . Other Pitx3 target genes , such as the classical dopaminergic markers DAT , Vmat2 , Drd2 are important for neurotransmitter identity and are subject to the cooperative action of Pitx3 with Nurr1 [22] . On the other hand , Pitx3 expression has been reported to be itself regulated by GDNF , especially during development [23] . GDNF is the only neurotrophic factor for which conditional inactivation in the adult mouse has provided strong evidence of its absolute requirement for the cell-autonomous survival of brain cathecholaminergic neurons , including mDA neurons [24] . Although , many studies described the implication of Pitx3 in post-natal maturation and developmental survival of mDA neurons , Pitx3 has not been conclusively linked to mechanisms of survival and maintenance of mDA neurons in the adult , especially as it pertains to degenerative processes . We hypothesized that Pitx3-controlled genes and pathways , in addition to their known role in development , may also be implicated in the neuroprotective pathway required to maintain the integrity of specific subset of mDA neurons throughout adulthood . In a screen of expression profiling data comparing Pitx3-dependent and -independent FACS-purified mDA neurons of SNc and VTA , the Regulator of G-protein Signaling 6 ( Rgs6 ) ( Entrez Gene ID: 9628 ) was identified as a putative survival factor that is preferentially expressed in vSNc mDA neurons and whose expression is positively regulated by Pitx3 . Rgs6 belongs to the R7 subfamily of Rgs and functions as a GTPase activating protein to terminate signaling downstream of ligand-bound G-protein coupled receptors ( GPCR ) . It does so by accelerating the conversion from the active Gα-GTP bound state ( dissociated from Gβγ subunit ) to the inactive Gα-GDP bound state ( associated to Gβγ subunit ) . The R7 subfamily of Rgs regulators , including Rgs6 , are known to have preference for catalysis of pertussis toxin-sensitive Gi/o heterotrimeric G-proteins through recognition of their Gαi by a C-terminal Rgs protein domain [25] . Activated Gαi subunits inhibit adenylate cyclase such that cAMP production from ATP is halted and PKA/cAMP-dependent protein kinase pathways are inhibited . Conversely , the activated Gβγ subunit opens Girk channels to allow efflux of potassium ions outside the cell resulting in hyperpolarization [26] . The neuronal GPCRs previously associated with Gi/o proteins include dopamine receptors ( Drd2 , Drd3 ) , acetylcholine receptors ( m2 , m4 ) , GABAB receptor , metabotropic glutamate receptors ( mGluR2 , 3 , 4 , 6 , 7 , 8 ) . Thus in cerebellum and heart , phenotypes resulting from inactivation of Rgs6 are consistent with over-activation of signaling downstream of GABAB , serotonin 5-HT1A , M2 acetylcholine receptors , respectively . [26]–[29] . In the present work , we identified Rgs6 and investigated its role in the midbrain dopaminergic system . Rgs6 is shown to be preferentially expressed in vSNc mDA neurons and its knockout in mice results in progressive loss and alterations of Pitx3-positive mDA neurons specifically within the vSNc of aged animals . This late-onset degeneration is associated with markers of increased Drd2 signaling , down-regulation of Pitx3 expression and deregulated expression of its target genes , Aldh1a1 , Bdnf , Vmat2 , TH and Fgf10 . Further , the pattern of mDA degeneration observed in Rgs6−/− mice is a close phenocopy of DJ-1−/− mice suggesting that these two genes may act through similar pathways .
In order to identify genes responsible for the differential vulnerability of vSNc mDA neurons , we devised a strategy to isolate FACS-purified Pitx3-dependent and Pitx3-independent mDA neurons and compare their transcriptomes ( Fig . 1 ) . By birth , the SNc of Pitx3−/− pups is completely depleted of Pitx3-positive neurons but the dorsal Pitx3-negative neurons are spared [12] . After dissection of SN and VTA from midbrain slices of mice expressing TH-EGFP , FACS-sorting of TH-EGFP-positive neurons yielded a pure dSNc Pitx3-negative population from Pitx3−/− brains and mixed ( ∼80% Pitx3-positive and ∼20% Pitx3-negative ) SN mDA populations from wild-type animals . The comparison of their transcriptomes defined vSNc- and dSNc-enriched genes ( Fig . 2A ) . In VTA , Pitx3 deficiency leaves the 50% Pitx3-expressing mDA neurons intact at birth but they die within the next three months [12] . Comparison of VTA TH-EGFP cell expression profiles from Pitx3−/− and WT mice will thus identify genes which have Pitx3-dependent expression . RNA extracted from FACS-sorted cells ( Fig . 1 ) was used to generate probes for hybridization in duplicates to Affymetrix Mouse Gene 1 . 0ST microarrays and determination of expression profiles . Unbiased clustering of the 1813 differentially expressed genes ( fold changes >1 . 5 , p≤0 . 05 and signal ≥60 ) into seven clusters defined genes that are expressed in specific subsets of mDA neurons and/or that are Pitx3-dependent in VTA ( Fig . 2A ) . qRT-PCR analyses confirmed the expected enrichment for vSNc ( Girk2 , DAT ) , dSNc ( Calb1 ) and VTA ( Otx2 , Calb1 ) markers ( Fig . 2B ) . Further , many genes previously characterized for their subset-specific expression ( marked by stars in Fig . 2A ) validate the profiling data; these include Calb1/2 for dSNc , Otx2 for VTA , Kcnj6 ( Girk2 ) for vSNc , Lpl for VTA , Aldh1a1 for vSNc , Slc6a3 ( DAT ) for vSNc , Lix1 for SNc [20] , [30]–[32] . The complete list of genes in each cluster is provided in Table S1 . We chose to investigate genes enriched in vSNc and Pitx3-dependent . Rgs6 immediately appeared as an interesting candidate because it is known to negatively modulate signaling downstream of heterotrimeric Gi protein-coupled receptors through its instrinsic GTPase-stimulating protein activity [26] , [28] , [29] . The Rgs6 protein was detected by immunohistofluorescence in TH-positive SNc mDA neurons and not in VTA ( Fig . 3A ) . Most dSNc mDA neurons were negative for Rgs6 ( Fig . 3B upper left , arrowheads ) as were VTA cells ( Fig . 3B , bottom left ) . The SNc distribution of Rgs6 is very similar to that of Pitx3 but they differ in VTA ( Fig . 3B ) . Triple immunohistofluorescence staining against TH , Calb1 and Pitx3 ( Fig . 3C ) showed that the majority of Pitx3-positive vSNc mDA neurons are negative for Calb1 whereas dSNc mDA neurons are calbindin-positive . The SNc thus has two major subsets of mDA neurons with differential vulnerability to Pitx3 knockout: TH+/Pitx3+/Rgs6+/Calb− neurons in vSNc ( Pitx3-dependent , PD vulnerable ) and TH+/Pitx3−/Rgs6−/Calb+ cells in dSNc ( Pitx3-independent , PD resistant ) . In order to define the in vivo role of Rgs6 , we investigated the mDA system of Rgs6−/− mice by TH immunohistochemistry at 6 , 180 and 356 days of age ( Table S2 ) . Only 1 y-old Rgs6−/− midbrains were markedly different from controls and we could identify two major phenotypes in different mice . The first phenotype was a partial loss of SNc TH-positive cells on one side of the brain ( random unilateral ) ( Fig . 4A and Figure S1 ) . Quantification of TH-positive cells in SNc and VTA indicated a loss of about 35±6% ( SD ) in the SNc of the affected side ( Fig . 4B ) . The loss of TH-positive cells was further supported by fewer Nissl-positive cells in SNc of Rgs6−/− mice ( Fig . 4C ) . In addition , the Nissl stain revealed the presence of cells with abnormal elongated morphology that were found unilaterally within the SNc ( Fig . 4C ) . The loss of TH-positive cells is correlated with a loss of Pitx3-positive cells ( Fig . 4D ) . A second group of 1 y-old Rgs6−/− mice exhibited dysmorphic mDA neurons that displayed low levels of TH immunoreactivity ( THlow ) , aberrant morphology , pronounced cell shrinkage and disrupted TH-positive fiber network ( Fig . 5A and Figure S2 ) . These dysmorphic neurons were all localized in the vSNc , while mDA neurons in dSNc and VTA had normal appearance and unaffected Calb1 expression ( Figure S3 ) . We then determined whether the Rgs6−/− dysmorphic mDA neurons are undergoing degeneration by staining with Fluoro-Jade C ( FCJ ) , as previously shown in the MPTP or 6-OHDA-induced mouse PD models [33] , [34] and zitter mutant rats [34] . A high proportion of the dysmorphic mDA neurons stained positive for FJC ( Fig . 5B , C ) and are THlow . Cell counts indicated that degeneration is mostly unilateral and limited to the vSNc ( Fig . 5C ) . Some 180 days-old Rgs6−/− mice exhibited mild unilateral degeneration in the most lateral part of SNc , while newborn mice did not , supporting the progressive appearance of degeneration with age ( Table S2 ) . We then characterized degenerating mDA neurons in Rgs6−/− mice for the presence of pathologocial markers observed in PD and in other neurodegenerative diseases . Notably since degenerating neurons in PD suffer from oxidative stress and increased autophagocytosis due to the presence of protein aggregates , we assessed and observed increased expression of LC3B , a marker of activated autophagosome in degenerating neurons ( Fig . 5D ) . In order to further characterize the degenerating neurons , we verified expression of key genes implicated in development of familial forms of PD [5] . For example , DJ-1 and PINK1 are two genes whose mutated forms cause PD in an autosomal recessive manner and LRRK2 is the most frequently mutated gene causing PD and it acts in an autosomal-dominant fashion . Pink1 and Lrrk2 mouse null mutants do not however display degeneration of mDA neurons [5] , [35] but a rat model of Pink1 knockout does [8] . Surprisingly , the THlow neurons of Rgs6−/− vSNc specifically exhibit decreased DJ-1 expression , while Pink1 and Lrrk2 expression is increased in those same neurons ( Fig . 6 ) ; this contrasts with the fairly widespread expression of DJ-1 [36] , Pink1 and Lrrk2 [5] , [37] in midbrain neurons . Interestingly , the late-onset degeneration observed in Rgs6−/− midbrain appears to be a close phenocopy of the DJ-1−/− mice [7] , especially in terms of the initial unilateral nature of defects in aging SNc mDA neurons and the kinetics of cell loss . The loss of DJ-1 may thus contribute to Rgs6-dependent degeneration . Degenerating neurons were shown to overexpress the phosphorylated cell cycle inhibitor p27Kip1 in Alzheimer's disease [38] . Interestingly , Rgs6 was implicated in control of cell cycle and apoptosis [39] and we observed cytoplasmic accumulation of phospho-p27Kip1 only in degenerating THlow mDA neurons of one year-old Rgs6−/− mice ( Fig . 5E ) . In order to define molecular correlates of Rgs6-dependent degeneration , we assessed Pitx3 expression that was previously shown to have a survival function in SNc [12] . Nuclear Pitx3 was greatly diminished in the dysmorphic mDA neurons while cytoplasmic Pitx3 staining increased , suggesting a shift in sub-cellular localization observed with two different polyclonal Pitx3 antibodies in different cells of the same sections ( Fig . 7A ) . We assessed protein levels of some Pitx3 target genes by immunohistochemistry: these include Bdnf , Aldh1a1 ( Adh2 ) , TH , Drd2 , DAT , Vmat2 and Fgf10 [19] , [23] , [40] . The expression of Vmat2 and Bdnf is decreased in THlow neurons ( Fig . 7B , C ) . Expression of Aldh1a1 ( Adh2 ) is also decreased in degenerating vSNc mDA neurons ( Fig . 7D ) : this decrease may in part account for the low TH expression [19] . Our profiling data suggested that Fgf10 expression is repressed by Pitx3 in VTA and indeed , we found de-repression of Fgf10 expression only in THlow degenerating mDA neurons ( Fig . 7D ) . At post-natal day 6 ( not shown ) , we did not observe any change in expression of these genes consistent with a late onset phenotype . The regulatory action of Rgs6 was associated with various GPCRs , in particular the dopamine receptor D2 ( Drd2 ) [25] . Ventral SNc mDA neurons are subject to regulatory negative feedback mediated by Drd2 autoreceptors . Expression of Drd2 itself is not affected in vSNc mDA neurons of Rgs6−/− mice ( Fig . 8A ) despite its dependence on Pitx3 [19] . Dopamine signalling in these neurons leads to activation of Erk1/2 [41] and accordingly , we observed significant phospho-Erk1/2 only in THlow neurons of Rgs6−/− vSNc ( Fig . 8B ) . In addition , enhanced DA signalling downstream of Drd2 [42] would be expected to increase expression of the dopamine transporter DAT ( SLC6A3 ) which is otherwise dependent on Pitx3 . It was indeed observed that activated glycosyl-DAT is high in THlow neurons of Rgs6−/− vSNC ( Fig . 8C ) . Thus , high DAT would presumably increase intracellular DA levels in these neurons by promoting DA uptake [43] . Cytoplasmic DA would be further enhanced by the decrease of Pitx3-dependent [43] Vmat2 ( Fig . 7B ) which is responsible for sequestration of DA into vesicles . Thus , the combined elevation of DAT with decreased Vmat2 is very likely to maintain high levels of free DA that may be toxic [44] and contribute to the degenerative process and cell death [45] . Increased DA signalling thus constitutes a putative mechanism to explain the late-onset neurodegeneration observed in Rgs6−/− vSNc mDA neurons ( Fig . 9A ) . The two Rgs6−/− phenotypes appear to reveal incremental penetrance of similar defects: in this case , it would be expected that the contralateral midbrain of affected mice will be affected at some point . We thus further scrutinized the affected Rgs6−/− midbrains with the clear unilateral deficits described above and found two that exhibited on the contralateral side , small clusters of mDA neurons of relatively normal appearance but that are THlow , have cytoplasmic Pitx3 , high glycosylated DAT and enhanced phospho-Erk1/2 expression ( Figure S4 ) . Such cells were not observed in control midbrains . These data suggest that elevation of DA signalling subsequent to the loss of Rgs6 is closely associated with translocation of Pitx3 to the cytoplasm . In summary , the loss of Rgs6 is associated with three phenotypes each individually associated with mDA neuron degeneration , cell loss and Parkison's disease namely , 1 ) the loss of Pitx3 expression and of its target genes , 2 ) the loss of DJ-1 and finally 3 ) excessive dopaminergic tone .
A few gene expression profiling studies have compared gene expression in SNc versus VTA [46]–[48] . These studies identified large numbers of SN or VTA restricted genes but did not include criteria to relate these specificities to function . Our reliance on Pitx3 gene dependence to identify genes with preferential expression in Pitx3-independent dSNC versus Pitx3-dependent vSNc neurons allowed the prioritization of candidate genes based on the pro-survival Pitx3 gene . This approach also allowed definition of the unique expression profiles for dorsal compared to ventral SNc mDA neurons . We focused on vSNc-enriched and Pitx3-dependent genes in order to identify candidates for role ( s ) in vSNc mDA neurons survival . The list of 10 Pitx3-activated and 6 Pitx3-repressed genes in this subset includes only one known gene encoding a regulator of a signaling pathway , Rgs6 . Further , Rgs6 is the most dependent on Pitx3 for expression in VTA . The list also includes a transcriptional co-regulator Lmo3 that may contribute to Pitx3-dependent survival pathways but it's putative involvement in survival appeared less likely than Rgs6 to contribute to an age-dependent phenotype . Thus , the present study focused on characterization of neuronal loss and on molecular features of degenerating vSNc neurons as a result of Rgs6 inactivation . We observed late-onset degeneration of vSNc mDA neurons in Rgs6−/− midbrains ( Fig . 9B ) . This phenotype is detected at about 6 months of age and becomes more important in 1y-old mice . At that age , clear mDA cell loss is observed in a subset of mice ( Table S2 ) . It is likely that the degenerating THlow mDA phenotype ( Fig . 5 ) and the mDA cell loss ( Fig . 4 ) represent progressive steps of the same defects; accordingly , THlow neurons are also observed in the vSNc of midbrains with cell loss . This late-onset degeneration is similar to another mouse model of monogenic PD , the DJ-1 ( Parkin 7 ) mutant , that also initially exhibits unilateral defects [7] . Indeed , DJ-1−/− mice present unilateral loss of mDA cell bodies as early as 2 months after birth , with a transition to bilateral cell loss occurring at 1 year of age . Our study of Rgs6−/− mice showed unilateral cell loss , as evidence by decreased number of TH+ and NissL+ neurons ( Fig . 4A–C , Table S2 ) at 12 months after birth , while evidence of THlow degenerating neurons is readily apparent earlier at 6 months of age ( Fig . 5A , Table S2 ) . The comparison of phenotypes for DJ-1 and Rgs6 knockout mice indicates that: 1 ) they both have selective degeneration of SNc , but not VTA , mDA neurons , 2 ) they both have progressive degeneration and loss of mDA neurons , 3 ) degeneration begins unilaterally , 4 ) degeneration eventually becomes bilateral with earlier transition in DJ-1−/− than Rgs6−/− mice . A distinguishing feature of the Rgs6−/− model is the bias towards degeneration of Calb-negative vSNc mDA neurons compared to Calb-positive dSNc mDA neurons that remain largely unaffected ( Fig . 5A , Figure S4 ) , as is usually observed in PD . The slow degeneration of THlow vSNc neurons provided an opportunity to define the molecular features that accompany the dysmorphology . These neurons display increased expression of markers previously associated with pathological changes such as FluroJade C , LC3B and phospho-p27Kip1 . Moreover , degenerating THlow vSNc neurons show decreased DJ-1 and elevated Pink1 and Lrrk2 protein expression , suggesting a relationship between Rgs6 signaling and pathways implicated in PD pathology ( Fig . 6 ) . Future studies should address the relationships between DJ-1 , Pink1 and Lrrk2 in degeneration pathways of Rgs6−/− mice in terms of their known roles in mitochondria dynamics , calcium , balance , redox state and cell signaling . Collectively , the data show that Rgs6 signaling is necessary for maintenance of vSNc mDA neurons in the aging animal and that its downstream action may be mediated , at least in part , by Pitx3-dependent mechanisms ( Fig . 9A ) . Indeed , the THlow vSNc neurons exhibit low levels of Pitx3 and of its target gene products TH , Aldh1a1 , Bdnf , Vmat2 , together with enhanced expression of Pitx3-repressed Fgf10 ( Fig . 7 ) . The THlow neurons also exhibit cytoplasmic Pitx3 staining suggesting that there may be regulation of nuclear-cytoplasmic localization: this effect could be mediated through phosphorylation of Pitx3 as it was suggested that phosphorylated Pitx1 has greater affinity for nuclear DNA binding than its de-phosphorylated form [49] . One of the documented Pitx3-activated factors , Bdnf , is an important mediator of the neuroprotective action of Pitx3 during development and could contribute to trophic impairment in adult degenerating neurons . The degenerating neurons also exhibit a loss of determinants of postmitotic dopaminergic identity ( TH , Aldh1a1 , Vmat2 ) and this likely affects their neuronal activity . Aberrant neuronal activity is a hallmark [45] of the pre-symptomatic stage of PD and this likely transitions to major cellular disruptions ( proteosome dysfunction , mitochondrial integrity , calcium permeability… ) associated with degeneration and cell death . What could be the target of Rgs6 action ? One likely possibility is the dopamine receptor D2 ( Drd2 ) . Indeed , Rgs6 is a negative modulator of GPCR activity , including the dopamine Drd2 receptor [25] . Drd2 expression itself was not affected in degenerating Rgs6−/− mDA neurons ( Fig . 8A ) . However , the vSNc THlow neurons showed evidence of increased DA signaling , namely accumulation of phospho-Erk1/2 ( Fig . 8B ) and enhanced glycosylated dopamine transporter ( Slc6a3/DAT ) expression in the mutant ( Fig . 8C ) , consistent with a putative loss of negative Rgs6 input on dopamine signaling [50] . Since the Pitx3−/− midbrain exhibits decreased DAT and Drd2 [19] , the observed increase in DAT together with phospho-Erk1/2 are consistent with a primary action of Rgs6 inactivation on DA signaling . Rgs6 may thus contribute to the auto-regulatory negative feedback of the dopaminergic system and its absence may lead to dopamine-dependent oxidative stress and neuronal loss [2] , [50] . The enhanced phospho-p27Kip1 and FluoroJade staining support the interpretation that these cells are under stress . Alternatively , Rgs6 may have GPCR-independent actions: those could involve the GDNF pathway that is essential for catecholaminergic neuron survival [24] or involve direct action on apoptotic pathways [39] . An important aspect of the expression changes discussed above is that they only occur in THlow dysmorphic neurons that normally express Rgs6 and not in other mDA neurons of VTA and dSNc that are negative for Rgs6 ( Fig . 3 ) . Therefore , the concordance between Rgs6 midbrain expression and observed cell degeneration patterns suggests that the changes are cell-autonomous and directly related to Rgs6-dependent signaling operating in Rgs6+Pitx3+ vSNc neurons . We cannot however rule out the contribution of other brain systems affected by Rgs6 deficiency . Collectively , the data show that Rgs6 signaling is necessary for maintenance of vSNc mDA neurons in the aging animal and that its downstream action may be mediated , at least in part , by Pitx3-dependent mechanisms ( Fig . 9 ) . The present work identified a critical signalling pathway that controls survival of the mDA neuron subset that preferentially degenerates in PD . Further dissection of this pathway may lead to therapeutically useful insights on the unique properties of this group of mDA neurons .
All experimental procedures with laboratory animals were approved by the IRCM Animal Protection Committee and followed guidelines and regulations of the Canadian Council of Animal Care . All mice were maintained as heterozygous carriers in the C57Bl/6J background and maintained on a 12 h light-dark cycle with food and water ad libitum . Rgs6-null [26] and TH-EGFP [51] mice were described previously . Pitx3-null mice were generated in this laboratory [52] . Ventral midbrain dissections were performed on WT and Pitx3−/− newborn mice ( P1–P4 ) crossed onto TH-EGFP heterozygous background . Mouse brains were quickly washed in ice-cold PBS and then placed into cold Hibernate-A/1%B27 solution ( Gibco ) to dissect EGFP+ ventral midbrain ( vMB ) tissue under the fluorescence stereoscope ( Leica DFC300 FX ) . Tissue blocks of vMB were then further microdissected so as to separate SNc ( lateral ) from VTA ( medial ) . VTA and SNc tissue blocks were digested using the papain dissociation kit ( Worthington ) . Dissociated cells were then resuspended in warm Hibernate-A/1%B27 solution containing propidium iodide ( PI , 1 µg/ml ) , passed through 100 µm mesh and sorted by flow cytometry using the MOFLO™ instrument ( Beckman Coulter ) . PI−/EGFP+ live sorted cells were deposited in 30 µl of RNAlater solution ( Ambion ) ( max . of 3000 cells per 30 µl of RNAlater ) to preserve RNA integrity . Sorted EGFP+ cells in RNAlater were processed in batches of approximately 5000 cells for purification of total RNA using RNeasy Micro kit ( Qiagen ) . Briefly , 350 µl of RLT lysis buffer was added per 30 µl RNAlater-suspended cells . After vortexing for 1 min , 1 volume of 70% ethanol was added and the content loaded into single pre-equilibrated RNeasy MinElute column . This was done in duplicate for each of the four different preparations of EGFP+ cells ( Pitx3+/+ SNc , Pitx3+/+ VTA , Pitx3−/− SNc , Pitx3−/− VTA ) . Column-bound RNA was washed as recommended and eluted with 14 µl of RNAse-free water . Quality of total RNA was verified with the Agilent RNA 6000 Nano kit adapted for Agilent 2100 Bioanalyzer . For RT-qPCR , first-strand cDNA was synthesized using Superscript III RT enzyme and accompanying kit ( Invitrogen ) . Primers for PCR amplification are displayed in Table S3 . qPCR was performed using Perfecta reagents ( Quanta ) on a MX-3005 device ( Stratagene ) , and results were analyzed using the accompanying software . All quantifications were relative to Gapdh mRNA . For microarray hybridization , total RNA with minimal degradation was used . Prior to hybridization onto Affymetrix Gene1 . 0ST expression arrays , which was done at the Genome Quebec/McGill Innovation Centre , total RNA was linearly amplified using WT-Ovation Pico RNA Amplification kit ( NuGen ) . Gene expression summary values were computed by RMA Express ( Bolstad et al , 2003 ) and raw data was normalized with the LPE algorithm embedded in the FlexArray suite of programs ( Genome Quebec ) . Differentially expressed genes were chosen on the basis of their p-value≥0 . 05 , fold-enrichment ≥1 . 5 and raw array signal ≥60 . Hierarchical clustering and heat map display was done using Genesis ( Institute for Genomics and Bioinformatics , Austria ) Mice were anesthetized and perfused intra-cardially with fresh 4% paraformaldehyde/PBS buffer . Brains were collected and post-fixed for 24 h at 4°C . After inclusion in paraffin , brains were cut in 5–6 µm coronal sections using microtome and mounted on Superfrost Plus ( Fisher Scientific ) slides . Immunonohistochemistry was performed after paraffin removal and hydration through xylene and graded alcohol series . Antigen retrieval was performed in 10 mM sodium citrate ( pH 8 . 5 ) at 80°C water bath for 30 min . Sections in citrate solution were left to cool to room-temperature ( RT ) after which a step of endogenous biotin block was performed ( Streptavidin/biotin kit , Vector Labs ) . Blocking with 5% normal serum for 1 h at RT preceded primary antibody incubation ( overnight at 4°C ) . Primary antibodies used were against Pitx3 ( rabbit home-made , 1∶400 ) , Rgs6 ( rabbit home-made 1∶25 ) , Th ( Millipore MAB318 , 1∶1000 ) , Th ( Millipore AB152 , 1∶500 ) , Calb1 ( R&D Systems AF3320 , 1∶40 ) , Fgf10 ( Millipore ABN44 , 1∶1000 ) , Slc6a3 ( Santa Cruz sc-32258 , 1∶250 ) , Drd2 ( Santa Cruz sc5303 , 1∶100 ) , Aldh1a1 ( Abcam ab24343 , 1∶400 ) , BDNF ( Abcam ab108319 , 1∶25 ) , phospho-p27 ( Abcam ab32096 , 1∶50 ) , phospho-Erk1/2 ( Cell Signaling 4376 , 1∶25 ) , DJ-1 ( Abcam ab18257 , 1∶500 ) Lrrk2 ( Epitomics 3514-1 , 1∶50 ) , Pink1 ( Novus Biologicals BC100-494 , 1∶50 ) , LC3B ( Cell Signaling 3868 , 1∶50 ) , and Vmat2 ( Millipore AB1598P , 1∶200 ) . Secondary antibodies were either biotinylated ( Vector Labs , 1∶250 ) or directly coupled to fluorochromes such as AlexaFluor 488/546 ( Invitrogen , 1∶250 ) . For ABC method of amplification , AlexaFluor350/480/546- or HRP-conjugated ( PerkinElmer NEL750 , 1∶1000 ) streptavidin compounds were used . For immunoperoxidase staining , 1% hydrogen peroxide treatment was done just after antigen retrieval step . Mounting of sections was in Mowiol ( Sigma ) ( fluorescence ) or in Permount ( Fisher Scientific ) ( DAB chromogen reaction ) . Immunofluorescence sections were observed on Leica DM6000B light microscope and Carl Zeiss LSM700 confocal microscope . FJC ( Molecular Probes , 0 . 01% stock in water ) staining was performed just after the immunohistochemical procedure [33] , [34] . Briefly , sections were washed in water , dipped for 5 min in 0 . 06% of potassium permanganate , rinsed again in water and placed in 0 . 0001% FJC/0 . 1% acetic/0 . 0001% DAPI solution for 10 min . Finally , slides were rinsed in water and mounted in 0 . 1% acetic acid/80% glycerol ( v/v ) . The Fluorescein/FITC filter system was used to visualize FJC . Nissl staining was performed on deparaffinized sections by immersion into warm 0 . 1% cresyl violet solution for 10 min , rinsing three times in distilled water and differentiating in 95% ethanol . Slides were then dehydrated in 100% alcohol , cleared in xylene and mounted with Permount . Cell counts were performed using ImageJ ( National Institutes of Health ) . For cell counts of degenerating neurons , TH-stained or TH/FJC/Dapi-stained coronal sections were loaded on ImageJ; the sections spanned regular intervals ( 30 or 100 µm ) across rostro-caudal extent of midbrain of WT ( n = 2 ) and 1-yo Rgs6−/− mice ( n = 2 ) . For each section , total numbers of TH+ , TH+/FJC+ and TH−/FJC+ cells were separately counted for SNc and VTA in both hemispheres . The percentage of TH+ degenerating neurons for each anatomical region reflects the ratio between the total number of TH+/FJC+ events and the total number of TH+ events for all rostro-caudal series . Values are reported as means +/−S . D . Statistical significance was calculated using Student T-test .
|
The locomotor deficits associated with Parkinson disease result from the death of a specific subset of dopamine neurons in the ventral part of the midbrain . The reason for the greater sensitivity to degeneration of those , relative to other , neurons is not clear . Prior work showed that the Pitx3 transcription factor is specifically expressed in these neurons where it has a survival role during development . The present work identified a cell signaling component , Rgs6 , that is also restricted to the sensitive neurons in the midbrain and that exerts a protective function , particularly late in life . While the loss of Rgs6 function may predispose or contribute to Parkinson disease , its stimulation may provide a novel therapeutic avenue to treat Parkinson disease .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"biology",
"and",
"life",
"sciences",
"medicine",
"and",
"health",
"sciences"
] |
2014
|
Rgs6 is Required for Adult Maintenance of Dopaminergic Neurons in the Ventral Substantia Nigra
|
Merkel Cell Polyomavirus ( MCPyV ) is associated with Merkel Cell carcinoma ( MCC ) , a rare , aggressive skin cancer with neuroendocrine features . The causal role of MCPyV is highly suggested by monoclonal integration of its genome and expression of the viral large T ( LT ) antigen in MCC cells . We investigated and characterized MCPyV molecular features in MCC , respiratory , urine and blood samples from 33 patients by quantitative PCR , sequencing and detection of integrated viral DNA . We examined associations between either MCPyV viral load in primary MCC or MCPyV DNAemia and survival . Results were interpreted with respect to the viral molecular signature in each compartment . Patients with MCC containing more than 1 viral genome copy per cell had a longer period in complete remission than patients with less than 1 copy per cell ( 34 vs 10 months , P = 0 . 037 ) . Peripheral blood mononuclear cells ( PBMC ) contained MCPyV more frequently in patients sampled with disease than in patients in complete remission ( 60% vs 11% , P = 0 . 00083 ) . Moreover , the detection of MCPyV in at least one PBMC sample during follow-up was associated with a shorter overall survival ( P = 0 . 003 ) . Sequencing of viral DNA from MCC and non MCC samples characterized common single nucleotide polymorphisms defining 8 patient specific strains . However , specific molecular signatures truncating MCPyV LT were observed in 8/12 MCC cases but not in respiratory and urinary samples from 15 patients . New integration sites were identified in 4 MCC cases . Finally , mutated-integrated forms of MCPyV were detected in PBMC of two patients with disseminated MCC disease , indicating circulation of metastatic cells . We conclude that MCPyV molecular features in primary MCC tumour and PBMC may help to predict the course of the disease .
Polyomaviruses are small , non enveloped double stranded DNA viruses which infect many species with a restricted host range . The initial discovery of Murine polyomavirus ( MPyV ) and Simian vacuolating 40 ( SV40 ) was closely linked to the demonstration of their experimental tumorigenic properties [1] . Infections by the human polyomaviruses BK ( BKPyV ) , JC ( JCPyV ) , and the recently identified KI ( KIPyV ) and WU ( WUPyV ) are highly prevalent in most populations [2] , [3] . Polyomaviruses persist latently in the host and may reactivate , causing disease in the immunocompromised [4] , [5] , but have not been firmly associated with cancer in humans [6] . Therefore , the discovery in a rare but aggressive skin cancer , Merkel Cell Carcinoma ( MCC ) , of a fifth human Polyomavirus , named Merkel Cell Polyomavirus ( MCPyV ) has raised new interest in the oncogenic potential of human Polyomaviruses [7] . MCPyV DNA was shown to be monoclonally integrated into most MCC , and tumour cells were found to express the major viral oncoprotein , large T antigen ( LT ) [8] . Remarkably , MCPyV present in MCC tissue exhibited a molecular signature , consisting of mutations which truncate LT and suppress its helicase domain required for viral replication [9] . These features are similar to molecular defects observed in MPyV [10] and bring strong evidence for a causative role of the virus in MCC . MCC is a carcinoma of neuroendocrine cells which affects principally elderly and immunocompromised patients . The close association between MCPyV and MCC has been confirmed in several case series worldwide [11] , [12] , [13] , [14] , [15] , [16] , [17] , [18] , [19] , [20] , [21] , [22] . However , MCPyV has been detected in normal skin [23] , benign skin lesions [24] and non MCC skin cancer [25] , and also found in various tissues such as the oral cavity [26] , gastrointestinal and urogenital tracts [23] , tonsils [27] , respiratory tract [28] , blood [9] , and rarely in non skin cancers [19] , [25] , [29] , [30] . Seroepidemiologic studies have revealed that most healthy individuals are infected with MCPyV , as with other human Polyomaviruses [31] , [32] , [33] , [34] , [35] . Therefore , numerous questions related to the persistence and replication of the virus in the host and mechanisms of oncogenesis remain unanswered . We investigated whether MCC patients' clinical data were related to the presence of MCPyV and specific molecular features in tumour and non tumour tissues . In a cohort of 33 MCC patients , we collected MCC and non MCC samples , and compared the frequencies of MCPyV DNA detection , viral load and nucleotide sequences . We looked for integration of MCPyV in MCC cells and for the presence of integrated/mutated forms of MCPyV in non tumour samples .
This study was conducted according to the principles expressed in the Declaration of Helsinki . According to French regulation , information was delivered to the patients on research performed on biological samples , and written informed consent was obtained for participation in the study , which was approved by the institutional review board of the Comité de protection des personnes Ile de France 3 . All patients with Merkel cell carcinoma who attended the Dermatology Departments of Cochin and Bichat hospitals from March 2008 to April 2010 were prospectively included in the study . Histopathologic confirmation of MCC diagnosis relied on tumour morphology consistent with MCC on hematoxylin-eosin-stained tissue sections , paranuclear dot immunostaining pattern for CK20 or positive immunostaining for synaptophysin and chromogranin A . Clinical data were retrieved from hospital case records , and included sex , age at diagnosis , site and size of primary MCC , and stage of the disease at diagnosis , according to Allen's classification [36] based on primary tumour diameter ( <2 cm = I , >2 cm = II ) and the presence of metastasis ( regional lymph node metastasis = III , distant = IV ) . Samples from primary and/or metastatic MCC lesions were recovered and included retrospectively retrieved formalin-fixed paraffin-embedded ( FFPE ) sections and/or fresh-frozen specimens for patients newly diagnosed or who relapsed . FFPE sections from non MCC cancer tissue were retrieved if possible . In addition , blood and/or nasal swab and/or urine were sampled in patients at inclusion and follow-up visits when possible , and disease stage and status were recorded . Status was defined according to the presence or absence of tumour ( primary and/or metastatic ) as alive with disease ( AWD ) or in complete remission ( CR ) respectively . Date and cause of death were recorded . For FFPE samples , two consecutive 10 µ-thick sections were retrieved . Frozen fragments were cut and mechanically dissociated . DNA was extracted using the QIAmp DNA tissue extraction kit ( Qiagen ) according to the manufacturer instructions . Nasal swabs discharged in 200 µl sterile saline buffer and urine were immediately frozen and kept at –20°C until analysis . PBMC were obtained by Ficollpaque centrifugation ( Eurobio ) of 3 mL EDTA blood , suspended in 200 µl of PBS , immediately frozen and kept at –20°C until analysis . DNA was extracted using the QIAmp blood DNA extraction kit ( Qiagen ) . For tissue material , lysis was extended to 24 h . DNA was eluted in 200 µl elution buffer ( Macheret ) , and concentration was measured by UV-spectrometry ( NanoDrop Technology , Willmington ) . To detect MCPyV sequences , a sensitive real time PCR assay was designed using primers which encompass a 91 bp fragment of the LT oncoprotein gene , located upstream of the Rb-binding encoding sequence ( Table S1 ) [7] . Amplification was performed on 100 ng DNA with 300 nM each primer and 100 nM probe in 50 µl Taqman Master Mix ( Applied Biosystems , Courtaboeuf ) . After a 10 min denaturation step at 95°C , cycling conditions consisted of 50 cycles of 15 sec at 95°C and 1 min at 60°C on an ABI 7500 platform ( Applied Biosystems ) . To check for carryover between the samples , mock samples were included in each series . Confirmation LT3 and VP1 PCR assays were performed as described [7] . Absolute quantification of MCPyV viral load in patients' samples was obtained by establishing a calibration curve with ten-fold serial dilutions ( from 10 to 106 copies ) of known concentrations of plasmid MCCIC13 which contains 1 copy of the MCPyV genome inserted in the pCR-XL-TOPO vector [18] . To quantify MCPyV DNA in copy number per cell in MCC and non MCC cancer samples , the housekeeping apolipoprotein B gene ( hapb ) [37] was amplified by real time PCR and the delta-delta Ct method was used . To detect BKPyV and JCPyV sequences , real time PCR assays were designed to amplify a 79 and a 110 bp fragment of the LT antigen coding sequence in BKPyV and JCPyV genome respectively ( Table S1 ) . Amplification was performed using the same cycling and control conditions as above . To characterize integration of MCPyV , we used the DIPS-PCR technique which amplifies junctions between the host and MCPyV genomes as previously described [18] , [38] . Briefly , DNA extracted from MCC tumours was digested overnight with Taq I ( Invitrogen ) , an enzyme that does not cleave the MCV350 sequence ( GenBank accession number EU375803 ) , and ligated to adapters . Ligated fragments were subjected to 40 cycles of linear amplification with forward primers sequentially designed along the MCPyV LT gene ( Table S1 ) , followed by a 30 cycle exponential amplification using internal primers coupled to a reverse adapter-specific primer . PCR products were visualized and cut from agarose gel , purified with QIAmp gel extraction kit ( Qiagen ) , and directly sequenced using Big-Dye terminator DNA-Sequencing technology ( Applied Biosystems ) . To discriminate integrated versus non integrated forms of the MCPyV genome in patients' samples , we amplified short fragments bracketing each of the characterized integration sites . DNA ( 100 ng ) was subjected to PCR using a forward MCPyV-specific primer located 100 to 300 bp upstream of the integration site , and a reverse downstream primer , specific either to the human integration locus or to the MCPyV non-integrated genome ( Table S1 ) . PCR was performed in 50 µl Ampli Taq Gold Master Mix ( Applied Biosystems ) for 50 cycles with conditions specific to each primer set . Overlapping fragments of the MCPyV LT encoding gene were amplified using four primer pairs in order to cover the whole second exon of LT ( base pair number 861 to 3080 according to the MCV 350 genome ( Table S1 ) . PCR was performed on 100 ng DNA by using HotStartTaq Master Mix kit ( Qiagen ) containing 0 . 5 µM primers in a final volume of 50 µL . PCR products were then directly sequenced as above . The cumulative rates of survival in complete remission relative to MCPyV load in primary MCC ( <1 or ≥1 copy per cell ) , and overall survival relative to the presence or absence of MCPyV DNA in PBMC were estimated by the Kaplan-Meier method . Patients with follow-up under 3 months were excluded from analysis . Survival in complete remission was calculated from the date of diagnosis to the date of first tumour recurrence in patients AWD or died of disease ( DOD ) at last follow-up . Patients in complete remission were censored at their last follow-up visit . Analyses were performed using the XLStat software ( Addinsoft , Paris , France ) . Multivariable analysis of survival in complete remission was done by using a stepwise Cox proportional hazards model that used forward covariate entry to the model . The proportions of patients either CR or AWD , or at stages I or II or III versus IV who had either a MCPyV negative or MCPyV positive PBMC sample were compared using the exact Fischer test . All P values were two-sided , and P values less than . 05 were considered statistically significant .
Thirty nine patients with MCC attended the Dermatology Departments of Bichat and Cochin hospitals . Six patients without retrieved MCC material were excluded from the study . The remaining 33 patients included 14 males and 19 females ( sex ratio = 0 . 6 ) . Their median age at diagnosis was 77 years ( range 39–88 ) . Four patients were immunocompromised , because of corticoid therapy for rheumatoid arthritis , hepatic transplantation , lymphopenia and recurring hairy cell leukaemia . Thirteen ( 39% ) patients had a history of cancer other than MCC ( non MCC skin cancer and/or non skin cancer ) ( Table S2 ) . Primary MCC was localized to the limbs , head , and trunk in 21 ( 64% ) , 11 ( 33% ) and 1 ( 3% ) cases respectively . MCC median diameter was 25 mm ( range 7–70 mm ) . At diagnosis , patients were at Allen's stages I , II , III and IV in 9 ( 27% ) , 16 ( 48% ) , 7 ( 21% ) and 1 ( 3% ) cases respectively [36] . The median delays from diagnosis until inclusion and last follow-up were 7 months ( up to 112 months ) and 16 months ( up to 134 months ) respectively . At last follow-up , 18 ( 54% ) patients were in CR , 8 ( 24% ) patients were AWD and 7 ( 21% ) patients had died of disease ( DOD ) ( Table 1 ) . We analyzed 43 MCC samples consisting of 26 primary MCC ( 15 fresh-frozen and 11 formalin fixed paraffin-embedded ( FFPE ) specimens ) , 14 skin metastasis ( 12 fresh-frozen and 2 FFPE specimens ) and 3 fresh-frozen regional lymph node metastasis samples . Viral DNA was detected in 41/43 samples from 31/33 patients , with a median viral load quantified in 37 samples of 3 copies per cell ( range 3 . 10−3 to 3 . 103 ) . Negative results observed in one FFPE section and one fresh frozen sample from primary MCC were confirmed using the previously described LT3 and VP1 PCR assays [7] . In twenty four patients with follow up greater than three months , viral load in the primary tumour was analyzed with respect to survival . Median survival in complete remission was longer in patients who had ≥1 copy per cell ( n = 15 ) , than in patients who had no detectable viral DNA ( n = 2 ) or <1 copy per cell ( n = 7 ) ( 34 months , 95% CI = 26 to 42 vs 10 months , 95% CI = 7 to 14 , Kaplan Meier log-rank P = 0 . 037 ) ( Figure 1 ) . Among clinical parameters analyzed ( sex , age , limb site and size of primary tumour , and presence or absence of lymph node metastases at diagnosis ) , which didn't differ in the two groups , only female sex was associated with a better outcome . Adjusted for sex , the relative hazard for survival in complete remission was 4 . 8 ( 95% confidence interval 0 . 90–26 , P = 0 . 066 ) with primary tumour containing more than 1 copy per cell . We then asked if , in MCC patients , MCPyV was restricted to MCC tissue . Viral DNA was detected in 27/28 ( 96% ) nasal swabs from 21/21 patients , with a median load of 3 . 103 copies per sample ( range 5–2 . 106 ) . MCPyV DNA was also found in 22/38 ( 58% ) urine samples from 18/28 ( 64% ) MCC patients , with a median load of 6 . 102 copies/ml ( range 100–4 . 105 ) . In addition , MCPyV was amplified from 20/49 ( 41% ) PBMC samples from 15/30 ( 50% ) patients with a median load of 102 copies per ml whole blood ( range 10–5 . 104 ) . MCPyV DNA was detected in 1/6 FFPE non MCC cancer samples from 5 patients ( Table S2 ) . We wondered whether the high rate of detection of MCPyV was common to other human Polyomaviruses . BKPyV and JCPyV DNA were amplified from 9% each of nasal swabs , 31% and 7% of urine samples , and 3% and 6% of PBMC respectively . Since MCPyV DNA was detected infrequently in urine and in the PBMC of about half of the patients , we asked whether MCPyV DNAuria and/or DNAemia were linked to the stage and/or evolution of the disease . No correlation between MCPyV detection in urine was found with any of these parameters . In contrast , detection of MCPyV in PBMC was more frequent in patients sampled alive with disease ( AWD ) than in patients in complete remission ( CR ) ( 18/30 or 60% versus 2/19 or 11% , P = 0 . 00083 ) . Moreover , among AWD patients , MCPyV tended to be more frequently detected in patients with distant metastasis than at less advanced stages of the disease ( 6/6 MCPyV positive PBMC when sampled at stage IV versus 12/24 or 50% when sampled at stages I , II or III , P = 0 . 057 ) . In twenty eight patients with follow up greater than three months , the detection of MCPyV in PBMC was associated with poorer outcome , since patients with at least one positive sample had a median survival of 28 months ( 95% IC = 19 to 36 ) whereas all patients with no detectable MCPyV survived after a median follow-up of 71 months ( Kaplan Meier log-rank P = 0 . 003 ) ( Figure 2 ) . All clinical parameters analyzed were comparable in the PBMC-positive and PBMC-negative groups except sex , since significantly more male patients had MCPyV-positive PBMC ( P<0 . 009 ) . Among these parameters , only primary tumour size above 2 cm was associated with higher risk of death . We then looked at PBMC results according to viral load in primary MCC . Among 15 patients with ≥1 copy per cell , 4/4 patients who relapsed had positive PBMC , compared with 3/11 disease-free patients . Among 9 patients with <1 copy per cell , 8 had a PBMC sample tested . Two of four patients who relapsed tested positive while two with MCPyV-negative tumours tested negative . The four patients who were disease free at last follow-up had negative PBMC . Using the DIPS-PCR method , we demonstrated the integration of MCPyV DNA in six MCC cases . We first confirmed viral integration in metastatic tissue from two patients , and found virus-host genome junctions identical to those previously reported in their primary tumours [18] . In three new cases , integration of MCPyV was found to interrupt the second exon of LT downstream from the Rb binding coding sequence , whereas it interrupted the 3′ end of the VP1 gene in another one . Integration was located on four distinct chromosomal loci , next to or into known human genes ( Table 2 ) . Two of them , PARVA and DENND1A genes , encode proteins involved in cell junctions and in formation of clathrin coated vesicles or cell adhesion and cytoskeleton organization respectively . A third gene , TEAD1 , encodes a transcriptional activator reported to be used by the SV40 enhancer to activate expression of the early T oncoprotein gene [39] . Finally , in one case , integration resulted in fusion of the MCPyV LT gene and successive truncated fragments of the seventh and the tenth introns of the GMDS gene , two regions separated by approximately 200 kb in the human genome , demonstrating that large rearrangements occurred . We sequenced the whole second exon of LT gene in MCC and non MCC tissues . Fifty-two sequences from 26 patients displayed >99% homology with prototypes MCC350 , MCC339 , MKL-1 and TKS published sequences [7] , [9] . To characterize strain-specific and/or tumour-specific markers , a local consensus reference was constructed by alignment of all sequences , and variations were indicated as silent , non synonymous or stop mutations , deletions or insertions . The total number of silent and non synonymous mutations in MCC and non MCC samples with respect to the consensus differed by three and sixteen fold respectively . Single nucleotide polymorphisms ( SNPs ) characterized strain specific signatures in eight patients , and a common single silent mutation was identified in three other patients ( Figure 3A ) . Fourteen MCC sequences ( 13 complete and 1 partial ) were obtained from 12 patients . In two patients , we verified that sequences from distinct metastasis exhibited 100% homology . Mutations which truncate LT were characterized in nine cases , consisting in stop mutations ( 5 cases ) , or deletions causing frameshifts which generate stop codons ( 3 cases ) ( Figure 3A ) . In the remaining case , a 250 bp insertion of the third intron of the DENND1A gene was lying inside LT , upstream the VP1-host genome junction identified in the second intron of the same human gene . Overall , mutations preserved the Rb fixation domain but inactivated the helicase domain of the oncoprotein ( Figure 3B ) . In one additional patient , we failed to amplify the 3′ end 600 last bp of LT , despite a high viral load and repeated attempts using different sets of primers , suggesting a truncation of this sequence . Finally , in the last two cases , full length sequences were obtained and encoded a non truncated protein . Interestingly , in 5 MCC cases where the DIPS-PCR characterized integrated-truncated LT , a complete LT sequence was amplified downstream the truncation site , suggesting the coexistence of integrated concatemers or latent episomes of the MCPyV genome and truncated-integrated viral sequences . Then , we analyzed 16 nasal and 8 urine sequences . All sequences were complete except three obtained from weak positive samples , probably reflecting sensitivity limits of the method used rather than truncations . Two sequential nasal sequences from the same case showed 100% homology . All full-length sequences were wild-type . None of the 9 non MCC samples from 7 patients who displayed a truncated LT in MCC harboured the tumour-specific molecular signature ( Figure 1A ) . These results suggest that nasal swabs or urine are likely to contain MCPyV DNA from either excreted or episomal virus . PBMC sequences were obtained in 14 cases , including partial sequences obtained from 7 weak positive samples . Complete PBMC and MCC sequences from 4 patients were compared ( Figure 1A ) . In one case , both PBMC and MCC sequences encoded intact LT antigen open reading frame ( ORF ) . In another case , the premature stop codon observed in MCC LT was absent from MCPyV recovered from PBMC . In contrast , in two other cases , the specific MCC signatures ( a 5 and a 25 bp deletion respectively ) were recovered from the two patients' PBMC samples . Notably , in one of these patients , sequences from nasal swabs and urine were also analyzed and didn't harbour the tumour signature . Since the two patients presented disseminated metastatic lesions at the time of sampling and were both DOD at the end of the study , we assume that the presence of mutated MCPyV DNA in PBMC reflected circulation of metastatic MCC cells . To further confirm this last hypothesis , we amplified a portion of the LT gene bracketing the predicted integration site and the viral-host junction sequence in tumour and non tumour samples of five patients . Virus-host junction sequences were amplified from MCC tissue in all cases , from PBMC in two cases , but from neither urine nor nasal swabs . Both integrated and non integrated products were amplified from MCC samples , further suggesting the coexistence of integrated-truncated viral sequences and integrated concatemers or episomes in MCC samples . Altogether , these results suggest that MCPyV sequences recovered at peripheral urine and respiratory sites merely correspond to free excreted virus or episomal DNA , whereas the presence in the patients' PBMC of tumour-like sequences argued for the presence of circulating tumour cells .
Since the discovery by Feng et al who identified MCPyV in 6/8 MCC [7] , several large studies have demonstrated that MCPyV is associated with most cases of MCC except in Australia [11] , [12] , [13] , [17] , [19] , [22] , [24] , [26] , [33] . We detected MCPyV in MCC from 31/33 patients . Negative results from FFPE specimens could be due to poor conservation of DNA . Only one fresh frozen tumour tested negative . We estimated MCC median viral load at 3 copies per cell , with a 6 log variation between samples , consistent with other reports [13] , [18] , [22] , [40] , [41] , [42] , [43] . Variations may be due to tissue quality , the proportion of non tumour cells in samples , or mutations in the target viral sequence . However , variations in rates of LT expressing MCC cells were also reported [8] , [43] . The fact that some MCC cases do not contain MCPyV DNA nor express LT suggests that MCC is a heterogeneous disease with at least two etiologies , despite a lack of phenotypic markers able to distinguish between MCPyV positive and MCPyV negative cases [8] , [44] . Interestingly , patients with MCC containing at least 1 copy of viral genome per cell had better outcome than patients with lower values of MCPyV . Although the low number of patients studied impairs definite conclusions , it is striking that two previous studies also reported poorer survival rate in patients with the lowest viral DNA load and LT expression in MCC [22] , [40] . Although the mechanisms of MCC pathogenesis are unknown , variations in MCPyV load and in patients outcome further argue for heterogeneity and variable implication of the virus in the disease [11] , [12] , [16] , [21] , [45] , as previously described in HPV related and unrelated carcinoma [46] . Several observations support the causal role of MCPyV in most MCC . In particular , cell transformation by MCPyV was shown to depend on LT , as in other polyomaviruses-induced oncogenesis [1] . First , MCPyV LT is able to bind and sequester the tumor-suppressor protein Rb through a conserved LxCxE motif [9] . Second , the transformed phenotype of MCPyV-positive MCC cell lines depends on LT expression , since cells undergo growth arrest and/or death upon LT silencing [47] . However , in two models of adenovirus and polyomavirus-induced oncogenesis , the dependence of transformed cells on viral oncoproteins was reversed upon time , since cells conserved an oncogenic phenotype while viral expression was shut-down in one case and viral sequences were lost in the other [48] , [49] . These findings suggest that transformed cells acquire subsequent genetic alterations which circumvent their need for a continued expression of viral oncoproteins . Therefore , more cellular genetic alterations may be necessary in virus-unrelated than in virus-related oncogenesis . In this respect , the number of chromosomal alterations and amplifications was significantly higher in HPV-unrelated than in HPV-related carcinomas , and correlated with unfavourable prognosis [46] . Recurrent genomic changes have been described in MCC [16] , [18] , but their link with MCPyV has not yet been extensively investigated . Monoclonal integration of MCPyV is viewed as a key element in oncogenesis . We identified in metastatic lesions from two patients the same virus-host genome integration characteristics previously described in their primary tumours , sustaining the hypothesis that viral integration constitutes an early event in MCC oncogenesis [7] , [18] . We also showed the integration of truncated LT in four cases , and in one of these this led to a complex rearrangement between LT , VP1 and the target human gene . All chromosomal integration sites identified so far differ from each other [7] , [18] . We are currently verifying whether expression of putative target human genes , located in the vicinity or at the site of integration , is modified in tumour cells , notably TEAD1 which was reported to be used by another Polyomavirus , SV40 , as a transcriptional enhancer factor . In addition , MCC LT sequences revealed various point or frameshift mutations which preserve the Rb binding domain but truncate the oncoprotein before the helicase domain , as in the tumour-specific molecular signatures previously described [9] , [50] . Such truncations preserve the transformation ability of LT through Rb sequestration , but prevent viral DNA replication . A similar loss of full-length LT has been observed in vitro and in vivo in models of SV40 and MPyV-induced carcinogenesis [51] , [52] In addition , replication-defective polyomaviruses with loss of LT binding to the origin of replication showed enhanced transforming properties [53] . Our results extend previous observations and reinforce the hypothesis that acquisition of mutations within LT is a common feature and may be a prerequisite for carcinogenesis induced by polyomaviruses . However , in three cases of this series and in two previously reported cases , mutations truncated LT upstream an identified nuclear localization signal , which could prevent nuclear expression of the protein [9] . Lastly , mutations in LT were not observed in all cases in this nor in other studies [43] , [54] . We can't exclude that these cases display mutations at other sites critical for MCPyV replication . A point mutation in a pentanucleotide sequence of the replication origin was observed in a MCC strain and prevented replication [55] . Finally , the fact that the full length second exon of LT was sequenced in five MCC samples although integration interrupted LT suggests that , as previously observed with Southern Blot analysis [9] , truncated/integrated and probably whole genomic copies of MCPyV coexist in tumour cells , as confirmed by PCR assay which discriminates integrated versus non integrated MCPyV genomes . The lifecycle of MCPyV in the host is unknown . Serological studies showed that infection is common in the general population and occurs before the third decade [33] , long before development of MCC . Routes of transmission and sites of excretion are not completely known . We showed presence of MCPyV in the respiratory tract of most MCC patients , in serial samples drawn at a several-month interval , in contrasts with low detection rate ( below 17% ) in non MCC patients reported in the literature and observed with our own detection method ( data not shown ) [4] , [27] , [28] , [50] , [56] , [57] , [58] . MCPyV DNA excretion in urine , which was previously reported in one MCC case [59] , was observed in almost half of patients , above rates ( below 25% ) reported in control subjects [23] , [26] . Comparative LT sequencing from MCC and non MCC samples revealed strain-specific SNPs . Whereas most MCC sequences displayed tumour-specific molecular signatures , all nasal swabs and urine sequences were wild-type , suggesting that the latter correspond to excreted or episomal virus , whereas the former belong to integrated genomes . Thus , high rates of MCPyV excretion both in the respiratory tract and urine may be a hallmark of MCC patients . Urine excretion of BKPyV or JCPyV is frequent in immune competent subjects and increases with age , during pregnancy or immune suppression [60] , [61] . Since excretion rates of BKPyV and JCPyV were comparable in MCC patients to rates of non MCC patients [61] , [62] , we hypothesize that patients present a specific failure to control latency of MCPyV but not all Polyomaviruses . This hypothesis is supported by the fact that high levels of antibodies directed towards the major viral capsid protein VP1 of MCPyV but not other human Polyomaviruses were more frequently observed in MCC patients than in the general population [31] , [34] . Indeed , in the case of BKPyV and JCPyV infection , reactivation and active shedding were positively correlated with serum antibody responses to VP1 [63] , [64] . Lastly , our results show evidence of high rates of MCPyV DNA in MCC patients' PBMC , in contrast with low rates ( 0–8% ) reported in the serum , whole blood or PBMC of non MCC subjects [17] , [19] , [27] , [29] , [50] . Moreover , MCPyV DNA detection in PBMC was significantly correlated with the disease stage and outcome since patients with at least one PBMC positive sample had shorter survival in remission that patients in whom MCPyV had not been detected in any PBMC sample . In one patient , MCPyV recovered from PBMC had a wild-type genotype whereas the viral genome recovered from MCC had a LT truncating mutation . We hypothesize that MCPyV DNAemia may correspond to active viral replication following reactivation , as observed with other human polyomaviruses [2] , [61] . Indeed , MCPyV DNA detection was reported in activated circulating monocytes of one MCC patient and one control [59] . In two patients in our study however , viral sequences recovered from PBMC displayed the patient's MCC-specific molecular signature . As both patients were sampled at a metastatic stage and subsequently died of their disease , we believe that PBMC viral DNA revealed metastatic circulating cells , since MCC cells were previously identified in the peripheral blood of one MCC patient [65] . Altogether , our results provide new insights in the life cycle of MCPyV during MCC pathogenesis . The low number of cases studied might weaken the statistical power of our results . However , we suggest that quantitative and qualitative molecular analysis of MCPyV in tumour and non tumour sites of MCC patients may be a useful tool to characterize their disease stage and manage their follow-up . We are currently designing a prospective study to confirm these results in large series of patients .
|
Merkel cell polyomavirus ( MCPyV ) is a recently discovered virus highly associated with a rare skin cancer , Merkel cell carcinoma ( MCC ) . The causal role of MCPyV in cancer is suggested by integration of viral sequences into the cell genome and by a specific molecular signature . We looked for and compared molecular species of MCPyV in tumour and non tumour samples of 33 MCC patients . We showed that a tumour viral load greater than 1 copy per cell was associated with a better outcome , and that detection of the virus in blood but not in urine correlated with a shorter overall survival . A tumour–specific molecular signature was found in the blood of two patients with metastatic disease , but did not occur in their respiratory nor urine samples . We propose that molecular analysis of MCPyV in tumour and blood be used as a biomarker of infection and cancer progression in MCC patients .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"Methods",
"Results",
"Discussion"
] |
[
"dermatology/skin",
"cancers,",
"including",
"melanoma",
"and",
"lymphoma",
"oncology/skin",
"cancers",
"virology/viruses",
"and",
"cancer",
"virology/emerging",
"viral",
"diseases"
] |
2010
|
Distinct Merkel Cell Polyomavirus Molecular Features in Tumour and Non Tumour Specimens from Patients with Merkel Cell Carcinoma
|
Breast cancer is the most common cancer in women in developed countries , and the contribution of genetic susceptibility to breast cancer development has been well-recognized . However , a great proportion of these hereditary predisposing factors still remain unidentified . To examine the contribution of rare copy number variants ( CNVs ) in breast cancer predisposition , high-resolution genome-wide scans were performed on genomic DNA of 103 BRCA1 , BRCA2 , and PALB2 mutation negative familial breast cancer cases and 128 geographically matched healthy female controls; for replication an independent cohort of 75 similarly mutation negative young breast cancer patients was used . All observed rare variants were confirmed by independent methods . The studied breast cancer cases showed a consistent increase in the frequency of rare CNVs when compared to controls . Furthermore , the biological networks of the disrupted genes differed between the two groups . In familial cases the observed mutations disrupted genes , which were significantly overrepresented in cellular functions related to maintenance of genomic integrity , including DNA double-strand break repair ( P = 0 . 0211 ) . Biological network analysis in the two independent breast cancer cohorts showed that the disrupted genes were closely related to estrogen signaling and TP53 centered tumor suppressor network . These results suggest that rare CNVs represent an alternative source of genetic variation influencing hereditary risk for breast cancer .
Breast cancer is the most common malignancy affecting women . It is a complex disease with a well-established genetic component [1]; however , most of the familial and young breast cancer cases still remain unexplained by inherited mutations in the known susceptibility genes [2] . Multiple genome-wide association studies ( GWAS ) have identified several breast cancer associated single nucleotide polymorphisms ( SNPs ) , but these have only modest effect sizes and explain much less of the heritability than originally anticipated [3] . Consequently , the contribution of rare variants with moderate to even high disease penetrance is now beginning to be more widely accepted . With the exception of some specific founder mutations , these rare variants are individually infrequent , and even specific to single cases or families . Much of the work with rare genomic variants has been conducted through candidate gene re-sequencing studies mainly concentrating on DNA damage response genes , Fanconi anemia/BRCA pathway genes in particular , and their coding region variations [2] . However , rare genomic microduplications and microdeletions , also known as structural variants or copy number variants ( CNVs ) , could represent an alternative class of genetic variation responsible for increased cancer risk . Recent reports have suggested a role for genomic structural variants in susceptibility to various diseases , particularly neurodevelopmental disorders [4] , [5] . Association of common CNVs with breast cancer susceptibility has been ruled out by a recently performed large case-control study [6] , but the contribution of rare CNVs still remains poorly explored . As alleles in this variation class will be individually rare , the studies remain statistically underpowered to identify any specific loci involved , but the overall involvement can be tested by comparing the collective frequency of rare variants in cases with that in controls [5] . Moreover , the functional profiling of the disrupted genes will have a potential to reveal biological processes , which when defective could predispose to breast cancer . The known susceptibility genes are already considered to cause cancer predisposition through different mechanisms . Whereas BRCA1 and BRCA2 function in DNA repair [7] , other high-risk susceptibility genes , TP53 and PTEN , participate in cell cycle control and regulation of cell proliferation [8] , [9] . Here we have examined whether rare CNVs throughout the genome display an increased frequency in familial and young breast cancer cases when compared to healthy controls , and whether the biological pathways or processes , to which the disrupted genes relate to differ between the groups . Our results provide evidence that rare CNVs contribute to breast cancer susceptibility and that the disrupted genes are closely related to the TP53 tumor suppression network and to estrogen signaling .
Genome-wide scans for structural variants were performed on 103 familial breast cancer cases and 128 controls , using high-resolution Illumina HumanOmni1-Quad BeadChips . Stringent quality control criteria were applied to ensure that ascertainment of CNVs was consistent between cases and controls . The frequencies of common CNVs were monitored in both groups , and their frequency did not significantly differ ( mean 9 . 7 CNVs for cases and 9 . 13 CNVs for controls ) . Rare variants were defined as those that did not overlap over 60% with the common CNVs in Toronto Database of Genomic Variants , and all CNVs fulfilling the rare variant criteria were confirmed by independent method . In the studied 231 subjects we observed 65 microdeletions and microduplications , ranging in size from 25 kb to 612 kb . In cases , there were 15 deletions ( mean length 123 kb , median 61 kb ) and 20 duplications ( mean 216 kb , median 173 kb ) , whereas in controls 14 deletions ( mean 146 kb , median 133 kb ) and 16 duplications ( mean 242 kb , median 186 kb ) were observed . Among familial breast cancer cases the total number of rare CNVs was slightly higher than in controls: their proportion was also higher when only considering those rare CNVs involving genes , and those directly disrupting genes . This trend stayed the same when analyzing the independent young breast cancer cohort of 75 patients ( Table 1 ) . The difference was most profound when considering CNVs disrupting genes and restricting the analysis to variants not shared between cases and controls . Familial cases showed almost twice , and young breast cancer cases 1 . 5 times the number of rare CNVs compared to controls , but none of the differences were statistically significant . The genes within each rare CNV locus were identified ( Tables S1 , S2 and S3 ) , and functions and pathways of the involved genes ( Table S1 ) were assessed by using the Ingenuity Pathway Analysis ( IPA ) classification system . Analyses were restricted to genes , which were either disrupted by the breakpoints or deleted entirely , as mutations disrupting only part of the gene are likely to have biological consequences , and entirely deleted genes in the case of tumor suppressors follow the rationale of Knudson's two hit model or haploinsufficiency [10] . Only a few of the disrupted genes were part of known canonical pathways , and neither cases nor controls showed significant increase in any of them . The genes disrupted in familial cases showed , however , a significant overrepresentation in functions involving the maintenance of genomic integrity ( Table 2 ) , whereas no particular functions were overrepresented among controls . Three of the genes disrupted in cases were directly involved in double-strand break ( DSB ) repair signaling: BLM participates in BRCA1-mediated DNA damage response [11] , RECQL4 is involved in DNA replication and DSB repair [12] , and DCLRE1C operates in DSB repair by non-homologous end joining [13] . Both BLM and RECQL4 are RecQ family DNA helicases with an integral role in the maintenance of genomic stability . Their defects result in recessive cancer predisposition syndromes , Bloom and Rothmund-Thompson syndrome , respectively [14] , [15] . DCLRE1C encodes ARTEMIS , which is essential for V ( D ) J recombination . Biallelic mutations result in severe combined immunodeficiency ( SCID ) , in which lymphoma has been described [16] . Curiously , the currently observed DCLRE1C allele is one of the most frequent mutations reported among SCID patients . This null allele comprises a gross deletion of exons 1–4 and the adjacent MEIG1 gene and results from homologous recombination of DCLRE1C with the pseudo-DCLRE1C gene , located 61 . 2 kb upstream [17] . Based on their biological functions BLM , RECQL4 and DCLRE1C all represent attractive susceptibility genes , although to date clearly deleterious , breast cancer related mutations have not been reported in any of them . Although not significantly overrepresented , it should be noted , however , that another DNA repair gene , MCPH1 , was found to be disrupted in one of the studied controls . MCPH1 is an early DNA damage responsive protein , the dysfunction of which leads to recessive primary microcephaly without any reported malignancies [18] . The observed CNV deletes exon 13 and is predicted to lead to out of frame translation of the last exon , number 14 , thereby disrupting one of the three BRCT domains of MCPH1 . The carrier was still healthy at the age of 59 years , supporting the previous notion that all DNA damage response gene deficiencies do not necessarily predispose to malignancy . The genes disrupted in familial cases were also highly overrepresented among genes connected to diabetes mellitus ( P = 0 . 000268 ) ; this connection was mediated mainly through SNP associations observed in GWAS [19] . This overrepresentation was also seen in the young breast cancer cohort ( P = 0 . 0246 ) , but not in controls . Of the 16 diabetes associated genes 6 were under β-estradiol regulation . The strict pathway-based approach has several limitations as the function of many genes is currently unknown and cannot be assigned to any predetermined pathways [20] . Consequently , we next analyzed IPA networks , which map the biological relationships of the uploaded genes . Curiously , analysis with familial cases revealed a network centered on TP53 and β-estradiol ( score 29 ) . The same TP53 and β-estradiol centered network was observed when analyzing genes disrupted in the young breast cancer cohort ( score 28 ) ( Figures S1 and S2 ) . When analyzing both case cohorts together the network with the highest scores ( 35 , 31 ) centered on TP53 , β-estradiol and CTNNB1 ( encoding β–catenin , the oncogenic nuclear accumulation of which occurs in several malignancies , including breast cancer [21] ) and the other around β-estradiol ( Figure 1 , Table 3 ) . Neither the TP53 nor β-estradiol centered network was observed in controls , strongly arguing in favour of the possibility that dysregulation of these networks is disease related . The TP53 centered network appears to have obvious tumor suppressive function , as p53 itself is a key regulator in preventing cells from malignancy . Somatic TP53 mutations occur frequently in human malignancies , and germline lesions associate with the cancer prone Li-Fraumeni syndrome [22] . In the studied breast cancer cases , six genes disrupted by the observed rare CNVs were directly linked to TP53 ( Figure 1A ) , and all encode proteins functioning in pathways with a potential role in malignancy prevention . Two of these , RECQL4 and BLM , were DNA damage response proteins . Network interactions were based on the repression of RECQL4 transcription by p53 [23] , and the requirement of BLM for p53 localization to stalled replication forks [24] . The other four interactions were based on direct binding of p53 with HECW2 [25] , DAB2IP and EIF2C2 [26]; for CASP3 p53 has been shown to increase its activation [27] . The HECW2 disrupting allele was observed in two familial cases , whereas the others were all singletons ( Table S1 ) . The other network indicated in both of the studied breast cancer case cohorts centered on β-estradiol ( Figure 1A and 1B ) , which is the primary biologically active form of estrogen . Exposure to both exogenous and endogenous estrogens is a well-established risk factor for breast cancer , and disruptions in estrogen signaling and metabolism have a potential to affect this risk . The physiological effects of estrogens are mediated by their ability to alter the expression of their target genes . Estrogens play a key role in proliferation and differentiation of healthy breast epithelium , but also contribute to the progression of breast cancer by promoting the growth of transformed cells [28] . Many of the estrogen actions are mediated by intracellular estrogen receptors ESR1 and ESR2 [29] . The β-estradiol centered network consisted of several β-estradiol responsive genes , ANKS1B [30] , NXPH1 , MEP1B [31] , CASP3 [32] and ACSL1 [31] , whereas when separately tested in IPA none of the genes disrupted in controls were found to be under β-estradiol regulation . Of the network genes ESR2 , STRN and ANKS1B exhibited recurrent disrupting alleles among cancer cases ( Table S1 ) , emphasizing their potential role in breast cancer predisposition .
The results from our high-resolution genome-wide scans for structural variants provide evidence that rare CNVs contribute to breast cancer susceptibility . When compared to controls , the studied breast cancer cases showed a slight but consistent increase in the frequency of rare CNVs . The difference was not as profound as seen in psychiatric disorder studies where the observed changes , typically involving large genomic regions and numerous genes , can have very severe effects on patients' phenotype and many of which are de novo mutations [4] , [5] . However , in our study the biological networks affected by the disrupted genes differed between breast cancer cases and controls , supporting their role in cancer predisposition . The genes disrupted in familial cases showed a significant overrepresentation in functions involving the maintenance of genomic integrity . This included DSB repair , which is consistent with the prevailing paradigm that defects in this pathway contribute to breast cancer predisposition [2] . The three DSB repair genes , BLM , RECQL4 and DCLRE1C , disrupted in the case group all represent attractive breast cancer susceptibility genes . Moreover , IPA analysis demonstrated that the genes disrupted by rare CNVs in the studied breast cancer cases formed a network centered on TP53 and β-estradiol , a notion confirmed in two independent cohorts . Both networks are coherent and biologically meaningful , and their identification through the used genome-wide approach provides strong evidence for a role in breast cancer predisposition . TP53 network genes encode proteins functioning in pathways with potential role in malignancy prevention , including DNA damage response and apoptosis [25] , but also RNA interference [33] . They all represent attractive susceptibility genes , which could harbor also other cancer predisposing mutations; thus being excellent candidates for re-sequencing studies . Of the disrupted TP53 network genes DAB2IP and CASP3 were particularly interesting . DAB2IP is a member of the Ras GTPase-activating gene family and has been reported to act as a tumor suppressor . Inactivation of DAB2IP by promoter methylation occurs in several malignancies , including prostate and breast cancer [34] , and it has been shown to modulate epithelial-to-mesenchymal transition and prostate-cancer metastasis [35] . CASP3 is an apoptosis related gene , which encodes a member of a highly conserved caspase protease family , caspase 3 . Caspases are key intermediaries of the apoptotic process , failure of which can lead to cancer [36] . Various molecular epidemiological studies have suggested that SNPs in caspases may contribute to cancer risk , and a common coding variant in caspase 8 has been associated with breast cancer susceptibility [36] , [37] . Curiously , apoptosis is also one of the numerous genomic integrity maintenance functions of BRCA1 . Caspase 3 has been reported to mediate the cleavage of BRCA1 during UV-induced apoptosis , and the cleaved C-terminal fragment triggers the apoptotic response through activation of BRCA1 downstream effectors [38] . The rare CNVs disrupting the DAB2IP and CASP3 genes were both predicted to result in null alleles ( Table 3 ) . For estrogen , there are multiple lines of evidence for its profound role in breast cancer development , and disruptions in estrogen signaling and metabolism have long been considered to affect breast cancer risk . The estrogen network was largely explained by the genes under β-estradiol regulation , but two of the disrupted genes , ESR2 and STRN , had a more straightforward role in estrogen signalling . ESR2 encodes the estrogen receptor β , which is one of the main mediators of estrogen actions within the cell [29] . It binds estrogens with a similar affinity as estrogen receptor α , and activates expression of estrogen response element containing genes [39] . ESR2 has previously been suggested to harbor common breast cancer predisposing variants [40] , [41] , and ESR2 variation has been suggested to influence the development of breast cancer also by in vitro studies [42] . In contrast , striatin acts as molecular scaffold in non-genomic estrogen-mediated signaling [43] . It physically interacts with calmodulin 1 [44] and estrogen receptor α , and also forms a complex with protein phosphatase 2A , which also regulates the function of estrogen receptor α [45] . The identification of a recurrent deletion allele in CYP2C19 , encoding an enzyme involved in estrogen metabolism [46] and with an increased frequency in familial cases ( Table S2 ) , further emphasizes the role of estrogen in breast cancer predisposition . One CYP2C19 allele , CYP2C19*17 , defining an ultra-rapid metabolizer phenotype , has previously been associated with a decreased risk for breast cancer . This suggests that increased catabolism of estrogens by CYP2C19 may lead to decreased estrogen levels and therefore reduced breast cancer risk [47] . Correspondingly , decreased activity of CYP2C19 through haploinsufficiency might potentially increase the risk of breast cancer . Curiously , based on their function both ESR2 [40] , [41] and CYP2C19 [47] have long been considered strong candidate genes for breast cancer susceptibility . However , no structural variants have previously been reported in either of them , and it is possible that CNVs might represent a new class of cancer predisposing variation in both genes . Functionally relevant structural variants might be present also in other CYP genes that locate in gene clusters , like CYP2C19 [48] . The clustering of similar genes increases the potential for unequal crossing-over between sister chromatids and thus for creation of CNV alleles . The genes disrupted in both studied breast cancer cohorts were also significantly overrepresented among genes connected to diabetes mellitus . This unexpected result likely represents shared risk factors predisposing to both breast cancer and diabetes . Indeed , these two diseases have already been reported to share several non-genetic risk factors , including obesity and a sedentary lifestyle . The hormonal factors altered in diabetes include several hormonal systems that may also affect the development of breast cancer , including insulin , insulin-like growth factors , and other growth factors as well as estrogen [49] , [50] . Our results support estrogen being the key link in the association between diabetes and breast cancer , as over one third of the diabetes associated genes in the two studied breast cancer cohorts were part of the β-estradiol network . In conclusion , rare CNVs should be recognized as an alternative source of genetic variation influencing breast cancer risk . This notion is further supported by a recent study which also provided evidence for rare CNVs' contribution to familial and early-onset breast cancer [51] . The results from the current network analysis with two independent breast cancer cohorts provide strong evidence for the role of estrogen mediated signaling in breast cancer predisposition and reinforce the concept of TP53 centered tumor suppression in the prevention of malignancy . The variety of disrupted genes belonging to these networks underscores that diverse mechanisms are likely to be relevant to breast cancer pathogenesis .
The studied familial breast cancer cohort consisted of affected index cases of 103 Northern Finnish breast , or breast-and ovarian cancer families . 73 of the families were considered as high risk ones: 67 had three or more cases of breast cancer , potentially in combination with single ovarian cancer in first- or second-degree relatives , and 6 had two cases of breast , or breast and ovarian cancer in first- or second-degree relatives , of which at least one with early disease onset ( <35 years ) , bilateral breast cancer , or multiple primary tumors including breast or ovarian cancer in the same individual . The remaining 30 families were indicative of moderate disease susceptibility , and had two cases of breast cancer in first- or second-degree relatives , of which at least the other breast cancer was diagnosed under the age of 50 . The median at the age of diagnosis for the familial cases was 49 years ( variation 26–89 years ) , and all families were negative for Finnish BRCA1 , BRCA2 , TP53 and PALB2 founder mutations [52] . The studied young breast cancer cohort consisted of 75 Northern Finnish patients that were diagnosed with breast cancer at or under the age of 40 ( median 38 , variation 25–40 years ) . These patients were unselected for a family history of the disease , and tested negative for Finnish BRCA1 , BRCA2 and PALB2 founder mutations . This independent breast cancer cohort was collected as a validation group for the studied familial cases , based on the assumption that when a woman under the age of 40 years develops breast cancer , a hereditary predisposition may be suspected regardless whether there is a family history or not [53] . All biological specimens and clinical information of the familial and young breast cancer cases investigated were collected at the Oulu University Hospital , with the written informed consent of the patients . The geographically and ancestrally matched control group consisted of 128 anonymous cancer-free female Northern Finnish Red-Cross blood donors ( median age at monitoring was 56 , variation 50–66 years ) . Permission to use the above mentioned patient and control materials for studies on hereditary predisposition to cancer has been obtained from the Finnish Ministry of Social Affairs and Health ( Dnr 46/07/98 ) , and the Ethical Committee of the Northern-Ostrobothnia Health Care District ( Dnr 88/2000+amendment ) . All genomic DNA samples analyzed derived from blood samples extracted using either the standard phenol-chloroform method , Puregene D-50K purification kit ( Gentra , Minneapolis , MN , USA ) , or UltraClean Blood DNA Isolation Kit ( MoBio , Carlsbad , CA , USA ) and no DNA samples from immortalized lymphoblastoid cell lines were used . CNV discovery for both the familial and young breast cancer cohort as well as for the healthy controls was performed by using Illumina HumanOmni1-Quad BeadChips ( Illumina Inc . , San Diego , CA , USA ) . This provides high-resolution coverage of the genome with over one million genetic markers , including those derived from the 1 , 000 Genomes Project and all three HapMap phases , and enables precise definition of the breakpoints . All samples included in the array had to pass the standard quality control ( QC ) measures , which included agarose gel runs to confirm the integrity of the DNA sample , and accurate concentration determination with three-step dilution measurements . To control the confounding effects resulting from the handling of the samples and subsequent CNV analysis , all cases and controls were given new IDs and were blindly analyzed without knowing their disease status . All samples were analyzed following the Illumina provided protocol in the same laboratory ( Laboratory of Cancer Genetics , University of Oulu ) with same arrays at the same period of time , with random places on the chip . Samples were analyzed with GenomeStudio Genotyping module ( Illumina ) and Nexus Copy Number Discovery Edition 5 . 1 software ( BioDiscovery Inc . , El Segundo , CA , USA ) . Projects were created in GenomeStudio , and samples having Call Rates over 98% were transported to Nexus where samples with quality score <0 . 15 were passed on for further analysis . In order to obtain a high-quality CNV dataset , we restricted the analysis to CNVs called by two independent algorithms . In Nexus the SNP-FASST2 segmentation algorithm was used . The significance threshold was set to 1 . 0E-06 , and +0 . 25 for gains and −0 . 25 for losses . The minimum number of probes needed for segment calling was set to 25 , and minimum loss of heterozygosity length to 10 000 kb . Quadratic correction was used as a systematic correction of artifacts caused by GC content and fragment length . Samples passing all the QCs but showing over 50 copy number changes in Nexus were excluded . The sensitivity of detection in Nexus was evaluated by analyzing 11 samples containing known deletions/amplifications confirmed by independent methods , and all changes were detected under the parameters used . All observed CNVs had to be confirmed by Illumina cnvPartition 2 . 4 . 4 software , using a confidence level of over 50 in order to be included in the analysis: values of 50 or higher tend to reflect a region with high confidence . The breakpoints of the observed aberrations were defined using the information obtained from both Nexus and GenomeStudio , and CNVs that appeared to be artificially split by the algorithm were joined . The focus of our interest was on rare duplications and deletions . Rare events were defined as those which were called by two independent algorithms and did not overlap over 60% with the common CNVs in the CNV track defined in Nexus , based on the Toronto Database of Genomic Variants ( DGV ) . However , as the DGV database presents several known cancer susceptibility genes as containing polymorphic CNVs , each CNV not fulfilling the rare variant criteria were individually inspected before exclusion . As a result , we decided to include “common” CNV in the rare variant analysis if fulfilling all three of the following criteria: 1 ) the CNV disrupts the involved gene partially , or deletes it entirely , 2 ) affected gene is a known breast cancer susceptibility gene , or based on it biological function it is a highly likely breast cancer susceptibility gene , and 3 ) biallelic defects in the involved gene lead to a rare genomic disorder , indicating that the defective allele is highly unlikely to be polymorphism . This led to inclusion of three alleles disrupting the following genes: RECQL4 , MCPH1 and DCLRE1C . All “rare” events which were present at polymorphic frequencies in the pooled population of 250 cases and controls , except those that were specific or showed a clear enrichment in cancer cases , were excluded from further analyses . All potential events of interest , rare CNV variants , were validated by another independent method , either by Affymetrix Genome-Wide Human SNP Array 6 . 0 platform ( Affymetrix , Santa Clara , CA , USA ) or quantitative real time PCR ( qPCR ) . Affymetrix chip analysis was performed following all the QC measures recommended by the protocol , and Affymetrix CEL files were transported to Nexus for analysis with the SNP-FASST2 segmentation algorithm . Confirmation with qPCR was done with BioRad CFX96 using SsoFast EvaGreen Supermix ( BioRad , Hercules , CA , USA ) . Samples with rare CNVs and at least 3 wildtype controls were analyzed in triplicate , and quantitation was done with CFX manager software ( version 1 . 5 ) under gene expression analysis . RAD50 and CtIP were used as reference genes . Rare variant carrier frequencies between cancer cases and controls were compared using Fisher's exact test . The frequency of common CNVs and the size of duplications and deletions was monitored both in cases and controls and tested for differences with Mann-Whitney U-test ( PASW Statistics 18 . 0 for Windows , SPSS Inc . , Chicago , IL , USA ) . All tests were two-sided and considered to be statistically significant with a P-value≤0 . 05 . For pathway and biological function analysis , Ingenuity Pathway Analysis ( IPA , http://www . ingenuity . com/ ) was used . The list of disrupted genes [defined as genes ( including also their promoter region ) disrupted by the breakpoints or deleted entirely , and not shared between cases and controls] were uploaded to IPA , which is an online exploratory tool with a curated database for over 20 , 000 mammalian genes and 1 . 9 million published literature references . Together with several databases , including Entrez Gene , Gene Ontology and GWAS database , IPA integrates transcriptomics data with mining techniques to predict and build up networks , pathways and biological function clusters . The software maps the biological relationships of the uploaded genes according to published literature included in the Ingenuity database . The output results are given as scores and P-values computed based on the numbers of uploaded genes in the cluster or network and the size of network or cluster in the Ingenuity knowledge database . Benjamini-Hochberg multiple testing correction P-values ( to monitor the false discovery rate ) were used to determine the probability that each biological function or overrepresentation in diseases is due to change alone . Scores for IPA networks are the negative logarithm of the P-value , and they indicate the likelihood of the genes analyzed in a network for being found together due to random chance . Scores 2 or higher have at least a 99% likelihood of not being generated by chance alone .
|
Although genetic susceptibility to breast cancer has been well-established , the majority of the predisposing factors still remain unidentified . Here , we have taken advantage of recent technical and methodological advances to examine the role of a new class of genomic variation , rare copy number variants ( CNVs ) , in hereditary predisposition to breast cancer . By examining 103 BRCA1/2 and PALB2 mutation negative familial and 75 young breast cancer cases , together with 128 geographically matched healthy female controls , we show that the frequency of rare CNVs is increased in cases when compared to controls and that the genes disrupted in individuals of specifically the two case groups are closely related to estrogen signaling and TP53 centered tumor suppressor network . The variety of disrupted genes belonging to these networks underscores that diverse mechanisms are likely to be relevant to breast cancer pathogenesis . The current results warrant the investigation of rare CNVs as new susceptibility factors in other cancer types as well .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"oncology",
"systems",
"biology",
"medicine",
"public",
"health",
"and",
"epidemiology",
"genomics",
"epidemiology",
"genetics",
"biology",
"computational",
"biology",
"genetics",
"and",
"genomics",
"clinical",
"genetics"
] |
2012
|
Rare Copy Number Variants Observed in Hereditary Breast Cancer Cases Disrupt Genes in Estrogen Signaling and TP53 Tumor Suppression Network
|
Chunking is the process by which frequently repeated segments of temporal inputs are concatenated into single units that are easy to process . Such a process is fundamental to time-series analysis in biological and artificial information processing systems . The brain efficiently acquires chunks from various information streams in an unsupervised manner; however , the underlying mechanisms of this process remain elusive . A widely-adopted statistical method for chunking consists of predicting frequently repeated contiguous elements in an input sequence based on unequal transition probabilities over sequence elements . However , recent experimental findings suggest that the brain is unlikely to adopt this method , as human subjects can chunk sequences with uniform transition probabilities . In this study , we propose a novel conceptual framework to overcome this limitation . In this process , neural networks learn to predict dynamical response patterns to sequence input rather than to directly learn transition patterns . Using a mutually supervising pair of reservoir computing modules , we demonstrate how this mechanism works in chunking sequences of letters or visual images with variable regularity and complexity . In addition , we demonstrate that background noise plays a crucial role in correctly learning chunks in this model . In particular , the model can successfully chunk sequences that conventional statistical approaches fail to chunk due to uniform transition probabilities . In addition , the neural responses of the model exhibit an interesting similarity to those of the basal ganglia observed after motor habit formation .
When a sequence of stimuli is repeated , they may be segmented into “chunks , ” which are then processed and stored as discrete units . This process , called “chunking” or "bracketing" [1] , takes place during various cognitive behaviors that require hierarchical sequence processing [2–5] . For instance , in motor learning , a sequence of smaller movements may be executed as one compound movement after repetitive practice [1 , 6–9] . During language acquisition , continuous vocal sounds are segmented into familiar groups of contiguous sounds that are processed as words [10 , 11] . Chunking is believed to reduce the complexity of sequence processing and hence the associated computational cost [1 , 12–13] . In this regard , chunking constitutes a crucial step in representing the hierarchical structure of sequential knowledge in neural circuits [14] . Chunking is believed to occur through two consecutive processes . Long and complex sequences are first segmented into shorter and simple sequences , while frequently repeated segments are concatenated into a single unit [15] . Various mechanisms of chunking have been proposed based on Bayesian computation [4 , 16] , statistical learning guided by prediction errors [17] , and a bifurcation structure ( stable heteroclinic orbits ) in nonlinear dynamical systems [18 , 19] . In addition , a neuromorphic hardware has been proposed [20] . However , none of these mechanisms have been shown to chunk with the level of flexibility that the brain offers . It also remains unclear whether a bifurcation theoretic mechanism exists that enables the chunking of arbitrary complex sequences . Many studies evaluating event segmentation in biological and artificial systems have focused on mechanisms to detect boundaries between events by transient increases in surprise signals , which are thought to form based on unequal transition probabilities among sequence elements [4 , 14 , 21–22] . However , human subjects can segment sequences of visual stimuli that have uniform transition probabilities and hence cannot be chunked by any variation of such mechanisms [23] . These findings suggest that biological neural networks favor a mechanism of chunking that is based on temporal community detection , in which stimuli that frequently go together are grouped into a chunk . A similar mechanism may also account for the brain’s ability to detect repetitions of patterned stimuli in random sequences [24–26] . However , the logic and neural mechanism of flexible and automatic chunking by the brain remain unknown . In this study , we propose a novel mechanism of unsupervised chunk learning based on a unique computational framework that differs from any of the previous proposals . In this mechanism , neural networks learn the low-dimensional dynamical trajectories embedding stereotyped responses to recurring segments ( chunks ) of a temporal input . We achieve this mechanism in a framework of cortical computation [27 , 28] by extending reservoir computing ( RC ) to unsupervised learning . RC is a high-dimensional dynamical system consisting of a recurrent neural network , readout units , with feedforward and feedback projections between them , and supervised learning in its original form [29] . We were able to attain unsupervised learning in a pair of independent RC modules supervising each other without any external instructive signal . As a consequence , they learned to mimic , or predict , the preferential responses of partner modules to chunks in a given temporal input . The primary interest of this study was determining the novel computational mechanism to segment information streams . However , an unexpected finding included a surprising similarity between the temporal response patterns of readout units in our model and a functional subtype of basal ganglia neurons , called stop cells , which are observed after habituation [7–9] . This finding suggests that the proposed paradigm of sequence processing has a biological relevance .
To demonstrate the basic framework of our model , we first consider the case where the input sequence alternates a single chunk ( i . e . , a-b-c-d ) and random sequences of discrete items , which are chosen from the remaining 22 letters of the English alphabet ( e to z ) ( Fig 1A ) . In reality , each letter may correspond to a brief stimulus in any sensory modality , such as a brief tone signal , and is given to the model through phasic activation of a specific input neuron ( Iμ ( t ) in Eq 4 in the Methods ) with slow rise and decay constants ( Fig 1B ) . Thus , the number of input neurons coincides with that of letters . The random sequence components are introduced to unambiguously define the initial and end points of a chunk , and their lengths vary with every repetition cycle within the length range of 5 to 8 . Our network model comprises two mutually non-interacting RC modules , each of which consists of a recurrent network ( reservoir ) of rate-based neurons and a readout unit . Each RC module receives an identical input sequence ( Fig 1C ) . Each reservoir neuron receives a selective input from one of the input neurons and hence has a preferred stimulus . As shown later , however , this constraint is not essential for chunking and can be relaxed . Within each reservoir , all neurons are mutually connected and project to a readout unit , which projects back to all neurons belonging to the same reservoir . Note that the two reservoirs have different recurrent wiring patterns and hence are not identical . The activity of each readout unit z ( t ) is given as a weighted sum of the activities r ( t ) of the reservoir neurons projecting to the readout: z ( t ) = wTr ( t ) . Note that one readout unit per reservoir is sufficient for learning a single chunk . We will consider more complex cases later . The weight vector w is modifiable through the FORCE learning algorithm [29] , whereas the recurrent and feedback connections are non-plastic because the model can solve the present task without modifying these connections . The initial states of the reservoirs are weakly chaotic as in the previous model [29] . See the Methods for details of the model and values of the relevant parameters . A unique feature of the present model is that the output of each readout unit is used as a teacher signal to train the readout weights of the other reservoir module , implying that the two RC modules supervise each other . As a consequence , although the FORCE learning per se is a supervised learning rule , the entire network , which we call the "dual RC system , " is subject to unsupervised leaning because teaching signals originate from the system itself . The details of the teaching signals will be shown later . The design of the teaching signals is the key for successful chunk learning in the present model . The teaching signals should be symmetric with respect to the interchange of the two readout units , and should be determined such that the two systems stop learning when the two readout units output similar response patterns . In other words , the teaching signals eventually become identical between the two RC modules during learning . The following teaching signals fi enable chunk learning in the proposed dual RC system: fi ( t ) =[tanh ( z^j ( t ) /β ) ]+ . ( i , j=1 , 2;i≠j ) ( 1 ) where z^i is the normalized output of the i-th readout unit ( Methods ) , the threshold linear function [x]+ returns 0 if x≦0 , and [x]+ = x if x>0 , and the constant was set as β = 3 . Defining error signals as ei ( t ) = zi ( t ) – fi ( t ) , we trained the pair of RC modules through the FORCE learning algorithm until the error signals become sufficiently small ( typically , about 0 . 01 ) and the readout weights converge to equilibrium values ( within small fluctuations ) . The sigmoidal function allows the system to learn nontrivial solutions zj ( t ) ≠ 0 , while maintaining the outputs ( and hence the teaching signals ) to be finite during learning . Furthermore , the saturation part of sigmoidal function prevents the model from responding too strongly to a specific chunk and makes it easier to detect all the chunks embedded in the input sequence . This activity regulation is particularly important in the learning of multiple chunks studied later . The threshold linear function makes the outputs positive; these nonlinear transformations greatly improved the performance of learning . Importantly , the teaching signals do not explicitly contain information about the structure and timing of chunks in the input sequence . This dual RC system converged to a state of stable operations when the two RC systems produced similar teaching signals ( hence similar outputs ) that were consistent with the temporal structure of the input sequence ( S1 Fig ) . The readout units did not respond to the chunk before learning ( Fig 1D ) . After learning , the responses of the readout units were tested for the input sequences that had not been used for the training . The test sequences contained the same chunk “a-b-c-d , ” but the random sequence part was different . Given these inputs , the readout units exhibited steady phasic responses time-locked to the chunk ( Fig 1E ) . The readout activity piled up gradually in the beginning of the chunk , rapidly increased at its end , and then returned to a baseline level . The selective responses to the chunk were also successfully learned when each reservoir neuron was innervated by multiple input neurons . As shown in Fig 1F , the system succeeded in learning when randomly-chosen 10% or 40% of input neurons projected to each reservoir neuron , but failed when the fraction was 70% . Thus , responses of the individual reservoir neurons should be sufficiently independent of each other to robustly capture the recurrence of chunks . We can extend the previous learning rule for learning multiple chunks without difficulty . To show this , we embedded three chunks into a random input sequence ( Fig 2A , top ) . The three chunks had the same occurrence probability of 1/3 . To process this complex input sequence , we made two modifications to the previous model . First , each reservoir was connected to three readout units ( z1 , z2 , z3 for the 1st reservoir and z4 , z5 , z6 for the 2nd reservoir ) , each responsible for the learning of one of the three chunks ( Fig 2B ) . Second , we modified the teaching signals as follows: fa ( t ) =[tanh ( ( z^a′ ( t ) −γ∑c=4 , 5 , 6′z^c ( t ) ) /β ) ]+ ( a=1 , 2 , 3 ) ( 2 ) fb ( t ) =[tanh ( ( z^b′ ( t ) −γ∑c=1 , 2 , 3′z^c ( t ) ) /β ) ]+ ( b=4 , 5 , 6 ) ( 3 ) where a’ and b’ refer to the corresponding readout units of the partner RC modules ( i . e . , a’ = a+3 and b’ = b-3 ) , and dashes in the second term indicate that the corresponding readout unit should be excluded from the summation . The constant γ was set as 0 . 5 . Thus , teaching signals were exchanged between the RC modules as in the previous case . Each readout unit receives a triplet of teaching signals from the partner network , in which one is cooperative and the other two are competitive ( S2A Fig ) . These signals allow each readout unit to adopt to a specific chunk , but the chunk to be learned by a readout unit is not a priori specified because the teaching signals are symmetric with respect to the permutation of indices per reservoir . A further extension of the learning rule to an arbitrary number of chunks is straightforward . As in the case with a single chunk , each readout unit displayed a ramping activity selective to a specific chunk , signaling successful chunk learning ( Fig 2C ) . During this learning , teaching signals also self-organized such that each pair of the readout units eventually exhibited a selective response to a specific chunk , indicating that the teaching signals work adequately ( S2B Fig ) . The complex form of teacher signals looks somewhat biologically unrealistic , but they can easily be implemented by inhibitory neurons ( S2C Fig: see Methods ) to generate chunk-selective phasic readout responses ( S2D Fig ) . Below , inhibitory neurons are not explicitly modeled . The question then arises whether the RC system could also learn multiple chunks when they occur continuously without temporal separations by random sequences . To study this , we trained the model by using input sequences in which three chunks appear randomly and consecutively with equal probabilities , without any interval of random sequences ( Fig 2A , bottom ) . Thus , the same RC system as before could easily learn multiple chunks ( Fig 2D ) . A notable difference was that , outside of the chunks , the readout activity decayed faster for undisturbed sequences than for temporally separated ones ( Fig 2E ) . In fact , learning proceeded faster for the former sequences ( Fig 2F ) , suggesting that learning is more effective when chunks are not disrupted by random sequences . Throughout this study , one learning step corresponds to 15 sec . Next , we investigated how the activities of reservoir neurons encode chunks . Here , the network was trained on sequences containing three chunks and random sequences . In each reservoir , a subset of neurons selectively responded to each chunk after learning ( S3A Fig ) . Therefore , we classified reservoir neurons into three ensembles according to the selectivity of their responses to each chunk ( Methods ) . Some reservoir neurons responded to more than one chunk , but they were excluded from the following analysis for the sake of simplicity . Each neural ensemble received slightly stronger inputs from the specific chunk it encoded , which then determines the selectivity of the encoding ensemble ( S3B Fig ) . Through learning , the neural ensemble encoding a particular chunk developed stronger projections to the corresponding readout unit compared with other neural ensembles ( S3C Fig ) . Consistent with this , the distribution of readout weights was more positively skewed in encoding ensembles than in non-encoding ensembles ( S3D Fig ) . Moreover , the readout unit projected back to the corresponding encoding neuron ensemble more strongly than to the other ensembles ( S3E Fig ) . Because feedback connections were not modifiable , these results imply that readout connections were strengthened between readout units and reservoir neurons that happened to receive relatively strong feedback from the readout unit . To gain further insight into the mechanism of chunking , we explored the low-dimensional characteristics of the dynamics of reservoir networks . In our model , the two RC modules , termed R1 and R2 , are thought to mimic others . This would be possible when the two recurrent networks receiving the same input sequence predict the responses of other modules well . To see how this prediction is formed , we calculated the principal components ( PCs ) of the post-learning activity of trained recurrent networks in the example shown in Fig 1 . After learning , the lowest principal component ( PC1 ) but not the other PCs , of each reservoir resembled the phasic response of the corresponding readout unit during the presentation of chunks ( Fig 3A , left ) . The learned trajectories wandered in the low-dimensional PC space outside the chunks where teacher signals vanished , while , inside the chunks , non-vanishing teacher signals rapidly constrained both trajectories in narrower regions showing similar PC1 values ( Fig 3A , right ) . This behavior is understandable because the eigenvalues of PCs decay rapidly ( Fig 3B ) . Interestingly , the correlation coefficient between each PC and the readout activity decayed more dramatically ( Fig 3C ) . Accordingly , the direction of readout weight vector was more strongly correlated with that of PC1 compared to other PCs ( Fig 3D ) . These results suggest that the low-dimensional characteristics of neural dynamics play a pivotal role in encoding the chunks . We then determined to what extent the responses of R1 and R2 are represented by the low-dimensional dynamical characteristics of R1 . We calculated the PCs of recurrent network dynamics in R1 , and expanded its population rate vector and readout weight vector up to the M-th order of these PCs ( M ≦ NG ) . Then , we reconstructed the output of R1 by using the M-th order rate vector and the M-th order weight vector on the low-dimensional subspace spanned by the first M PCs ( Methods ) . In Fig 3E , we calculated the differences between the reconstructed R1-output and the full outputs of R1 ( within-self difference ) and R2 ( between-partner difference ) . Before learning , both differences remained large as M was increased . After learning , the “within-self” difference rapidly decreased for M < 30–40 and then gradually approached zero . The “between-partner” difference also rapidly dropped for relatively small values of M , but it stopped decreasing for M > 50 , remaining at relatively large values . These results suggest that R1’s reservoir , as well as R2’s reservoir , learns to mimic the partner's response by using the low-dimensional characteristics of its recurrent neural dynamics . The role of low-dimensional neural dynamics in a broad range of computation was recently explored in a class of recurrent network models with a minimal connectivity structure [30] , which is a combination of a low-rank structured matrix and a random unstructured matrix . The low-rank matrix may be trained to give task-related low-dimensional dynamics whereas the random matrix may generate chaotic fluctuations useful for learning . The RC system can be approximately viewed as such a network , where the product of readout weight vector and feedback weight vector ( JGZ ) Tw defines a rank-one matrix and recurrent connections in the reservoir gives a random matrix . It will be intriguing to study the present chunk learning in the theoretical framework . Chunk learning may be easier and more accurate if chunks were shorter or network size is larger . However , we did not find a sharp drop of performance when the size of chunks was increased . To observe this , we first measured learning performance for two chunks with the sizes 4 and 7 by varying the network size . Instantaneous correlations were calculated between the activity of a readout unit and a reference response pattern , which takes the value 1 during the presentation of a chunk and is 0 otherwise , every 15 s during learning and were averaged over 20 independent simulations . Note that the maximum value of the correlation was 0 . 5 if the readout activity grows linearly from 0 to 1 during the chunk presentation . S4A Fig shows the correlations for input sequences containing the short or long chunk in networks of sizes NG = 30 , 300 , and 500 . Correlations were nearly zero before learning , but reached similar maximum values approximately within ten steps of learning . The average value of the correlations was generally larger for chunk size 4 than for chunk size 7 , but the differences were not significant ( S4B Fig ) . Second , we measured learning performance by varying the size of chunks with the network size fixed ( NG = 300 ) . In this simulation , we alternately presented a single chunk with the size s and random sequences of the sizes s + 2 to s + 5 , where each element of the random sequences was chosen from a set of 4s elements . Thus , the dual RC system had 5s input neurons . When the chunk size exceeded 10 ( S4C Fig ) , the value of correlation rapidly dropped , suggesting the existence of a critical chunk size beyond which learning performance is degraded . For s = 4 , learning performance showed unexpectedly large fluctuations due to some unknown reason . The explicit evaluation of the critical chunk size requires an analytic approach , which is beyond the scope of this study . In addition , a larger network did not necessarily show better performance . The magnitude of the post-learning instantaneous correlation was not significantly increased when the network size was 200 or greater ( S4B Fig ) . Thus , the performance of chunk learning does not scale with the network size . This is not so surprising because increasing the size of the reservoirs does not necessarily increase the variety of neural responses useful for learning if the size is already sufficiently large . This seems to be particularly the case in the proposed mechanism because it heavily relies on the low-dimensional characteristics of neural dynamics ( S3A Fig ) . We found that external noise plays an active role in successful chunking . We demonstrated this in the case where the input only contained a single chunk ( Fig 4A ) . In the absence of noise readout units , phasic responses were still observed , but these responses were not necessarily time-locked to chunks ( Fig 4A , vertical arrow ) . As shown later , the two RC modules in principle may agree on an arbitrary feature of the input sequence , which implies the RC system may converge to a local minimum of learning . Noise may help the system to escape from the local minima . On the other hand , too strong of noise completely deteriorated the phasic responses to chunks . Thus , the RC system could learn chunks only when a modest amount of external noise existed ( Fig 4B ) . In the presence of adequate noise ( σ = 0 . 25 ) , the average weight of the readout connections rapidly decreased to a small equilibrium value during learning ( Fig 4C ) , leaving some readout weights much stronger than the majority ( Fig 4D ) . This reduction was expected because external noise gives a regularization effect on synaptic weights in error-minimization learning [31] . The strong weights were obtained for readout connections from the reservoir neurons responding to the chunk , hence they were crucial for chunk detection . However , this was not the case in the absence of noise ( σ = 0 ) . We counted the fraction of strong readout connections that emerged from chunk-encoding reservoir neurons , where strong connections included those that were greater than the standard deviation of the weight distribution . Such a fraction was significantly larger in the presence of adequate noise than in the absence of noise . Under strong noise ( σ = 1 ) , although the weight distribution becomes more bimodal , the noise disrupted learning and the system failed to capture the chunks ( Fig 4E ) . Another possible mechanism in which the external noise would improve the learning performance is that the dynamics of RC modules with weak noise are too far in the chaotic regime and the external noise suppresses chaos to enable proper chunk learning [32] . To test this possibility , we compensated a decrease of σ by decreasing the strength of recurrent conections gG , which weakens the influences of chaos , and investigated whether the deterioration of performance is suppressed . The noise intensity was decreased from a modest level ( σ = 0 . 5 ) , and the values of σ and gG were decreased at the same rate . Although the improvement was not significant , the dual RC system better resisted the performance deterioration ( Fig 4B ) , suggesting that proper chunk learning requires a certain balance between external noise and chaotic dynamics . Though our results so far suggest that mutual supervision enables the RC system to learn recurring groups of items in a sequence , these results do not indicate how the system chooses particular groups for learning . The question then arises whether our model detects a “chunk” if a sequence merely repeats each letter randomly without temporal grouping . To study this , we constructed a set of input sequences of ten letters , where all the letters appeared equally often in each sequence . We then exposed the RC system with a readout unit to these sequences . We found that the system learned to respond to one of the letters with approximately equal probabilities ( S5A Fig ) . We then made the occurrence probability of letter “a” twice as large as the occurrence probabilities of the others and found that the system detected “a” about twice as frequent as the others ( S5B Fig ) . These results indicate that the learning performance of the dual RC system relies on the occurrence frequency of repeated features if there are no other characteristic temporal features in the input sequence . The frequency dependence of our model partially accounts for the features of sequences that are grouped into chunks . As demonstrated in Fig 3 , a pair of RC modules engage in the mutual prediction of the partners' response . This prediction would be easier for the items in the input that repeatedly occur in a fixed temporal order . However , the explicit role of temporal grouping in chunking remains to be further clarified . We then demonstrate that the RC system can simultaneously chunk multiple sequences with overlaps , where input sequences share some letters as common items . In some sequences , common subsequences appeared in the beginning or the end of chunks ( Fig 5A ) , whereas other sequences involved common subsequences in the middle of chunks ( Fig 5D ) . In both cases , the RC system ( with two readout units ) successfully chunked these input sequences without difficulty ( Fig 5B and 5E ) . Interestingly , the activity of the readout units averaged over repetitive presentations ceased to increase during the presentation of the overlapping part of the chunks ( Fig 5C and 5F ) . This seems reasonable as overlapping in part does not contribute to the prediction of the following items in the chunks and hence needs not be learned . So far , we have only studied discrete sequences of letters with varying complexity . However , the applicability of the proposed mechanism is not restricted to this relatively simple class of temporal inputs . We first showed the potential advantage of this mechanism over the conventional statistical methods , considering a system with three readout units ( per RC module ) for processing sequence inputs generated by a random walk through a graph ( Fig 6A ) . This was previously used in examining the learning ability of event recognition by human subjects [23] . Each node of this graph has exactly four neighbors , and hence is visited by random walk with uniform transition probabilities over all neighbors . Despite this uniformity , the graph has three clusters of densely connected nodes , which define the communities in the graph [33 , 34] . Human subjects and our model ( Fig 6A ) can easily chunk these clusters according to community structure , but machine-learning algorithms based on surprise signals ( e . g . , [21] ) cannot [23] . We further demonstrated that the proposed system can learn to detect two images recurring in visual input streams with ( Fig 6B ) and without ( Fig 6C ) random intervals of Gaussian noise stimuli . We examined whether learning speed depends on the resolution of images and found that such a dependence was weak if the network size was unchanged ( Fig 6D ) . Our results show the potential ability of the proposed mechanism in analyzing the community structure of a broad class of temporal inputs .
Conventional statistical methods of chunking use unequal transition probabilities between sequence elements as cues for sequence segmentation . In contrast , we propose a conceptually novel framework in which the neural system self-organizes its internal dynamics to respond preferentially to chunks ( i . e . , frequently recurring segments ) with a temporal input , rather than attempts to predict the temporal patterns of input sequences . We achieved this unsupervised learning in a network of paired RC modules mutually learning the responses of the partners . Sequence leaning with RC has been studied in motor control [35–37] and decision making [38 , 39] . Theoretical extensions to spiking neuron networks [40] and/or reward-based learning [41] have thus been proposed . In this study , we showed that RC can be used for the unsupervised learning of hidden structure of continuous information streams . Chunking has often been accounted for by predictive uncertainty or surprise [17 , 42–44] . However , recent evidence suggests the existence of an alternative mechanism of chunking in which events are segmented based on the temporal community structure of sequential stimuli [23] . Indeed , it has been shown that individual items in a sequence are concatenated into an event when they frequently go together in the sequence . This dual RC system automatically chunks a continuous flow of stimuli based on temporal clustering structure and the occurrence probabilities of the stimuli without relying on predictive uncertainty or surprise . In addition , the model can chunk clusters of sequence elements that cannot be chunked by conventional statistical methods based on unequal transition probabilities ( Fig 6A ) . Unsupervised chunk learning was previously modeled by using heteroclinic orbits in a dynamical neural system [19] . Though this mechanism enables the learning of prescribed chunks , whether it also offers flexible learning of arbitrary chunks remains unclear . Our model has some advantages over the previous models of chunking . Our model can detect multiple chunks embedded into random background sequences . To our knowledge , the detection of multiple chunks has not been seriously attempted in the presence of various types of input noise on chunking . Further , as shown in Fig 5 our model can learn multiple partially overlapping chunks without additional mechanisms , which was also previously difficult . On the other hand , a weak point is that our model requires specially designed teaching signals , which depend on the structure of chunks . Related to this , mutually inhibitory teaching signals were introduced in an ad-hoc manner to prevent multiple readout units from learning the same chunk . A more flexible mechanism of learning should be further explored . The dual RC system described here shows good performance in the presence of external noise . Without noise , the system also learns to respond to certain segments of a sequence , but these segments may not coincide with any of the frequently repeated chunks . An adequate amount of external noise eliminates such spurious responses and enables the system to respond to the most prominent features of a sequence , namely repeated chunks . This finding is interesting because the initial state of the dual RC system is chosen on the so-called “edge of chaos , ” on which weakly chaotic neural dynamics provide an adequate amount of flexibility for supervised learning [29 , 45–46] . Moreover , the present system assumes a similar initial state , but additionally requires the regularization of synaptic weight dynamics by noise ( Fig 4C ) . Training a recurrent neural network with an explicit regularization term is known to eliminate the strange neuronal responses that are not observed in the motor cortex [37] . The dual RC system learns sequence in an unsupervised fashion by using two neural networks and , in this sense , is similar to Generative Adversarial Network ( GAN ) in deep learning [47] . A critical difference , however , exists between the two models . In GANs , a generative network learns to mimic the structure of training data and a discriminative network learns to distinguish between samples from the training data and those generated by the generative network . Because the generative model learns to deceive the discriminative model , GANs learn the structure of data distribution under a conflicting cost function . By contrast , in the dual RC system , two neural networks learn to help each other for the formation of a consensus about the structure of temporal inputs . Therefore , our model is conceptually different from GANs . The structure of our model has an interesting similarity to cortico-basal ganglia loops , where two reservoirs may represent bi-hemispheric cortical networks and readout units may correspond to striatal neurons . The responses of readout units and those of striatal neurons in the formation of motor habits also look similar . Sequential motor behavior becomes more rigid and automatic over the course of learning and practice , and the basal ganglia is thought to play a pivotal role in habit formation [9 , 48] . For instance , in rats running in a T maze , the majority of dorsolateral striatal neurons exhibit burst firing when the run is initiated or completed , or both [49] . Similarly , in mice an increased population of striatal neurons selectively responds to the initial ( Start cells ) , the last ( Stop cells ) , or both actions in the trained behavioral sequence [7 , 8] . In our model , readout units respond strongly to the last component of each chunk , similar to the Stop cells . Our model predicts that the Stop-cell’s response may decrease when two motor chunks have overlapping portions ( Fig 5 ) . On the other hand , our model does not show Start cell-like responses . Whether and how Start cells are formed is an intriguing open question . The proposed learning scheme works most efficiently when two RC modules are not interconnected , but rather work independently . In fact , the performance of chunk learning drops below 50% of the initial level when the connection probability between the two reservoirs exceeds about 10% ( S4D Fig , see the Methods ) . This suggests that each RC module can obtain maximum information about temporal input when it receives the teaching signal completely from its outside . The existence of inter-reservoir connections implies that some portion of the teaching signal originates from its inside . Where can such independent networks be located in the brain ? One possibility is that they are represented by mutually disconnected recurrent neuronal networks in a local cortical area . Because they are functionally equivalent , it is unlikely that they are implemented in functionally distinct areas . Another intriguing possibility is that they are distributed to functionally equivalent cortical areas in different hemispheres . Indeed , the inferior frontal gyrus and the anterior insula are bilaterally activated when human subjects chunk visual information streams [23 , 50] . Whether subnetworks of pyramidal cells perform chunking in these or other cortical areas [51] remains an intriguing open question . In summary , we propose an unsupervised learning system by combining two independent reservoir computing modules . During learning , the two systems supervise each other to generate coincident outputs , which in turn allows the entire system to consistently learn chunks hidden in irregular input sequences . As chunking is a fundamental step in the analysis of sequence information , our results have significant implications for understanding how the brain models the external world .
In this study , the proposed model is composed of two recurrent networks ( reservoirs ) . Each recurrent network is composed of NG neurons . Each neuron follows the following dynamics as i = 1 , 2 , … NG , τxi˙ ( t ) =−xi ( t ) +gG∑j=1NGJijGGrj ( t ) +JiGZz ( t ) +∑μ=1NIJiμGIIμ ( t ) +σξi ( t ) , ( 4 ) ri ( t ) =tanh ( xi ( t ) ) , ( 5 ) where Iμ ( t ) is the activity of input neurons , ξi ( t ) is a random ( Wiener ) process and σ is the standard deviation . NI is the number of input neurons . The parameter gG determines the complexity of the behavior of the reservoir , and shows chaotic spontaneous activity if gG > 1 . The instantaneous output is given by z ( t ) = wTr ( t ) , where w is the readout weight vector . The readout unit is connected with n reservoir neurons by the readout weights w . The readout weights are modified according to the FORCE learning rule in which the error between the actual output and the teaching signal is minimized [29] . The activity of the readout unit is transmitted to the reservoir via the feedback . The initial values of the readout weights w are generated by a Gaussian distribution with the mean 0 and variance 1/n . Each element of the feedback coupling JGz is randomly sampled from a uniform distribution [-1 , +1] . In the connection matrix JGG of the reservoir , each element is taken from a Gaussian distribution with mean 0 and variance 1/ ( pNG ) , where p is the connection probability of the reservoir neurons . In the connection matrix JGI between input neurons and the reservoir , each row has only one non-zero element drawn from a normal distribution of mean 0 and variance 1 . We simulated the model with time steps of 1 [ms] . The values of parameters used in simulations are as follows: in Figs 1 , 3 and 4 and S1 Fig , S4C Fig , S4D Fig , and S5 Fig , NG = 300 , p = 1 , n = 300 and σ = 0 . 3; in Figs 2 and 6 , S2 Fig , and S3 Fig , NG = 600 , p = 0 . 5 , n = 300 and σ = 0 . 1; in S4A Fig and S4B Fig , p = 1 , σ = 0 . 3 and n = NG while the values of NG were varied; in Fig 5 , p = 1 , n = 300 , and NG = 800 , σ = 0 . 15 ( b ) or NG = 500 , σ = 0 . 1 ( e ) . The number of input neurons was NI = 26 in all simulations except S4C Fig , in which NI was 5s with s being the size of the chunk . In all simulations , τ = 10 [ms] and gG = 1 . 5 . The learning rate was set as α = 100 because larger values could cause instability in the learning process . The network was trained typically for several hundreds of seconds except in Figs 2 , 5B and 5E where the simulation time was 5000 , 2500 and 25 , 000 [s] , respectively . In S2D Fig , the teaching signals were generated as fa ( t ) =[tanh ( ( z^a′ ( t ) −γ∑c=4 , 5 , 6′yc ( t ) ) /β ) ]+ ( a=1 , 2 , 3 ) , ( 6 ) where the activities of interneurons were calculated as τyc˙ ( t ) =−yc ( t ) +z^c ( t ) . ( 7 ) A similar formula applied to the partner network . Note that a dash in the second term of Eq 6 indicates that the corresponding readout unit should be excluded from the summation ( S2C Fig ) . In S4D Fig , the weights of recurrent connections in each reservoir module and those of connections between the modules were sampled from an identical Gaussian distribution with mean 0 and variance 1/{ ( 1 + q ) NG} , where q is the connection probability of inter-module connections . The recurrent connections were all-to-all . The value 1 in the denominator was introduced such that the limit q → 0 gives the disconnected RC modules studied in other panels in S4 Fig . In our learning rule , we changed the outputs of readout units such that the mean outputs coincide with zero and the standard deviation becomes unity: z ( t ) ⟶z^ ( t ) =z ( t ) −μ ( t ) σ ( t ) , ( 8 ) where μ ( t ) and σ ( t ) were calculated as μ ( t ) =1T∫t−Ttz ( t′ ) dt′ , ( 9 ) σ ( t ) =1T∫t−Ttz ( t′ ) 2dt′−μ ( t ) 2 , ( 10 ) with a sufficiently long period T ( = 15 [s] ) . The modified output z^ ( t ) was then transformed by two nonlinear functions to generate the teaching signal shown in the Results . In S3A Fig , the activities of all reservoir neurons were first averaged and then normalized . To define the response selectivity of neurons , we sorted all of the neurons by their mean activation phases defined as , t^i=Tπarg[∑t′=1Tr¯i ( t′ ) exp ( i2πt′T ) ∑t′=1Tr¯i ( t′ ) ][ms] , ( 11 ) where r¯ ( t ) is the normalized average response of each cell and T = 2400 [ms] . Each reservoir neuron generally showed a significantly large and prolonged phasic response to a particular chunk , which determined the selectivity of the reservoir neuron . We defined a phasic response as such transient activity that exceeded the threshold value μ + 3σ for more than 100 [ms] , where μ and σ stand for the average and standard deviation of its activity during the presentation of input sequence . Neurons that were not related to any chunks or responded to multiple chunks were discarded in the analysis . In Fig 3 , we projected the neural responses rR1 ( t ) of recurrent network in R1 onto the M ( ≦NG ) dimensional subspace: rR1 , M ( t ) =VMTrR1 ( t ) . ( 12 ) Here , the ( NG×M ) -dimensional matrix VM is defined as VM = ( φ1 ( R1 ) φ2 ( R1 ) …φM ( R1 ) ) in terms of the λ-th eigenvector of R1 reservoir φλ ( R1 ) . Similarly , we projected the readout weight vectors from R1 onto the same subspace as wR1 , M=VMTwR1 . ( 13 ) We then calculated the difference between the actual output of R1 and the output reconstructed on the subspace as Ez=1T∫0T|zR1 ( t ) −wR1 , MTrR1 , M ( t ) |2dt . ( 14 ) The difference between the actual output of R2 and the projected R1-output was calculated in a similar fashion . In Fig 6B and 6C , we constructed a pair of RC modules each having two readout units . A stream of two images with high ( 97x97 pixels x 3 RGB channels ) or low ( 32x32 pixels x 3 RGB channels ) resolutions was used as input , in which the presentations of two images ( and Gaussian noise images in Fig 6B ) were randomly switched at every 250 ms . Each reservoir neuron received input from randomly chosen 10% of pixels . In Fig 6D , the low-resolution versions of the images used in Fig 6C were created at the reduced size of 32 x32 pixels ( x 3 RGB channels ) . All codes for computer simulations were written in Python 3 and are available at https://github . com/ToshitakeAsabuki/dualRC_codes .
|
Varieties of information processing require chunking , but chunking arbitrary complex sequences as flexibly as the brain does remains a challenge . In this study , we solved this important but difficult problem of chunking by "reservoir computing" inferred from brain computation . In the method , chunking occurs automatically when a pair of such modules supervising one another agree on the recurring response patterns to remember . Our model revealed a unique dynamical mechanism for embedding the temporal community structure of inputs into low-dimensional neural trajectories . From a biological viewpoint , our model based on pairwise computing suggests the computational implications of brain's bi-hemispheric information streams . Owing to recent technological advances , the implementation of this model by electronic devices should be straightforward .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"learning",
"ecology",
"and",
"environmental",
"sciences",
"neural",
"networks",
"community",
"structure",
"signaling",
"networks",
"social",
"sciences",
"teachers",
"neuroscience",
"learning",
"and",
"memory",
"cognitive",
"psychology",
"recurrent",
"neural",
"networks",
"network",
"analysis",
"cellular",
"structures",
"and",
"organelles",
"computer",
"and",
"information",
"sciences",
"animal",
"cells",
"cell",
"signaling",
"structures",
"community",
"ecology",
"people",
"and",
"places",
"professions",
"cellular",
"neuroscience",
"psychology",
"cell",
"biology",
"ecology",
"neurons",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"population",
"groupings",
"cognitive",
"science"
] |
2018
|
Interactive reservoir computing for chunking information streams
|
Variation in the gut microbiome has been linked to colorectal cancer ( CRC ) , as well as to host genetic variation . However , we do not know whether , in addition to baseline host genetics , somatic mutational profiles in CRC tumors interact with the surrounding tumor microbiome , and if so , whether these changes can be used to understand microbe-host interactions with potential functional biological relevance . Here , we characterized the association between CRC microbial communities and tumor mutations using microbiome profiling and whole-exome sequencing in 44 pairs of tumors and matched normal tissues . We found statistically significant associations between loss-of-function mutations in tumor genes and shifts in the abundances of specific sets of bacterial taxa , suggestive of potential functional interaction . This correlation allows us to statistically predict interactions between loss-of-function tumor mutations in cancer-related genes and pathways , including MAPK and Wnt signaling , solely based on the composition of the microbiome . In conclusion , our study shows that CRC microbiomes are correlated with tumor mutational profiles , pointing towards possible mechanisms of molecular interaction .
The human gut is host to approximately a thousand different microbial species consisting of both commensal and potentially pathogenic members[1] . In the context of colorectal cancer ( CRC ) , it is clear that bacteria in the microbiome play a role in human cell signaling[2–11]; for example , in the case of CRC tumors that are host to the bacterium Fusobacterium nucleatum , the microbial genome encodes a virulence factor , FadA , that can activate the β-catenin pathway[12] . In addition , several attempts have been made to predict CRC status using the microbiome as a biomarker[13–16] . It has been shown that by focusing on F . nucleatum , it is possible to predict some clinically relevant features of the tumor present[17] . There is a positive , but not statistically significant , association between F . nucleatum presence in colorectal cancers in patients who eat a Western diet[18] . Regardless , as only a minority of CRCs are host to F . nucleatum , the use of this species as a sole maker are limited[18 , 19] . Other specific microbes have been linked to CRC , including Escherichia coli harboring polyketide synthase ( pks ) islands[20 , 21] and enterotoxigenic Bacteroides Fragilis ( ETBF ) [22–24] . The mechanism of action of these associations is still under investigation with F . nucleatum being the most clearly developed[12] . A major challenge in the research on the CRC-associated microbiome is the variability of the findings by different research groups . Comparisons among and between the findings from groups studying different patient populations arrive at differing sets of microbes that appear correlated with CRC . As highlighted above , several groups have focused on individual microbial taxa to identify functional associations that explain the statistical correlations between individual microbial species and cancer . It is clear that not all of the structural and compositional changes in the tumor-associated microbiome are functionally relevant and clinically actionable . Follow-up work will need to specifically assess these interactions in a directed , empirical manner . We know that in healthy individuals , host genetic variation can affect the composition of the microbiome[25–30] , and the associated human genetic variants are enriched with cancer-related genes and pathways[26] . However , it is still unknown whether somatic mutations causing disruptions in genes and pathways in the host’s cells can affect the composition of the microbiome that directly interacts with these tissues . It is also clear that individual taxa are likely to have differential interactions with host tissues dependent on the larger context of the community as a whole ( e . g . not all patients with high levels of F . nucleatum in their gut microbiomes will develop CRC ) . It is likely that the genetic and phenotypic heterogeneity of tumors also results in differential interactions with the microbiome . This variation , if not accounted for when assessing the tumor microbiome , might allow researchers to uncover generic or widely prevalent microbial changes at the site of the tumor . Including genetic information about the cancer in the assessment of tumor-microbiota interactions would allow for a more fine-grained analysis that unmasks the subtle interactions that might be lost in a generic tumor-normal comparison . To that end , we have performed a inter-tumor analysis accounting for the genetic heterogeneity present between them . In this work we show ( i ) associations between the tumor microbiome and variation in somatic mutational profiles in CRC tumors; ( ii ) which host genes and bacterial taxa drive the association; ( iii ) how these patterns can shed light on the molecular mechanisms controlling host-microbiome interaction in the tumor microenvironment; and ( iv ) how this correlation can be used to construct a microbiome-based statistical predictor of genes and pathways mutated CRC tumors . These findings provide a framework for discovery of sets of microbial taxa ( communities ) that should be further explored using direct functional assessment .
We performed whole-exome sequencing on a set of 88 samples , comprised of 44 pairs of tumor ( adenocarcinomas ) and normal colon tissue sample from the same patient , with previously characterized tissue-associated microbiomes[2] . The mutations in each of the tumors’ protein-coding regions were identified relative to the patient-matched normal sample and annotated as either synonymous , non-synonymous , or loss-of-function ( LoF ) mutations ( S1 and S2 Tables , S1 and S2 Figs ) . The ranges of mutations per tumor in our data set were between 19–6678 ( total ) , 10–3646 ( missense ) and 0–208 ( LoF ) . The mutations were collapsed by gene as well as by pathways using both Kyoto Encyclopedia of Genes and Genomes ( KEGG ) and pathway interaction database ( PID ) annotations[31–34] ( see Materials and methods ) . We performed quality control of the data and stringent filtering at every step , with the goal of reducing false positives in mutation calling and statistical prediction ( e . g . , requiring 30x coverage at a site in both the tumor and matched normal sample to call a mutation; see Materials and methods ) . While these requirements are likely to increase the frequency of false negatives ( true mutations that simply do not meet our criteria ) , this rigorous strategy is appropriate as a means of increasing the biological relevance of our findings . Of note , when comparing the common LoF mutations found in our dataset to those found in colorectal tumors sampled as part of The Cancer Genome Atlas ( TCGA ) project , we find several commonalities , including a high frequency of LoF mutations in APC , as well as numerous missense mutations in KRAS , NRAS , and TP53 , as expected ( S1 Table ) [35] . In general , the range of mutations across our sample set were also in line with those identified as part of TCGA and other CRC exome sequencing studies ( see S15 and S16 Tables for comparisons ) [35–39] . We first investigated the relationship between microbial communities and tumor stage ( Fig 1 ) . We hypothesize that the structure and composition of the associated microbiome can be affected by relevant physiological and anatomical differences between the tumors at different stages that would provide different microenvironmental niches for microbes . We identified the changes in the microbial communities surrounding each tumor as a function of stage by grouping the tumors into low stage ( stages 1–2 ) and high stage ( stages 3–4 ) classes , due to the low number of total tumors with available stage information , and applied linear discriminant analysis ( LDA ) effect size ( LEfSe ) to the raw operational taxonomic unit ( OTU ) tables corresponding to these tumors ( S3 , S5 and S6 Tables ) [40] . The set of taxon abundances was transformed to generate a single value representing a risk index classifier for membership in the low-stage or high-stage group ( Fig 1A; see Methods ) . To ascertain the error associated with these risk indices , a leave-one-out ( LOO ) cross-validation approach was applied . We also used the LOO results to generate receiver operating characteristic ( ROC ) curves and to calculate the area under the curve ( AUC; see Fig 1B ) . In addition , we performed a permutation test to assess the method’s robustness ( S5 Table ) . Using this approach , we demonstrate that the changes in abundances of 31 microbial taxa can be used to generate a classifier that distinguishes between low-stage and high-stage tumors at a fixed specificity of 80% and an accuracy of 77 . 5% ( P = 0 . 02 by Mann-Whitney U test , and P = 0 . 007 by a permutation test; S5 Table ) . The resulting changes seen in our analysis of the microbial communities that vary by tumor stage were similar to those found in previous studies , including one using a Chinese patient cohort[4 , 41] . In both cases , there were significant changes among several taxa within the phylum Bacteroidetes , including Porphyromonadaceae , and Cyclobacteriaceae ( Fig 1 and S5 Table ) . We applied the model generated from our data to two independent datasets that were analyses of the CRC microbiome and that also reported tumor staging[10 , 42] ( S3 Fig ) . In the case of Flemer et al . the same approach successfully separated the low stage from high stage tumors , whereas the trend was the same in the case of the Yoon , et al . data , it is likely that the n-value was too low to achieve significance ( S3 Fig ) . The comparisons of interest here are between tumor samples at high stage and tumor samples at low stage . This study utilized patient-matched normal samples , but these are not true normals , as they came from the same cancer patients who themselves have low or high stage cancer . To address the question of how the CRC samples , grouped by stage , compare to independent normal samples from healthy individuals , we obtained unpublished colonic mucosal microbiome data from a separate , but otherwise methodologically similar , study being performed at the University of Minnesota that included the tissue-associated microbiome from individuals without cancer ( n = 12 ) undergoing routine colonoscopy . The OTU tables from the normal samples as well as the low-stage and high-stage samples were merged and assessed using the aforementioned LEfSe LOO approach ( see Methods ) to generate a model that is relevant to the normal and cancer samples ( S4 Fig ) . The high-stage and low-stage samples were still able to be separated , as were the differences between the normal and the low-stage . Interestingly , the high-stage samples were not able to be statistically separated from the normal samples using this model . Next , we attempted to use a similar approach to classify tumors based on mutational profiles . We initially focused on individual genes that harbor loss-of-function ( LoF ) mutations , as those , we predicted , would be the most likely to have a physiologically relevant interaction with the surrounding microbiome . We applied a prevalence filter to include only those mutations that were present in at least 10 or more patients at the gene level , including 11 genes in the analysis . The purpose of the filter was to limit the number of statistical tests while maintaining a reasonable number of possibly cancer-driving mutations; see S7 Table for how changes in this cutoff affect the number of genes included in the analysis . The raw OTU table was collapsed to the level of genus for the analysis . A visualization of the correlations between gene-level mutational status and the associated microbial abundances revealed differing patterns of abundances that suggests an interaction between the 11 most prevalent LoF tumor mutations and the microbiome ( Fig 2 and S5 Fig ) . We hypothesized that the presence of mutation-specific patterns of microbial abundances could be statistically described by prediction of tumor LoF mutations in individual genes using the microbiome . For each of eleven genes that passed prevalence filtering cutoff , we identified the associated microbial taxa ( Fig 2A , S6 and S8 Tables ) , generated risk indices for each patient ( Fig 2B and 2C ) , and plotted the mean differences in abundances for a subset of microbial taxa interacting with each mutation ( Fig 2D ) . We found that we are able to use microbiome composition profiles to predict the existence of tumor LoF mutations in the human genes APC , ANKRD36C , CTBP2 , KMT2C , and ZNF717 ( Q-value = 0 . 0011 , 0 . 0011 , 0 . 019 , 0 . 019 , and 0 . 055 , respectively , by permutation test after False Discovery Rate ( FDR ) correction for multiple tests; Fig 2 ) . The risk indices for each mutation were generated using sets of microbial taxa that ranges from 22 ( ZNF717 ) to 53 ( ANKRD36C ) taxa ( S6 Table ) . The taxa that showed the most dramatic differences in abundance when comparing tumors with and without mutations are shown in Fig 2D . For example , the abundance of Christensenellaceae is relatively lower in tumors with APC mutations , but relatively higher in tumors with ZNF717 mutations . To determine if our ability to find associations between specific mutations and microbial signatures was simply the result of being confounded by stage ( e . g . perhaps high-stage tumors all have mutations in APC , meaning that the ability to separate tumors by mutations in this gene are potentially just a function of stage ) , we used mutation presence/absence information and tumor stage to look for confounding correlations between the two using Fisher’s exact test . The results show that there is no confounding associations between mutation status and tumor stage ( S9 Table ) . Next , we applied our interaction prediction approach , as described above , to the pathway-level mutational data ( Fig 3; see Methods ) . Following visualization of the pathway level abundances ( S6 and S7 Figs ) , we found that LoF mutations in each of the 21 KEGG pathways passing prevalence filter can be significantly predicted with a fixed specificity of 80% and an accuracy up to 86% ( Q-values < 0 . 02 by permutation test after FDR correction; Fig 3A–3D , S10 Table , S8 and S9 Figs ) . Similarly , microbiome composition significantly predicted LoF mutations in 15 of the 19 tested PID pathways ( Q-values < 0 . 04 by permutation test after FDR correction ) ( Fig 3E–3H , S10 Table , S10 and S11 Figs ) . The full sets of taxon abundances that were specifically associated with each of the LoF mutations in the genes and pathways can be found in S11 , S12 , S13 and S14 Tables and S12 Fig . As an additional control for our analysis pipeline , we also took the same sets of genes , KEGG , and PID pathways and tested them using an alternative analysis tool , MaAsLin ( see Methods ) [43] . When comparing these two approaches , we found a correspondence among the LoF mutations in genes and pathways with significant associations in the microbiome ( genes: 2/5 , KEGG: 16/21 , and PID: 15/15 ) ( S10 Table ) . In general , the number of taxa within each of the sets used to generate the risk indices was lower than that used for the gene-level analyses , with an average of 37 taxa per gene compared to 7 taxa per pathway . When comparing results using the gene-level interactions and the pathway level interactions , for instance looking at mutations in APC ( Fig 2 ) and comparing them to mutations in the KEGG-defined Wnt signaling pathway and the PID-defined Canonical Wnt signaling pathway ( Fig 3 ) , the interactions at the pathway level are more statistically significant ( AUC for APC = 0 . 81 , KEGG = 0 . 88 , PID = 0 . 90 ) . This trend is consistent and can be visualized as a density histogram of interaction prediction accuracies ( S13 Fig ) , indicating a stronger signal of interaction with the microbiome when using pathway-level mutational data . Lastly , we assessed the correlations between taxa among tumors with and without LoF mutations ( Fig 4; see Methods ) . We found striking differences in the structure of the network comparing tumors with and without a Lof mutation in APC the correlations between taxa ( Fig 4A ) . For example , in tumors with mutations in APC , the abundance of Christensenellaceae is positively correlated with Rhodocyclaceae and negatively correlated with Pedobacter . In tumors lacking LoF mutations in APC , these correlations are lost and Christensenellaceae is instead negatively correlated with Saprospiraceae and Gemm 1 . We also assessed the network of correlations across tumors with mutations in PID pathways ( Fig 4B ) . This analysis highlighted that some pathway-level mutations show a shared set of correlations between taxa , while others appear independent . It is interesting to note that there are several PID pathways that biologically linked ( e . g . Degradation of beta-catenin , Regulation of nuclear b-catenin signaling , Presenilin action in Notch and Wnt Signaling , and Canonical Wnt signaling pathway ) . These pathways do indeed share core elements , but as defined by the PID authors , are comprised of distinct combinations of genes ( S14 Fig ) [34] . Several of the taxa that can be used to predict LoF mutations in p75 ( NTR ) signaling share correlations among each other as well as with taxa associated with mutations in PDGFR-beta signaling and direct p53 effectors .
The link between colorectal cancer and the gut microbiome has been highlighted by a large number of recent studies[2–17 , 19] , with several hypotheses as to the causal role of microbes in the disease[9 , 12 , 44 , 45] . Given that host genetics is associated with microbiome composition , and since cancer is a genetic disease caused by mutations in host DNA , it is of interest to study the microbiome in the context of tumor mutational profiles[25–30] . Here , we jointly analyzed tumor coding mutational profile and the taxonomic composition of the proximal microbiome . We found that the composition of the microbiome is correlated with mutations in tumor DNA , and that this correlation can be used to statistically predict mutated genes and pathways solely based on the microbiome . The association of microbial taxa with tumor stage ( Fig 1 ) mirrors recent results , including a study of a Chinese population[4 , 41] . This concordance is relevant as it indicates that the microbial communities appear to be consistent even when comparing geographically distinct patient cohorts[46 , 47] . One of the predictive taxa , Porphyromonadaceae , has been shown to be altered in mouse models of CRC in other studies as well[7 , 14] . A study on the link between dysbiosis and colitis-induced colorectal cancer also showed similar results[48] . For instance , the bacterial genus Paludibacter was found to be associated with risk of developing tumors in a mouse model[48] . Additionally , other researchers have identified this taxa as a possible contributor to inflammation in bovine mastitis[49] . We find that Paludibacter is significantly associated with low-stage tumors , again , supporting the hypothesis that these bacteria are associated with cancer risk and may be contributing to early stage inflammation[48 , 49] . An equally likely hypothesis is that Paludibacter is using the inflammatory microenvironment to its own advantage , thereby explaining its association with the tissues at this state . Conversely , we found that the genus Coprococcus is associated with high-stage tumors and not low stage tumors . As microbes can have different effects dependent on their environmental contexts , further work that studies the overall functional effects of microbial communities will need to be performed to assess how individual members interact and contribute to , suppress , or have no influence on inflammation . Gene-level mutation data , visualized in S5 Fig , show intriguing patterns of microbial abundances that are associated with the tumors harboring different mutations . While our previous work demonstrated that specific taxa , including Providencia , were significantly associated with tumors when compared to patient-matched normals , we did not find that this genus significantly discriminated among tumors based on LoF mutation status[2] . This suggests that there are potentially microbial taxa that might act as generic tumor-associated microbes and others that might rely on specific alterations in the tumor . For instance , as reflected in the differing patterns within each gene ( rows ) in the heatmap , Aerococcus and Dorea both show higher abundances within tumors harboring LoF mutations in ZNF717 , CTBP2 , and APC , relative to tumors with LoF mutations in ANKRD36C and KMT2C . This highlights the different patterns in the microbiome that can be found when assessing genetically heterogeneous sets of tumors , as Dorea has been found to be differentially present in tumor microbiomes by several different groups . Our work highlights some potential genetic interactions that may explain the differences seen ( e . g . differences in the mutational statuses of the patients in the different cohorts could result in different findings ) [3 , 5–8] . Thus , incorporating host genetic profiles in studies of the microbiome in CRC may be beneficial and uncover patterns in clearly specified subsets of patients that are defined depending on specific tumor mutations . While it is not possible to definitively identify the biological mechanism behind the predicted interactions among mutated genes and microbial taxa due to the correlative nature of this work ( shown in Fig 2 ) , it is possible to generate hypotheses based on what is already known in the relevant literature . For example , we found that LoF mutations in APC correlate with changes in 25 different microbial taxa , including an increase in the abundance of the genus Finegoldia . This genus was identified in previous studies of colon adenomas and harbors species that are opportunistic pathogens at sites of epithelial damage[6 , 50 , 51] . Capnocytophaga has been identified as a potential biomarker for lung cancer[52] . Our results also indicate that changes in the abundance of Christensenellaceae are associated with mutations in both APC and ZNF717 . A recent study in twins has identified Christensenellaceae as a taxon that is highly driven by host genetics[27] . We found that mutations in ZNF717 , a transcription factor commonly altered in gastric , hepatocellular , and cervical cancers[53–55] , are associated with Verrucomicrobiaceae and Akkermansia , which are both known to increase in abundance in conjunction with colitis[56] . Alphaproteobacteria are significant contributors to our ability to predict mutations in CTBP2 , a repressor of transcription known to interact with the ARF tumor suppressor[57] . Changes in this bacterial taxon’s abundance has also been found to be associated with prostate cancer , although the mechanism of action is unknown[58] . We also show that mutations in KMT2C , a gene commonly co-mutated along with KRAS , could be predicted , in part , using the abundance of Ruminococcus[59] . These bacteria have been previously implicated in inflammatory bowel disorders and colorectal cancer by multiple groups[8 , 60–62] . Similar results were also evident when aggregating the mutations into KEGG and PID pathways ( Fig 3 , S6 and S7 Figs; see Methods ) [31–34] . As an example , we found that the abundance of microbes that predict mutations in KEGG pathways form two distinct clusters , and that the genus Escherichia has a higher scaled abundance in tumors with mutations in the KEGG pathways in cluster 1 relative to those in cluster 2 ( S6 Fig ) . Cluster 1 contains adherens junctions , which are partially responsible for maintaining the intestinal barrier; a disruption of the intestinal barrier in mice using cyclophosphamide was shown to cause a loss of adherens junction function and a concomitant increase in bacterial translocation into the intestinal tissue , including species of Escherichia[63] . When examining the heatmap with LoF mutation collapsed into PID pathways ( S7 Fig ) , we again identified differences in scaled microbial abundances between the tumors as a function of which pathways are mutated . For instance , we found lower abundance of Pseudomonas in tumors with LoF mutations in the pathways ‘regulation of nuclear β-catenin signaling and target gene transcription’ , ‘degradation of β-catenin’ , ‘presenilin action in Notch and Wnt signaling’ , and ‘canonical Wnt signaling pathways’ . Recent studies have shown that Pseudomonas strains that express the LecB gene can lead to degradation of β-catenin , providing hypothetical support for the concept that this genus may play a somewhat protective role in CRC by suppressing the Wnt signaling pathway[64] . The mechanism that might explain this phenomenon is still unclear , but may have to do with alterations in appropriate cell surface adhesion molecules for the LecB protein or a change in the content of the cellular microenvironment[64 , 65] . Many of the interactions identified here between bacterial taxa and mutations in PID pathways have already been demonstrated experimentally in the literature . For example , in human oral cancer cells , it was shown that bacteria of interest were able to activate EGFR through the generation of hydrogen peroxide[66] . In addition , the correlation between ErbB1 downstream signaling and increase in the abundance of Corynebacterium has been demonstrated mechanistically in a model of atopic dermatitis , whereby EGFR inhibition results in dysbiosis ( the appearance of Corynebacterium species ) and inflammation[67] . Specific depletion of Corynebacterium ablates the inflammatory response[67] . Moreover , our finding that the abundance of Fusobacterium is depleted in tumors with LoF mutations in the PDGFR-beta pathway may be explained by the dependence of several pathogenic strains of bacteria for functionally intact PDGFR signaling for adherence to intestinal epithelium[68] . In addition , p75 ( NTR ) signaling has been shown to operate as a tumor suppressor by mediating apoptosis in response to hypoxic conditions and reactive oxygen species[69–72] . Alterations in this pathway have also been shown to be useful as a biomarker for esophageal cancer[73 , 74] . Our study has several caveats . First , our study only shows correlations , and we cannot directly assess causal effects . Thus , we do not know whether the microbiome is altered before or after the appearance of specific mutations . Nevertheless , many of the predicted interactions described above have been previously tested , albeit across a wide variety of experimental systems and disease states , typically in isolation , for biological relevance and mechanism of action . Additionally , we have only profiled the taxonomic composition of the microbiome , and thus cannot detect interactions that are dependent on microbial genes or functions . We also do not have an exhaustive history of each of the patients and their backgrounds ( e . g . diet , family history , immune profile , etc . ) , without which it is possible there may be a confounding variable that explains some of our correlations . Similarly , using whole-exome sequencing does not allow us to include non-coding mutations and larger tumor structural variants and chromosomal abnormalities . Moreover , the study sample was relatively small ( n = 88 samples from 44 patients ) . Nevertheless , the sample size was sufficient to detect significant patterns . As ours is the first study , as far as we know , to present simultaneous exome sequencing of tumors alongside assessment of the microbial communities associated with these tumors , no validation datasets are available . In summary , we present an association between tumor genetic profiles and the proximal microbiome , and identify tumor genes and pathways that correlate with specific microbial taxa . We also show that the microbiome can be used as a predictor of mutated genes and pathways within a tumor , and suggest potential mechanisms driving the interaction between the tumor and its microbiota . Our proof-of-principle analysis can provide a starting point for the development of diagnostics that utilize microbiome profiles to ascertain CRC tumor mutational profiles , facilitating personalized treatments .
All research conformed to the Helsinki Declaration and was approved by the University of Minnesota Institutional Review Board , protocol 1310E44403 . 88 tissue samples from 44 individuals were used , with one tumor and one normal sample from each individual . These de-identified samples were obtained from the University of Minnesota Biological Materials Procurement Network ( Bionet ) , a facility that archives research samples from patients who have provided written , informed consent . These samples were previously utilized and are described in detail in a previous study[75] . The patient information provided for this retrospective cohort did not include information related to immune status or inflammation . To reiterate these points , all research conformed to the Helsinki Declaration and was approved by the University of Minnesota Institutional Review Board , protocol 1310E44403 . Tissue pairs were resected concurrently , rinsed with sterile water , flash frozen in liquid nitrogen , and characterized by staff pathologists . The criteria for selection were limited to the availability of patient-matched normal and tumor tissue specimens . In all cases , normal tissue was confirmed by the pathologist to be tumor free . The tumor stages were classified using a modified stage grouping based on the TNM scale ( collapsing the letter-scales into single numbers—eg . IIIA , IIIB , and IIIC are reported as stage 3 . Additional patient metadata are provided in S4 Table and in the indicated work[75] . Microbial community data from 16S rRNA gene sequencing of an independent cohort of healthy patients ( n = 12 ) who underwent routine colonoscopies was acquired from a separate , unpublished dataset and used in the comparison of low-stage , high-stage , and normal microbial communities . The microbiome data used in the study was generated previously and is described exhaustively in[75] . Briefly , microbial DNA was extracted from patient-matched normal and tumor tissue samples using sonication for lysis and the AllPrep nucleic acid extraction kit ( Qiagen , Valencia , CA ) . The V5-V6 regions of the 16S rRNA gene were PCR amplified with the addition of barcodes for multiplexing using the forward and reverse primer sets V5F and V6R from Cai , et al . [76] . The barcoded amplicons were pooled and Illumina adapters were ligated to the reads . A single lane on an Illumina MiSeq instrument was used ( 250 cycles , paired-end ) to generate 16S rRNA gene sequences . The sequencing resulted in approximately 10 . 7 million total reads passing quality filtering in total , with a mean value of 121 , 470 quality reads per sample . The forward and reverse read pairs were merged using the USEARCH v7 program ‘fastq_mergepairs’ , allowing stagger , with no mismatches allowed[77] . OTUs were picked using the closed-reference picking script in QIIME v1 . 7 . 0 using the Greengenes database ( August 2013 release ) [78–80] . The similarity threshold was set at 97% , reverse-read matching was enabled , and reference-based chimera calling was disabled . Genomic DNA samples were quantified using a fluorometric assay , the Quant-iT PicoGreen dsDNA Assay Kit ( Life Technologies , Grand Island , NY ) . Samples were considered passing quality control ( QC ) if they contained greater than 300 nanograms ( ng ) of DNA and display an A260:280 ratio above 1 . 7 . Full workflow details for library preparation are outlined in the Nextera Rapid Capture Enrichment Protocol Guide ( Illumina , Inc . , San Diego , CA ) . In brief , libraries for Illumina next-generation sequencing were generated using Nextera library creation reagents ( Illumina , Inc . , San Diego , CA ) . A total of 50 ng of genomic DNA per sample were used as input for the library preparation . The DNA was tagmented ( simultaneously tagged and fragmented ) using Nextera transposome based fragmentation and transposition as part of the Nextera Rapid Capture Enrichment kit ( Illumina , Inc . , San Diego , CA ) . This process added Nextera adapters with complementarity to PCR primers containing sequences that allow addition of Illumina flow cell adapters and dual-indexed barcodes . The tagmented DNA was amplified using dual indexed barcoded primers . The amplified and indexed samples were pooled ( 8 samples per pool ) and quantified to ensure appropriate DNA concentrations and fragment sizes using the fluorometric PicoGreen assay and the Bioanalyzer High-Sensitivity DNA Chip ( Agilent Technologies , Santa Clara , CA ) . Libraries were considered to pass QC as long as they contained more than 500 ng of DNA and had an average peak size between 200–1000 base pairs . For hybridization and sequence capture , 500 nanograms of amplified library was hybridized to biotinylated oligonucleotide probes complementary to regions of interest at 58°C for 24 hours . Library-probe hybrids were captured using streptavidin-coated magnetic beads and subjected to multiple washing steps to remove non-specifically bound material . The washed and eluted library was subjected to a second hybridization and capture to further enrich target sequences . The captured material was then amplified using 12 cycles of PCR . The captured , amplified libraries underwent QC using a Bioanalyzer , and fluorometric PicoGreen assay . Libraries were considered to pass QC as long as they contained a DNA concentration greater than 10 nM and had an average size between 300–400 base pairs . Libraries were hybridized to a paired end flow cell at a concentration of 10 pM and individual fragments were clonally amplified by bridge amplification on the Illumina cBot ( Illumina , Inc . , San Diego , CA ) . Eleven lanes on an Illumina HiSeq 2000 ( Illumina , Inc . , San Diego , CA ) were required to generate the desired sequences . Once clustering was complete , the flow cell was loaded on the HiSeq 2000 and sequenced using Illumina’s SBS chemistry at 100 bp per read . Upon completion of read 1 , base pair index reads were performed to uniquely identify clustered libraries . Finally , the library fragments were resynthesized in the reverse direction and sequenced from the opposite end of the read 1 fragment , thus producing the paired end read 2 . Full workflow details are outlined in Illumina’s cBot User Guide and HiSeq 2000 User Guides . Base call ( . bcl ) files for each cycle of sequencing were generated by Illumina Real Time Analysis ( RTA ) software . The base call files and run folders were then exported to servers maintained at the Minnesota Supercomputing Institute . Primary analysis and de-multiplexing was performed using Illumina’s CASAVA software 1 . 8 . 2 . The end result of the CASAVA workflow was de-multiplexed FASTQ files that were utilized in subsequent analysis for read QC , mapping , and mutation calling . The exome sequence data contained approximately 4 . 2 billion reads in total following adapter removal and quality filtering , inclusive of forward and reverse reads , with a mean value of 47 . 8 million high-quality reads per sample . The raw reads were assessed using FastQC v0 . 11 . 2 and the Nextera adapters removed using cutadapt v1 . 8 . 1[81 , 82] . Simultaneously , cutadapt was used to trim reads at bases with quality scores less than 20 . Reads shorter than 40 bases were excluded . The trimmed and filtered read pairs were aligned and mapped to the human reference genome ( hg19 ) using bwa v0 . 7 . 10 resulting in a bam file for each patient sample[83] . These files were further processed to sort the reads , add read groups , correct the mate-pair information , and mark and remove PCR duplicates using picard tools v1 . 133 and samtools v0 . 1 . 18[84 , 85] . Tumor-specific mutations were identified using FreeBayes v0 . 9 . 14-24-gc292036[86] . Following these steps , 94 . 0% of the remaining read pairs mapped to the reference genome , hg19 . Specifically , SNPs-only were assessed and a minimum coverage at each identified mutation position of more than 30X was required in both the patient normal and tumor samples . These mutations were filtered to only include those that were within protein-coding regions which were then compiled into a single vcf file . We did not remove variants found in dbSNP . This vcf file was assessed using SNPeffect v4 . 1 K ( 2015-09-0 ) in order to predict the potential impact of each of the mutations[87] . Based on these results , the mutations were grouped into three categories: ( 1 ) total mutations ( 2 ) non-synonymous mutations and ( 3 ) loss of function ( LoF ) mutations . The total mutations group The total mutations group is simply the sum of all the mutations that were found in each tumor . The non-synonymous mutations included all the mutations in the total mutations group that were non-silent . The LoF group only included those mutations that resulted in a premature stop codon , a loss of a stop codon , or a frameshift mutation . Mutations in genes were collapsed to pathways ( PID and KEGG ) based on pathway membership as defined by the relevant database authors[31 , 33 , 34] . Specifically , we relied upon Uniprot annotations to link genes to pathways with the database file , idmapping . dat . gz , available from Uniprot[88 , 89] . This , three column database file is used by the web-based mapping interface for cross-referencing between genes and pathway membership datasets , including KEGG and PID[88] . In our case , we annotated each gene in our dataset with pathway membership information by first reducing the size of the idmapping . dat file by including only entries for KEGG and PID . For each LoF mutation in our dataset , we found the corresponding gene in the reduced database file with a matching ENSEMBL Gene ID ( S1 Table ) and recorded which grouping it was found in ( KEGG or PID ) and what the name of the defined pathway was ( e . g . Canonical Wnt signaling ) . Many genes had multiple annotations as they are members of more than one of the defined pathways . These annotations were used to identify which tumor samples had mutations in KEGG and PID pathways , defined as having at least one gene harboring a LoF mutation that was a member of the pathway . In order to identify microbial taxa that were significantly associated with specific characteristics ( i . e . tumor stage—Fig 1 , LoF mutations in individual genes—Fig 2 , and LoF mutations within functional pathways—Fig 3 ) , the OTU table was divided into two groups of tumors , defined by the characteristic of interest ( e . g . Low Stage vs . High Stage , Mutation vs . No Mutation ) . As it would be invalid to generate a model that was built using the actual test sample , we used a “leave-one-out” approach when generating risk indices . To generate a risk index for a tumor , we started with an OTU table that contained all the tumors ( 43 ) , leaving out the one for which we were calculating the risk index . This 43-tumor OTU table was divided into two groups based on the characteristic of interest , as described above . These data were used as input into LEfSe , as a means of identifying the microbial taxa that were able to discriminate between the two groupings[40] . Taxa were considered significant discriminators if the base 10 logarithm of their LDA score was less than 2 , as recommended . The taxa that were significant discriminators were themselves of two categories: those that were more abundant , for instance , at tumors that harbored a mutation and those that were more abundant at tumors that did not harbor a mutation . The relative abundances of these taxa were arcsine root transformed and these transformed values—one for each category—were summed . The difference between these sums was calculated and is the risk index value used in each of the aforementioned figures . The use of the unweighted sum in the risk index , rather than relying entirely on the regression coefficients from LDA , is a simple way to control the degree of flexibility of the model when training on small sample sizes . More detail is described in a previous publication[90] . This procedure was repeated a total of 44 times ( once for each tumor , to complete the leave-one-out approach ) to obtain a risk index for each of the patients . The significance of the difference in risk indices between the patients found in one group vs . another ( e . g . low stage vs . high stage , with LoF mutation , without LoF mutation ) was assessed using a Mann-Whitney U test and a permutation test , in which we permuted the labels for a given group 999 times , each time deriving new held-out predictions of the risk indexes for each subject for that gene . Then the observed difference in means between the patients with LoF mutation and patients with LoF mutation risk index predictions using the method on the actual LoF mutation labels to the differences observed in the permutations to obtain an empirical P-value was compared . The resulting P-values were corrected using the false discovery rate ( FDR ) correction for multiple hypothesis tests . MaAsLin v0 . 0 . 4 was used as a test for method robustness to alternative statistical parameters[43] . Without performing LOO iterations , the same sets of genes , KEGG pathways , and PID pathways that passed mutation prevalence filtering as for the LEfSe-based pipeline were used as metadata input for MaAsLin . Patient age and sex were also included in the model as covariates . The unrarefied OTU table collapsed to the genus level was used as abundance input , with prevalence filters set to only include taxa found at relative abundances of more than 0 in at least 5 tumors and an abundance filter of 0 . 001 , relative abundance . Other parameters in MaAsLin were defaults . The p-values generated by MaAslin were used to identify sets of taxa that had altered abundances in tumors as a function of mutational status . These taxa were used to generate risk indices for each of the samples , as above , but in a non-iterative fashion as the LOO procedure was not used . The RIs for the samples harboring specific LoF mutations were tested against those without a detected LoF mutation , as above . The p-values generated were corrected for multiple testing by FDR . These values are reported in S10 Table . Receiving Operating Characteristic ( ROC ) curves were plotted and the area under the curve ( AUC ) values computed on a dataset containing 10 sets of predictions and corresponding labels obtained from 10-fold cross-validation using ROCR package in R[91] . A risk index threshold was also obtained that best predicts the membership in a stage group or the presence or absence of LoF mutation with a leave-one-out cross-validation on the risk index . Each held-out sample was treated as a new patient on whom the optimal risk index cutoff was tested and subsequently refined to separate patients who had a LoF mutation and patient who did not have a LoF . Correlation analysis was performed using SparCC on a reduced OTU table containing significant taxa identified using the above prediction methods collapsed to the genus level[92] . Pseudo p-values were calculated using 100 randomized sets . Networks of correlations were visualized using Cytoscape v3 . 1 . 0[93] . For easier interpretation , final network visualization was filtered to remove families that have potential for being miscalled . The patients in this study have associated clinical data as we described previously[75] , We used a linear model to determine the extent to which clinical factors may correlate with mutation load . These included patient sex , tumor stage , patient age , patient body mass index ( BMI ) , and microsatellite instability ( MSI ) status . None of these factors , alone or in combination , were found to significantly impact the mutational data , though it bears noting that MSI status was only available for a subset ( 13 out of 44 ) of the patients .
|
Although the gut microbiome—the collection of microorganisms that inhabit our gastrointestinal tract—has been implicated in colorectal cancer , colorectal tumors are caused by genetic mutations in host DNA . Here , we explored whether various mutations in colorectal tumors are correlated with specific changes in the bacterial communities that live in and on these tumors . We find that the genes and biological pathways that are mutated in tumors are correlated with variation in the composition of the microbiome . This study provides a step towards understanding interactions between tumor genes and the microbiome in colorectal cancer .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"microbial",
"mutation",
"medicine",
"and",
"health",
"sciences",
"immune",
"cells",
"microbiome",
"antigen-presenting",
"cells",
"immunology",
"microbiology",
"cancers",
"and",
"neoplasms",
"wnt",
"signaling",
"cascade",
"oncology",
"mutation",
"mathematics",
"forecasting",
"statistics",
"(mathematics)",
"mutation",
"databases",
"frameshift",
"mutation",
"microbial",
"genomics",
"research",
"and",
"analysis",
"methods",
"animal",
"cells",
"medical",
"microbiology",
"mathematical",
"and",
"statistical",
"techniques",
"biological",
"databases",
"signal",
"transduction",
"colorectal",
"cancer",
"cell",
"biology",
"database",
"and",
"informatics",
"methods",
"genetics",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"physical",
"sciences",
"genomics",
"cell",
"signaling",
"statistical",
"methods",
"signaling",
"cascades"
] |
2018
|
Colorectal cancer mutational profiles correlate with defined microbial communities in the tumor microenvironment
|
Influenza A viruses can adapt to new host species , leading to the emergence of novel pathogenic strains . There is evidence that highly pathogenic viruses encode for non-structural 1 ( NS1 ) proteins that are more efficient in suppressing the host immune response . The NS1 protein inhibits type-I interferon ( IFN ) production partly by blocking the TRIM25 ubiquitin E3 ligase-mediated Lys63-linked ubiquitination of the viral RNA sensor RIG-I , required for its optimal downstream signaling . In order to understand possible mechanisms of viral adaptation and host tropism , we examined the ability of NS1 encoded by human ( Cal04 ) , avian ( HK156 ) , swine ( SwTx98 ) and mouse-adapted ( PR8 ) influenza viruses to interact with TRIM25 orthologues from mammalian and avian species . Using co-immunoprecipitation assays we show that human TRIM25 binds to all tested NS1 proteins , whereas the chicken TRIM25 ortholog binds preferentially to the NS1 from the avian virus . Strikingly , none of the NS1 proteins were able to bind mouse TRIM25 . Since NS1 can inhibit IFN production in mouse , we tested the impact of TRIM25 and NS1 on RIG-I ubiquitination in mouse cells . While NS1 efficiently suppressed human TRIM25-dependent ubiquitination of RIG-I 2CARD , NS1 inhibited the ubiquitination of full-length mouse RIG-I in a mouse TRIM25-independent manner . Therefore , we tested if the ubiquitin E3 ligase Riplet , which has also been shown to ubiquitinate RIG-I , interacts with NS1 . We found that NS1 binds mouse Riplet and inhibits its activity to induce IFN-β in murine cells . Furthermore , NS1 proteins of human but not swine or avian viruses were able to interact with human Riplet , thereby suppressing RIG-I ubiquitination . In conclusion , our results indicate that influenza NS1 protein targets TRIM25 and Riplet ubiquitin E3 ligases in a species-specific manner for the inhibition of RIG-I ubiquitination and antiviral IFN production .
Influenza A viruses ( IAVs ) are highly infectious pathogens that have caused major pandemics and annual epidemics with serious economic and health consequences [1] , [2] . IAVs are naturally maintained in avian species but they also circulate in humans , horses , dogs and pigs [3] . Although multigenic host range restrictions exist , a combination of viral determinants can ultimately allow a virus to establish infection in a specific host [4] . This is particularly important because , although the current highly pathogenic avian IAVs that have been transmitted to humans lack the ability to spread from human to human , there is current concern that these avian viruses may adapt and develop the ability to spread efficiently among humans . In this respect , recent studies have demonstrated that only a few mutations in the hemagglutinin ( HA ) allow for transmissibility of highly pathogenic H5N1 viruses in ferrets [5] , [6] . Moreover , pigs can be infected with human and avian viruses and provide an environment for reassortment and the generation of new influenza virus strains capable of human transmission [7] . Therefore , it is essential to better understand the mechanisms that allow influenza viruses to adapt to a new host species , in order to predict and protect from future cross-species transmission . IAV is an enveloped virus that harbors a negative-strand RNA genome encoding eleven different proteins from 8 separate segments [8] . Individual viral proteins play critical roles in species-specific pathogenicity . For example , the hemagglutinin ( HA ) protein which binds in a species-dependent manner to sialic acid on host cell membranes; the neuraminidase ( NA ) protein which is important for viral release; and the polymerase components ( PA , PB1 , PB2 ) which are important for efficient replication [9] . The non-structural protein 1 ( NS1 ) , which is the product of the smallest RNA segment , acts as a virulence factor by inhibiting host immune responses [10]–[15] . An important host innate immune mechanism is the production of type-I interferons ( IFNs ) , which can establish an antiviral state by up-regulating interferon stimulated genes ( ISGs ) that interfere with distinct steps in the viral life cycle [16] , [17] . Type-I IFNs and other pro-inflammatory cytokines are produced when host pattern recognition receptors ( PRRs ) , including Toll-like receptors ( TLRs ) and RIG-I-like receptors ( RLRs ) , recognize pathogen-associated molecular patterns ( PAMPs ) such as nucleic acids or other structural components of the microbe [18]–[21] . Viral single-stranded or short double-stranded RNA bearing a 5′-triphosphate is recognized by the cytosolic RNA helicase RIG-I [22] , [23] . In addition , next-generation RNA sequencing revealed that in influenza A virus infected cells RIG-I preferentially associated with the shorter genomic segments as well as subgenomic defective interfering particles [24] . RIG-I is composed of two N-terminal caspase recruitment domains ( CARDs ) , a central DExD/H box helicase/ATPase , and a C-terminal regulatory domain ( CTD/RD ) . Under normal conditions , RIG-I is in an inactive , closed conformation . Upon RNA recognition , exposure of the N-terminal CARDs of human RIG-I leads to Lys63-linked ubiquitination on Lys172 induced by tripartite motif 25 ( TRIM25 ) ubiquitin E3 ligase . The ubiquitinated CARDs then allow RIG-I binding to MAVS ( also called IPS-1 , Cardif , or VISA ) and thereby downstream signaling for IRF3 phosphorylation and type-I IFN production [25] . Another ubiquitin E3 ligase , Riplet ( or RNF135 ) , has also been reported to ubiquitinate RIG-I and promote its signal transducing ability for antiviral IFN production [26] , [27] . Specifically , Riplet was shown to activate RIG-I by inducing Lys63-linked ubiquitination of the lysine residues 849 and 851 in the RIG-I CTD [26] , [27] . Many viruses are equipped with specialized mechanisms to suppress the host type-I IFN system . Influenza A viruses through their NS1 protein can inhibit type-I IFN production by various mechanisms , including inhibition of RIG-I [23] , [28]–[30] and , in some cases suppressing general host gene expression by targeting the 30 kDa subunit of cleavage and polyadenylation specificity factor ( CPSF30 ) [31]–[33] . The importance of this mechanism of antagonism is further supported by the fact that some highly pathogenic viruses encode NS1 proteins that are more efficient in suppressing the host antiviral response . For example , the NS1 protein of the highly pathogenic 1918 virus blocked the expression of IFN-regulated genes more efficiently than the NS1 from influenza A/WSN/33 virus [11] . In addition , the NS1 gene of a human IAV , when passaged in mouse , accumulated adaptive mutations that resulted in increased virulence and IFN antagonism [34] . Importantly , in addition to NS1 , also other viral genes , including PA , PB1 , PB2 and PB1-F2 have been shown to exert IFN-antagonistic functions [35]–[37] . We have recently uncovered the mechanism of RIG-I inhibition by NS1 [38] . NS1 binds to the central coiled-coil ( CCD ) region of TRIM25 ubiquitin E3 ligase , blocking its oligomerization and enzymatic activity to induce Lys63-linked ubiquitination of the RIG-I CARDs . NS1 thereby inhibits RIG-I CARD-dependent downstream signaling [38] . However there are differences in PRR signaling components of different species that could impact NS1 immunosuppressive function . For example , mouse RIG-I lacks Lys172 , which has been previously shown to be essential for ubiquitination-dependent activation of human RIG-I [25] . In addition , it appears that chicken lacks a functional RIG-I gene [39] . This host specificity in the RIG-I antiviral signaling pathway may provide a barrier for new incoming viruses , which in turn will need to adapt to inhibit these host-specific immune responses . In this study , we examined the ability of NS1 encoded by human ( Cal04 ) , avian ( HK156 ) , swine ( SwTx98 ) , and mouse ( PR8 ) adapted influenza A viruses to interact with TRIM25 orthologues from mammalian and avian species , and the impact that this interaction has in regulating the RIG-I-mediated type-I IFN response . We found that while NS1 efficiently inhibits TRIM25-mediated RIG-I ubiquitination and activation in human cells , in murine cells NS1 targets the ubiquitin E3 ligase Riplet to suppress RIG-I ubiquitination and signal transducing activity . Furthermore , we identified that the NS1 proteins of human viruses have evolved to bind and antagonize both human TRIM25 and human Riplet for the potent inhibition of the RIG-I-mediated IFN response .
Given that mammalian and avian TRIM25 proteins show significant differences in their CCD ( Supplemental Figure S1A ) , that in the case of human TRIM25 is necessary and sufficient for NS1 binding [38] , we tested the ability of NS1 encoded by human ( A/California/04/09 [Cal04] ) , avian ( A/Hong Kong/156/1997 [HK156] ) , swine ( A/Swine/Texas/4199-2/98 [SwTx98] ) and mouse-adapted ( A/Puerto Rico/8/34 [PR8] ) influenza A viruses to interact with TRIM25 orthologues . To facilitate these studies , we used NS1 proteins that lack the ability to bind CPSF30 and therefore are unable to block general host gene expression [40] . Furthermore , these NS1 proteins have conserved E96/E97 amino acids previously shown to be important for interaction with human TRIM25 ( Supplemental Figure S1B ) [38] . Consistent with our previous results [38] , co-immunoprecipitation ( Co-IP ) studies in transfected HEK293T cells showed that all tested NS1 proteins efficiently interacted with human TRIM25 ( hTRIM25 ) ( Figure 1A ) . Furthermore , chicken TRIM25 ( chTRIM25 ) bound strongly to NS1-HK156 , whereas only a weak or no interaction with chTRIM25 was observed for NS1 from PR8 or Cal04 and SwTx98 , respectively ( Figure 1A ) . Moreover , none of the NS1 proteins interacted with mouse TRIM25 ( mTRIM25 ) . It has been suggested that chicken lacks a functional RIG-I gene [39] . We thus asked whether chTRIM25 is important for IFN-β induction upon IAV infection , and further tested if the chTRIM25-NS1 interaction is functional . To this end , we silenced endogenous TRIM25 in chicken LMH cells and measured IFN-β mRNA expression upon viral infection ( Figures 1B–1D ) . Specifically , we used recombinant A/PR/8/34 viruses expressing NS1 proteins from PR8 , SwTx98 , and HK156 , which showed differential binding to chTRIM25 , as well as a PR8 recombinant virus expressing the NS1 mutant R38A/K41A , known to be deficient in IFN antagonism due to its abolished dsRNA and hTRIM25 binding ability [10] , [38] , [41] . Approximately 87% knockdown of chTRIM25 mRNA levels was achieved as determined by quantitative PCR ( qPCR ) ( Figure 1B ) . Compared to mock-infected control cells , IFN-β mRNA was highly up-regulated ( ∼1500-fold ) upon infection with the NS1 mutant R38A/K41A virus ( Figure 1C ) . This IFN-β induction was significantly reduced in TRIM25-knockdown LMH cells , suggesting that TRIM25 is indeed functional in chicken cells . In addition , IFN-β induction upon infection with PR8 , SwTx98 and HK156 recombinant viruses inversely correlated with the ability of their NS1 proteins to bind chTRIM25 ( Figure 1D ) . While PR8 and HK156 were poor inducers of IFN-β mRNA , infection with SwTx98 virus , whose NS1 protein did not bind chTRIM25 , resulted in ∼15-fold IFN-β induction . Furthermore , the IFN-β induction by SwTx98 virus was reduced to almost basal levels in TRIM25-knockdown cells , indicating that this virus induces IFN-β gene expression in chicken cells by a mechanism mostly dependent on TRIM25 ( Figure 1D ) . To exclude that the low levels of IFN production induced by the PR8 and HK156 viruses are due to inefficient replication in LMH cells , we measured the mRNA levels of viral M1 by qPCR ( Figure 1E ) . All viruses replicated well , and although only minor differences were observed , the levels of M1 mRNA inversely correlated with the IFN-β inducing activity of these viruses . This suggests that TRIM25 also plays an important role in establishing an antiviral response in chicken cells . We next tested whether the inability of the NS1 proteins to bind mTRIM25 correlates with their IFN-inducing activities in mouse cells . For this , we infected murine embryonic fibroblasts ( MEF ) with the different NS1 recombinant viruses . IFN-β mRNA induction was then determined by qPCR , and IFN protein levels were measured by bioassay ( Figure 1F , left panels ) . Surprisingly , although none of the NS1 proteins bound to mTRIM25 ( Figure 1A ) , all NS1 recombinant viruses induced very low amounts of IFN in MEFs as compared to the R38A/K41A mutant virus , suggesting that NS1 does not require binding to TRIM25 for suppressing IFN induction in murine cells . Binding of NS1 to CPSF30 could also not account for this inhibitory activity in MEFs , as all of these NS1 proteins are deficient in inhibiting CPSF30 and thus general gene expression [40] . Furthermore , all viruses replicated to similar levels in the MEF cells as determined by M1 mRNA and virus titrations ( Figure 1F , right panels ) . Mice are frequently used as a model to study influenza pathogenesis . It is therefore important to understand the detailed mechanisms by which influenza virus escapes the immune response in this particular host . Thus , we first sought to confirm the lack of NS1 interaction with mTRIM25 . To rule out that the inability of NS1 to bind to mTRIM25 was due to the absence of a co-factor required for NS1-TRIM25-interaction in human cells , we tested the binding of NS1-PR8 to endogenous TRIM25 in PR8-infected human HEK293T and murine Hepa1 . 6 cells ( Figure 2A ) . NS1 efficiently interacted with endogenous TRIM25 in HEK293T cells , whereas it did not bind to endogenous TRIM25 in Hepa1 . 6 cells ( Figure 2A ) . We also observed an efficient interaction of ectopically expressed NS1 with exogenous hTRIM25 , but not with mTRIM25 , in HEK293T and Hepa1 . 6 cells ( Figure 2B ) . In line with this , bacterially purified Glutathione-S-transferase ( GST ) -NS1 fusion protein bound to hTRIM25 , but not to mTRIM25 , in an in vitro binding assay ( Figure 2C ) . The CCDs of human and mouse TRIM25 proteins share only 56% identity at the amino acid level ( aa 180–450 ) with major differences between aa 350–400 . To test if these amino acid changes in TRIM25 CCD are responsible for the difference in NS1 binding we generated a human/mouse chimeric TRIM25 protein containing amino acids 191–379 of hTRIM25 ( mChimT25h191–379 ) and analyzed its ability to interact with NS1-PR8 ( Figures 3A and 3B ) . Co-IP experiments showed that while mTRIM25 did not interact with NS1 , hTRIM25 and the mChimT25h191–379 efficiently bound NS1 ( Figure 3B ) . As previously reported [38] , overexpressed NS1 localizes primarily to the nucleus with a minor cytoplasmic component; however , co-expressed hTRIM25 led to a marked increase of NS1 cytoplasmic localization , indicative of hTRIM25-NS1 interaction . As shown in Figures 3C and 3D , ectopic expression of hTRIM25 or mChimT25h191–379 markedly re-localized NS1 to the cytoplasm , where they extensively co-localized ( 92% and 80% of cells with cytoplasmic NS1 , respectively ) . In contrast , co-expression of mTRIM25 that did not interact with NS1 showed only a minor effect on NS1 sub-cellular localization ( Figures 3C and 3D ) . These results collectively demonstrate that amino acid changes in the CCD region of mouse TRIM25 are responsible for the loss of interaction with NS1 . Influenza A viruses lacking functional NS1 proteins induce a robust IFN response , and thus replicate poorly in immunocompetent mice , indicating that NS1 inhibits induction of IFN in mice [42] . However , while in human cells NS1 interacts with human TRIM25 for inhibition of RIG-I ubiquitination and activation [38] , the same mechanism cannot apply to mouse cells , as NS1 does not interact with mouse TRIM25 . In order to gain more insights on the impact of NS1 on mouse RIG-I activation we examined the effect of NS1 on the ubiquitination of human or mouse RIG-I 2CARD in human HEK293T or mouse Hepa1 . 6 cells ( Figure 4A ) . Consistent with our previous results [38] , NS1-PR8 potently inhibited the ubiquitination of GST-fused human RIG-I 2CARD ( GST-hRIG-I 2CARD ) in HEK293T cells in a dose-dependent manner . In striking contrast , NS1 did not block the ubiquitination of mouse GST-RIG-I 2CARD ( GST-mRIG-I 2CARD ) in Hepa1 . 6 cells ( Figure 4A ) . Furthermore , we analyzed the effect of NS1 on the Sendai virus ( SeV ) -induced ubiquitination of full-length RIG-I in human HEK293T and murine Hepa1 . 6 cells ( Figure 4B ) . Interestingly , exogenously expressed NS1-PR8 effectively suppressed the ubiquitination of full-length RIG-I in a dose-dependent manner in both cell lines ( Figure 4B ) . These results collectively suggest that while NS1 efficiently inhibits the CARD ubiquitination of human RIG-I , it suppresses the ubiquitination of mouse RIG-I by acting on its C-terminal region . Lys172 in the CARDs of human RIG-I ( hRIG-I ) is essential for RIG-I ubiquitination and activation by TRIM25 [25] . However , this residue is not conserved in mouse RIG-I ( mRIG-I ) ( Supplemental Figure S2 ) , raising the question whether mRIG-I is ubiquitinated by mTRIM25 as previously reported to occur for human RIG-I . Strikingly , ubiquitination of GST-h2CARD was undetectable in TRIM25 −/− MEFs , whereas it was robustly ubiquitinated in wild-type ( WT ) MEFs ( Figure 5A ) . In contrast , GST-m2CARD showed a decreased but still detectable ubiquitination in TRIM25 −/− MEFs compared to WT MEFs ( Figure 5A ) . Reconstitution of TRIM25 −/− MEFs with mouse or human TRIM25 showed an increase in both human and mouse RIG-I 2CARD ubiquitination ( Figure 5B ) , indicating that mTRIM25 is active and is able to ubiquitinate the mouse RIG-I CARDs under these conditions at a different Lys residue/s . Interestingly , NS1 inhibited the hTRIM25-dependent ubiquitination of human or mouse RIG-I 2CARD but was unable to inhibit mTRIM25-dependent ubiquitination of human or mouse RIG-I 2CARD ( Figure 5C ) . In support of this , NS1 co-immunoprecipitated with RIG-I 2CARD only when hTRIM25 was present , but not in the presence of ectopically expressed mTRIM25 ( Figure 5C ) . Since NS1 potently inhibited the ubiquitination of full-length mRIG-I ( Figure 4B ) , we next tested if NS1-PR8 interacts with endogenous RIG-I in non-complemented TRIM25 −/− MEFs , or in TRIM25 −/− MEFs reconstituted with either human or mouse TRIM25 ( Figure 5D ) . This showed that NS1 readily co-immunoprecipitated with endogenous RIG-I in TRIM25 −/− MEFs , and this interaction was further increased in the presence of exogenous human but not mouse TRIM25 ( Figure 5D ) . Furthermore , exogenously expressed NS1-PR8 strongly suppressed the SeV-induced ubiquitination of endogenous RIG-I in TRIM25 −/− cells in a dose-dependent manner ( Figure 5E ) , indicating that NS1 binds to mouse RIG-I and inhibits its ubiquitination in a TRIM25-independent manner . We next determined the ubiquitination of endogenous RIG-I in WT and TRIM25 −/− MEFs upon PR8 virus infection ( Supplemental Figure S4A ) . This showed that RIG-I ubiquitination was decreased in TRIM25 −/− MEFs compared to WT MEFs . Furthermore , PR8 infection markedly reduced the ubiquitination of endogenous RIG-I in both WT MEFs and TRIM25 −/− MEFs ( Supplemental Figure S4A ) . Finally , we examined the ability of NS1-PR8 to inhibit IFN-β promoter activation in WT MEFs or TRIM25 −/− MEFs . As expected , NS1 inhibited SeV-induced IFN-β promoter activation in WT MEFs ( Figure 5F ) . TRIM25 −/− MEFs showed a markedly reduced IFN-β promoter activity upon SeV infection as compared to WT MEFs; however there was still detectable IFN-β induction by SeV stimulation . This residual IFN-β induction in TRIM25 −/− MEFs was further suppressed by exogenous NS1 ( Figure 5F ) , demonstrating that NS1 inhibits the IFN induction in mouse cells by a TRIM25 independent mechanism . Collectively , these results demonstrate that human TRIM25 is the major ubiquitin E3 ligase for human RIG-I 2CARD and that NS1 inhibits its activity . Our results further indicate that in mouse cells there is an alternative ubiquitin E3 ligase that cooperates with TRIM25 for full activation of mouse RIG-I , and that this alternative ubiquitin E3 ligase is probably targeted by NS1 . Riplet ( also known as RNF135 ) has recently been shown to be important for IFN production in mice [27] and to ubiquitinate human RIG-I at Lys849 and Lys851 in its C-terminal domain [26] . Since these lysine residues are conserved in mouse RIG-I ( Supplemental Figure S3 ) we hypothesized that NS1 may bind mouse Riplet ( mRiplet ) , thereby suppressing RIG-I ubiquitination and downstream signaling in mouse . As shown in Figure 6A , NS1-PR8 readily bound exogenously expressed mRiplet in HEK293T cells in the presence or absence of SeV infection . We also observed an efficient interaction of NS1 with mRiplet in Hepa1 . 6 cells infected with the PR8 strain ( Figure 6B ) . Immunofluorescence assays further showed that , as seen with human TRIM25 , co-expressed mRiplet re-localized NS1 from the nucleus to the cytoplasm; approximately 90% of the cells exhibited primarily cytoplasmic NS1 ( Figure 6C ) . Furthermore , exogenously expressed NS1 of Cal04 and HK156 strains also efficiently interacted with mRiplet in murine Hepa1 . 6 cells ( Supplemental Figure S4B ) . In addition , Hepa1 . 6 cells infected with various recombinant A/PR/8/34 viruses expressing NS1 proteins from PR8 , Cal04 , HK156 , Tx91 , SwTx98 , and Pan99 , also showed an interaction between NS1 and mRiplet ( Figure 6D ) , indicating that this effect is not exclusive of the NS1 protein of the mouse-adapted PR8 strain . Finally , the NS1 E96A/E97A mutant that lacks the ability to bind human TRIM25 [38] was also able to interact with mRiplet in transfected and infected cells , whereas the NS1 R38A/K41A mutant , lacking both dsRNA and hTRIM25 binding , did not bind to mRiplet under the same conditions ( Figure 6E and Supplemental Figure S4C ) . We next asked whether NS1 specifically suppresses the Riplet-mediated ubiquitination of RIG-I in murine cells . For this , we examined the endogenous RIG-I ubiquitination in murine Hepa1 . 6 cells that were transfected with empty vector or Flag-mRiplet together with or without NS1-PR8 ( Figure 7A ) . This showed that exogenous expression of mRiplet markedly enhanced the ubiquitination of endogenous RIG-I , and this ubiquitination was potently suppressed by co-expression of NS1 ( Figure 7A ) . Consistent with this , ectopic expression of mRiplet led to a 4 . 6-fold increase of mRIG-I-induced IFN-β promoter activation , and this RIG-I activation was inhibited almost to basal levels by co-expressed NS1 ( Figure 7B ) . Finally , we sought to determine the role of endogenous Riplet in mouse RIG-I ubiquitination , and also addressed whether NS1 specifically antagonizes the mRiplet-induced RIG-I ubiquitination . For this , we examined the ubiquitination of RIG-I in Hepa1 . 6 cells in which endogenous Riplet was silenced using specific siRNA ( Figure 7C ) . 78% knockdown efficiency was achieved as determined by RT-PCR analysis compared to non-silencing control siRNA ( data not shown ) . Specific knockdown of endogenous Riplet strongly decreased endogenous RIG-I ubiquitination ( Figure 7C ) . Furthermore , co-expressed NS1 efficiently blocked the ubiquitination of RIG-I in cells transfected with non-silencing control siRNA , while NS1 had no effect on the residual ubiquitination of RIG-I observed in Riplet-knockdown cells ( Figure 7C ) . These results collectively demonstrate that Riplet is an important ubiquitin E3 ligase for mouse RIG-I , and that NS1 specifically targets Riplet to inhibit RIG-I ubiquitination in murine cells . To further demonstrate that Riplet contributes to the establishment of an efficient antiviral response against influenza virus , we silenced endogenous Riplet in WT or TRIM25 −/− MEFs using specific siRNA , and assessed influenza PR8 virus replication ( Figures 7D–7F ) . The knockdown of mRiplet was confirmed by RT-PCR ( Figure 7D ) . Infection of TRIM25 −/− cells with PR8 virus resulted in higher plaque forming units ( PFU ) as compared with WT MEF cells , underscoring the important role of TRIM25 for an efficient antiviral response in mouse . Furthermore , siRNA-mediated knockdown of endogenous Riplet in both WT and TRIM25 −/− cells yielded significantly higher levels of viral replication compared to transfection of non-silencing siRNA , with the strongest effect observed in TRIM25 −/− MEFs in which Riplet was silenced ( Figure 7E ) . The increased virus replication in Riplet-knockdown MEFs correlated with increased NS1 protein levels in these cells ( Figure 7F ) . Taken together , these data demonstrate that both Riplet and TRIM25 contribute to an effective antiviral response in mouse . Mice are not a natural host of influenza viruses , but the fact that all NS1 proteins tested interacted with mouse Riplet , including non-mouse adapted NS1 proteins ( avian or swine ) suggests that this is a conserved evasion mechanism across influenza viruses . Therefore , we tested if human Riplet interacts with the different NS1 proteins in the context of viral infection . Interestingly , the NS1 from human influenza viruses ( Tx91 and Pan99 ) interacted with human Riplet ( Figure 8A ) , suggesting that the capacity of NS1 binding to mouse Riplet might reflect an interaction of NS1 with Riplet derived from its specific host . To further demonstrate that the interaction of NS1 with human Riplet is biologically relevant , we compared the capacity of Tx91-NS1 and PR8-NS1 recombinant viruses to inhibit the ubiquitination of human RIG-I ( Figure 8B ) . We reasoned that Tx91-NS1 that binds both hTRIM25 and hRiplet , should be more efficient in inhibiting RIG-I ubiquitination ( and IFN induction ) than PR8-NS1 shown to bind only hTRIM25 . As predicted , Tx91-NS1 recombinant virus was more efficient in inhibiting human RIG-I ubiquitination as compared to PR8-NS1 virus ( Figure 8B ) . Accordingly , Tx91-NS1 recombinant virus induced significantly lower levels of IFN-β mRNA as compared to PR8 virus ( Figure 8C ) . Finally , we analyzed IFN-β production in PR8- and Tx91-infected A549 cells upon knockdown of TRIM25 , Riplet or both ( Figures 8D and 8E ) . siRNA-mediated knockdown reduced TRIM25 and Riplet mRNA levels by approximately 90% and 80% , respectively . Furthermore , the knockdown was highly specific , as targeting TRIM25 did not reduce Riplet mRNA levels and vice-versa ( Figure 8D ) . Silencing of endogenous TRIM25 in PR8-infected cells resulted in a slight but not significant decrease of IFN-β induction as compared to cells transfected with non-silencing control siRNA ( Figure 8E ) . In contrast , IFN induction by PR8 was significantly reduced in Riplet-knockdown cells as compared to controls , and almost to the same extent as the levels induced by the Tx91 virus in control cells . These results suggest that while PR8 inhibits IFN induction by targeting TRIM25 , the residual IFN induction ( not inhibited by the virus ) depends mostly on Riplet . Conversely , knockdown of either TRIM25 or Riplet did not have a significant effect on IFN induction in Tx91-infected cells . A significant reduction in IFN-β mRNA levels was only observed in cells in which both TRIM25 and Riplet were silenced ( Figure 8E ) . Collectively , these results indicate that both TRIM25 and Riplet are important for IFN-β production in human cells upon influenza virus infection . Furthermore , while PR8-NS1 mainly targets TRIM25 in human cells , Tx91-NS1 interacts with and suppresses both TRIM25 and Riplet for the inhibition of IFN production in human cells .
In this study we investigated the mechanisms by which the NS1 protein of influenza A viruses affects the TRIM25/RIG-I-mediated type-I IFN response in different host species ( Figure 9 ) . We demonstrated that NS1 proteins from avian and mammalian isolates are capable of binding to human , but not mouse TRIM25 . In mouse cells , NS1 protein inhibits RIG-I ubiquitination and downstream IFN promoter activation primarily by a mechanism that involves binding to and inhibiting the ubiquitin E3 ligase Riplet . This mechanism thus clearly differs from the previously described mechanism for NS1 antagonism of TRIM25 in human cells . As mice are not natural hosts of influenza virus infection , we reasoned that the ability of influenza virus NS1 to inhibit mouse Riplet most likely reflects its inherent ability to inhibit Riplet from other animal species . Indeed , our study showed that the NS1 protein of human viruses efficiently bound and suppressed human Riplet , indicating that specifically human viruses have evolved to inhibit both TRIM25 and Riplet for the suppression of RIG-I antiviral activity in human cells . Interestingly , Riplet is highly similar to TRIM25 sharing 60 . 8% sequence homology . Like TRIM25 , Riplet contains an N-terminal RING domain and a C-terminal SPRY domain . However , in contrast to TRIM25 , Riplet does not contain a B-box domain and thus it does not belong to the TRIM protein family [26] . Importantly , bioinformatic analysis also predicted the presence of a central CCD in Riplet , with structural similarities to the TRIM25 CCD ( data not shown ) . Further studies are needed to identify the precise binding site of NS1 in Riplet , and whether this interaction requires the Riplet CCD . Furthermore , our interaction studies showed that an avian NS1 protein ( HK156 ) strongly and preferentially interacted with chicken TRIM25 . Although RIG-I has been reported to be absent in chicken cells [39] , the fact that avian NS1 efficiently interacted with chicken TRIM25 and that the HK156 virus was a poor inducer of IFN in chicken cells , suggests that this interaction is likely to be functional . In addition , we observed higher levels of IFN induction in chicken cells by SwTx98 recombinant virus correlating with the inability of its NS1 protein to bind chicken TRIM25 . Knockdown of TRIM25 in chicken cells resulted in reduced IFN-β induction in infected cells , further evidencing that TRIM25 is active and that NS1 targets its function in chicken cells . Importantly , there are no Riplet sequences reported for the chicken or duck genomes , suggesting that influenza A viruses that are maintained in avian species have adapted to coexist with their host by targeting TRIM25 , but not Riplet ( Figure 9 ) . Further studies will be required to elucidate the target of TRIM25 in chicken , and whether the same principles apply to the natural avian reservoir of influenza viruses , migratory ducks , which are known to have RIG-I [39] . By studying RIG-I ubiquitination and signal-transducing activity in murine cells , we have also uncovered commonalities and differences between human and mouse RIG-I regulation . Although Lys172 , the critical residue for RIG-I ubiquitination in human [25] , is not conserved in mouse , we observed a robust ubiquitination of mouse RIG-I 2CARD , suggesting that other lysine residues in the mouse RIG-I CARDs are ubiquitinated . Furthermore , our studies using TRIM25 −/− cells indicate that the ubiquitination of mouse RIG-I 2CARD , in contrast to that of human RIG-I 2CARD , is less affected in the absence of TRIM25 , indicating that in addition to TRIM25 another ubiquitin E3 ligase mediates RIG-I 2CARD ubiquitination in mouse . The ubiquitin E3 ligase Riplet has been reported to ubiquitinate RIG-I , thereby leading to its optimal activation [26] , [27] , [43] . However , while Riplet has been shown to ubiquitinate Lys849 and Lys851 in the CTD of RIG-I , the role of Riplet for RIG-I CARD ubiquitination has been a subject of debate [26] , [27] , [43] . Our data suggests that , at least in the mouse system , Riplet is unlikely to ubiquitinate the 2CARD , because NS1 was unable to inhibit mouse RIG-I 2CARD ubiquitination . Therefore , in addition to TRIM25 there may be an additional ubiquitin E3 ligase in mice that ubiquitinates the RIG-I N-terminal CARDs , and this E3 ligase is most likely not targeted by NS1 . Nevertheless , our studies clearly showed that Riplet plays a crucial role for RIG-I ubiquitination-dependent activation in mouse , most likely by ubiquitinating the C-terminal region of RIG-I . Additional work will be needed to further define the roles and contributions of these ubiquitin E3 ligases for RIG-I activation in mouse . For our interaction studies , we used NS1 proteins of influenza viruses that are not able to bind CPSF30 . This facilitates the interpretation of the results , as their expression is not associated with inhibition of general host protein expression . However , we have also shown that the NS1 proteins of the Tx91 and Pan99 strains , which were previously shown to interact with CPSF30 [40] , are also able to interact with human and mouse Riplet in infected cells ( Figures 6D and 8A ) . This suggests that NS1 binding to Riplet or CPSF30 is not mutually exclusive . The NS1 mutant E96A/E97A was previously reported to lack human TRIM25 binding but retained dsRNA binding [38] . Our results showed that the E96A/E97A NS1 mutant , but not the R38A/K41A mutant , efficiently interacted with mouse Riplet , suggesting that the region in NS1 important for dsRNA binding is needed for interaction with mouse Riplet . Alternatively , the interaction between NS1 and mouse Riplet could be mediated by viral RNA . In our previous study [38] , we also showed that the PR8 recombinant virus carrying the E96A/E97A NS1 mutant was able to induce IFN in WT and TRIM25 +/− MEFs , but it lost this IFN-inducing capacity in TRIM25 −/− cells . In contrast , the R38A/K41A NS1 recombinant virus and a PR8 virus lacking NS1 ( ΔNS1 PR8 virus ) retained some IFN-inducing capacity in TRIM25 −/− cells [38] . Although we cannot completely rule out that in our experiments a low affinity interaction between NS1 and mouse TRIM25 undetectable in our system occurs , the fact that the E96A/E97A mutant virus was able to repress the residual IFN induction in TRIM25 −/− cells [38] implies that this virus can , at least in part , block IFN induction in mouse in a TRIM25-independent manner . This probably occurs by binding dsRNA and/or by interacting with other host factors including mouse Riplet . Our work raises an important question: why does NS1 target Riplet instead of TRIM25 in mouse cells ? Our results combined with previous studies [25]–[27] , [43] suggest that although both TRIM25 and Riplet contribute to optimal activation of both mouse and human RIG-I , their requirement might be different in human and mouse cells . Our experiments in TRIM25 −/− cells showed that whereas human RIG-I 2CARD ubiquitination is undetectable in TRIM25 −/− cells , mouse RIG-I 2CARD showed residual ubiquitination under these conditions . In addition , knockdown of endogenous Riplet in mouse cells strongly decreased RIG-I full-length ubiquitination . Taken together , this suggests that while TRIM25 is the primary E3 ligase for human RIG-I ubiquitination , in mouse other ubiquitin E3 ligases including Riplet may play a crucial role for RIG-I ubiquitination-dependent activation . Under this scenario , in mouse cells NS1 must target Riplet for efficient inhibition of RIG-I signaling ( Figure 9 ) . The mechanistic differences by how influenza virus NS1 inhibits the RIG-I pathway in human and mouse cells indicate that caution should be used to appropriately interpret future experiments on host responses during influenza virus infection using the mouse model . Our study also identified two NS1 proteins from human viruses ( Tx91 and Pan99 ) which efficiently bound to human Riplet . Furthermore , a recombinant virus expressing the Tx91 NS1 protein was more efficient in inhibiting endogenous RIG-I ubiquitination and IFN induction in human cells as compared to PR8 WT virus , whose NS1 protein does not bind human Riplet . These results are in agreement with the high capacity of Tx91 virus to inhibit the antiviral state in human dendritic cells [44] . Combined with our knockdown analyses in A549 cells ( Figures 8D and 8E ) , this suggests that human viruses such as Tx91 have evolved to interact with both TRIM25 and Riplet in human cells for the comprehensive inhibition of the RIG-I-induced IFN response ( Figure 9 ) . Finally , although we have used representative strains from different hosts , we cannot exclude that not all strains from a specific host have the observed specificity of antagonism . Future studies are needed to test NS1 proteins of various different strains for their ability to inhibit TRIM25 and Riplet from multiple hosts , and whether this correlates with the ability of these viruses to suppress the IFN response in the respective host . Taken together , our work suggests that influenza viruses can readily adapt to circumvent the type-I IFN system in different host species by targeting host-specific proteins involved in IFN induction . It stresses the importance of host-specific signaling pathways and immune responses when choosing animal models for studying influenza pathogenicity . Understanding the changes that viruses undergo during adaptation may provide us with better tools to predict potential future pandemics and develop vaccines and antivirals .
HEK293T ( human ) , HeLa ( human ) , A549 ( human ) , Hepa1 . 6 ( mouse ) , Vero , and L929 cells were purchased from the American Type Culture Collection ( ATCC ) . WT and TRIM25 −/− MEFs were previously described [25] . All cells were maintained in Dulbecco's Modified Eagle's Medium ( DMEM ) supplemented with 10% fetal bovine serum ( FBS ) , 2 mM L-glutamine and 1% penicillin-streptomycin ( Gibco-BRL ) . The LMH ( chicken ) cells were purchased from ATCC and cultured in flasks pre-coated with 0 . 1% gelatin in DMEM supplemented with 10% FBS , 2 mM L-glutamine and 1% penicillin-streptomycin . Transient transfections were performed with TransIT-LT1 ( Mirus ) , calcium phosphate ( Clontech ) or Lipofectamine 2000 ( Invitrogen ) according to the manufacturers' instructions . Mammalian expression constructs for untagged NS1 under the control of the chicken β-actin promoter ( pCAGGS vector [45] ) have been described previously for A/Puerto Rico/8/34 ( PR8 ) and R38A/K41A mutant NS1 [41] ) , A/Swine/Texas/4199-2/98 ( Sw/Tx/98 ) , A/Puerto Rico/8/34 E96A/E97A mutant , and A/Hong Kong/156/1997 [38] , and GST-PR8 and A/California/04/09 NS1 ( Cal04 ) [40] . Human TRIM25-V5 , human RIG-I and human GST-RIG-I 2CARD were described in [38] . Mouse TRIM25 was cloned from murine LA-4 cells by Superscript III one-step RT-PCR ( Invitrogen ) with forward primer ( 5′GAATTCACCATGGGCAAACCGATTCCGAACCCGCTGCTGGGCCTGGATAGCACCGCGGAGCTGAATCCTCTGGCC-3′ ) and reverse primer ( 5′CCTCTCTATCTGCTCCAAATAGGGATCC-3′ ) . The forward primer includes an EcoRI site and V5 sequence , and the reverse primer includes a BamHI site . V5-tagged mouse TRIM25 cDNA was then subcloned into the EcoRI/BglII sites of the mammalian expression vector pCAGGS . Chicken TRIM25 was cloned from primary chicken embryo fibroblasts by Superscript III one-step RT-PCR with forward primer ( 5′- GAATTCACCATGGGCAAACCGATTCCGAACCCGCTGCTGGGCCTGGATAGCACCGCGACGTTGACCAAAGCTCAGTC-3′ ) and reverse primer ( 5′- GGCACTGTTCTCTCTCTCTGGTAACTCGAG-3′ ) . The forward primer includes an EcoRI site and V5 sequence , and the reverse primer includes a XhoI site . V5-tagged chicken TRIM25 cDNA was then subcloned into the same sites of pCAGGS . For the generation of V5-tagged human/mouse TRIM25 chimera , the coiled-coil region of TRIM25 ( aa191–379 ) was amplified using human TRIM25-V5 plasmid as template , and then ligated into the BstBI sites of the mouse TRIM25 sequence in the pCAGGS vector . The Myc/DDK-tagged pCMV6 mouse Riplet was purchased from Origene ( Rockville , MD ) . The reporter plasmid carrying the firefly luciferase ( FF-Luc ) gene under the control of the IFN-β promoter ( p125Luc ) was kindly provided by Takashi Fujita ( Kyoto University , Japan ) [46] . The reporter plasmid carrying the Renilla luciferase gene ( REN-Luc/pRL-TK ) was purchased from Promega , WI . All sequences were verified by sequencing . Sendai virus ( SeV; Cantell strain ) was obtained from Charles River Laboratories and propagated in 10-day old embryonated chicken eggs . The A/Puerto Rico/8/1934 virus was propagated in 8-day old , specific pathogen-free embryonated eggs ( Charles River Laboratories; North Franklin , CT ) . Influenza virus titers were determined by plaque assay using MDCK cells . Rescue plasmids encoding the NS segments of A/Swine/Texas/4199-2/98 and A/Hong Kong/156/1997 , A/Texas/36/1991 and A/Panama/2007/1999 were generated by cloning the cDNA into the ambisense rescue plasmid pDZ [47] . The rescue plasmid encoding A/California/4/2009 NS was described previously [48] . Isogenic influenza A viruses ( A/Puerto Rico/8/1934 ) harboring the NS segment of A/Swine/Texas/4199-2/98 or A/Hong Kong/156/1997 were rescued by reverse genetics as described elsewhere [38] , [49] . The A/Puerto Rico/8/1934 viruses expressing the NS1 R38A/K41A or NS1 E96A/E97A mutant were described previously [38] . All viruses were amplified in 8-day old , specific pathogen-free embryonated hen's eggs ( Charles River Laboratories; North Franklin , CT ) . Interferon bioassays were conducted as described previously [38] , [50] . Immortalized mouse embryonic fibroblasts ( MEF ) , seeded in 12-well plates , were infected with the indicated isogenic influenza viruses at a multiplicity of infection ( MOI ) of 2 . At 12 h postinfection , supernatants were harvested and 2-fold dilutions of the influenza virus supernatants prepared . Diluted supernatants were then applied to bioassay using immortalized mouse fibroblast ( L929 ) cells; At 20 h post-treatment , L929 cells were infected with VSV-GFP at an MOI of 2 . GFP fluorescence was determined by fluorescence plate reader . Fluorescence values are reported on the y-axis as relative fluorescence units . HEK293T , Hepa1 . 6 and MEF cells were lysed in NP40 buffer ( 50 mM HEPES , pH 7 . 4 , 150 mM NaCl , 1% [v/v] NP40 and protease inhibitor cocktail [Roche] ) or RIPA buffer ( 50 mM TRIS-HCl pH 8 . 0 , 150 mM NaCl , 1% [v/v] NP40 , 0 . 5% sodium deoxycholate , 0 . 1% SDS , and protease inhibitor cocktail [Roche] ) . GST pull-down and immunoprecipitations were performed as described previously [25] . For immunoblotting , proteins were resolved by SDS-polyacrylamide gel electrophoresis ( SDS-PAGE ) and transferred onto a PVDF membrane ( Immobilon-P Millipore or BioRad Laboratories ) . The following primary antibodies were used: anti-V5 ( 1∶2 , 000 ) ( Invitrogen ) , anti-Flag ( 1∶5 , 000 ) ( Sigma ) , anti-HA ( 1∶5 , 000 ) ( Sigma ) , anti-GST ( 1∶2 , 000 ) ( Sigma ) , anti-TRIM25 ( monoclonal; 1∶2 , 000 ) ( BD Biosciences ) , anti-TRIM25 ( polyclonal; 1∶1 , 000 ) ( Santa Cruz Biotechnology ) , anti-RIG-I ( 1∶1 , 000 ) ( Alme-1 , Alexis ) , anti-NS1 ( 1∶2000 ) [38] , anti-β-actin ( Abcam; Cambridge , MA ) . Immunoblots were developed with the following secondary antibodies: ECL anti-rabbit IgG horseradish peroxidase conjugated whole antibody from donkey , and ECL anti-mouse IgG horseradish peroxidase conjugated whole antibody from sheep ( GE Healthcare; Buckinghamshire , England ) . The proteins were visualized by an enhanced chemiluminescence reagent ( Pierce ) . HeLa cells were seeded into Lab-Tek II 8-well chamber slides ( CC2 Glass slide , Nunc; Rochester , NY ) . After 12–16 h , plasmids harboring V5-tagged TRIM25 , Myc-mRiplet , and/or NS1 plasmids were transfected with Lipofectamine 2000 ( Invitrogen ) at a ratio 1∶1 . Six hours posttransfection the medium was changed . 24 h later , cells were washed with PBS , fixed with methanol-acetone ( 1∶1 ) , permeabilized with 0 . 5% NP-40 ( v/v ) in PBS , and blocked with 0 . 5% BSA 0 . 2% fish gelatin in PBS . Cells were stained with anti-V5 or anti-Myc and anti-NS1 antibodies as well as DAPI . Secondary antibodies conjugated to Alexa-fluor 488 and Alexa-fluor 555 ( Invitrogen ) were used to visualize the proteins . Images were taken on a Leica SP5 DM confocal microscope ( Leica Microsystems ) at a magnification of 63× . Confocal laser scanning microscopy was performed at the MSSM-Microscopy Shared Resource Facility . Hepa1 . 6 cells , WT MEFs and TRIM25 −/− MEFs were transfected in 24-well plates ( Falcon , Becton Dickinson , NJ ) with 50 ng of IFN-β reporter plasmid , 20 ng of Renilla luciferase and 10 ng NS1 plasmids using Lipofectamine 2000 at a ratio 1∶2 . 24 h later , cells were lysed and dual-luciferase assay was performed according to the manufacturer's instructions ( Promega , Madison , WI , USA ) . Luciferase values were normalized to Renilla values , and the fold induction was calculated as the ratio of SeV-stimulated samples , or samples transfected with inducing plasmid versus samples transfected with empty plasmid ( no stimulation ) . Recombinant human TRIM25 ( described in [25] ) and mouse TRIM25 purified from XL1blue bacteria as amino-terminal Maltose-Binding Protein ( MBP ) and carboxy-terminal Flag fusion proteins ( MBP-h/mTRIM25-Flag ) were incubated with bacteria-produced recombinant GST or GST-NS1 ( PR8 ) in 50 mM Tris HCl , pH 7 . 4 , 150 mM NaCl , and 0 . 1% NP40 . TRIM25-NS1 protein complexes were then precipitated with Amylose agarose resin ( New England Biolabs ) , and the binding reaction mix was incubated for 4 h at 4°C . Precipitated protein complexes were resolved by SDS-PAGE and visualized by Coomassie staining . Transient knockdown of endogenous Riplet in mouse Hepa 1 . 6 and MEF cells , seeded in 6-well plates , was achieved by transfection of siGenome SMARTpool siRNA specific for mouse Riplet ( Dharmacon ) with Lipofectamine and Plus reagent ( Invitrogen ) according to the manufacturer's instructions . A final concentration of 600 nM of mouse Riplet SMARTpool siRNA , or 600 nM of siGenome non-targeting siRNA ( Dharmacon ) was used per well . Riplet knockdown efficiency was determined by RT-PCR using specific primers . Knockdown of endogenous human TRIM25 and Riplet in A549 cells was performed in 24-well plates by transfection with Lipofectamine RNAiMAX ( Invitrogen ) . 10 picomols of siGenome SMARTpool siRNA specific for human TRIM25 or Riplet , or non-targeting control siRNA ( Dharmacon ) was used . 40 h later , cells were infected with the different influenza viruses at an MOI of 0 . 1 for 36 h . Cells were then harvested for qPCR analysis . Knockdown of endogenous TRIM25 in chicken LMH cells ( hepatocellular carcinoma cell line ) , seeded in 24-well plates , was achieved by transfection of a siRNA specific for chTRIM25 5′-GCGAGAUUUGCUGAGAGCUGAGUUU-3′ ( Invitrogen ) . As control , LMH cells were transfected with a non-targeting control sequence 5′-AAGGACGCUGAGGCCUAAUCCUGUU-3′ ( Invitrogen ) . Total RNA was isolated using RNeasy kit ( Qiagen ) and subjected to DNAse digestion with Turbo DNase ( Ambion ) . Reverse transcription was performed using the high capacity cDNA reverse transcription kit ( Applied Biosystems ) . Real-time RT-PCR was performed in 384-well plates in triplicate using gene-specific primers and Lightcycler480 SYBR green I master mix ( Roche ) in a Roche LightCycler 480 . Threshold cycles were calculated using the 2nd derivative max value of every amplification curve . Relative mRNA abundances were calculated using the ΔΔCt method [51] using 18S rRNA as a reference and plotted as fold change compared to mock-control samples . Statistical analysis was performed using Prism ( Version 5 . 0 , GraphPad Software , San Diego California USA ) . Student's paired t-test , or in defined cases Two-way ANOVA with Bonferroni post-test or Fisher's exact test were used . *p<0 . 05; **p<0 . 01; ***p<0 . 001 .
|
Influenza viruses cause annual epidemics and occasionally , major global pandemics . To establish productive infection these viruses have mechanisms to evade host immune responses , including the type-I interferon ( IFN ) response . An important component of the IFN system is the helicase RIG-I that recognizes viral RNA , and is subsequently ubiquitinated by TRIM25 ubiquitin E3 ligase to induce downstream signaling resulting in IFN-α/β production . The NS1 protein of influenza A viruses binds to human TRIM25 and inhibits TRIM25-dependent RIG-I ubiquitination and downstream RIG-I signaling . An important unresolved question is how viruses can inhibit the RIG-I pathway when infecting new hosts . Here we show that while human TRIM25 is able to bind to different NS1 proteins , chicken TRIM25 binds preferentially to the NS1 from an avian virus . Strikingly , mouse TRIM25 was unable to bind NS1 . We found that NS1 blocks RIG-I signaling in mouse and human cells by different mechanisms . While NS1 inhibits human TRIM25-mediated RIG-I ubiquitination , in mouse cells NS1 suppresses RIG-I signaling by binding to and inhibiting the ubiquitin E3 ligase Riplet . These results help understand the immune evasion strategies used by influenza virus in different species , and may partly explain the ability of this virus to adapt to different host species .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"medicine",
"infectious",
"diseases",
"avian",
"influenza",
"a",
"viruses",
"zoonoses",
"rna",
"viruses",
"viral",
"immune",
"evasion",
"immunity",
"virology",
"innate",
"immunity",
"viral",
"classification",
"biology",
"microbiology"
] |
2012
|
Species-Specific Inhibition of RIG-I Ubiquitination and IFN Induction by the Influenza A Virus NS1 Protein
|
Over the past decade , several targeted therapies ( e . g . imatinib , dasatinib , nilotinib ) have been developed to treat Chronic Myeloid Leukemia ( CML ) . Despite an initial response to therapy , drug resistance remains a problem for some CML patients . Recent studies have shown that resistance mutations that preexist treatment can be detected in a substantial number of patients , and that this may be associated with eventual treatment failure . One proposed method to extend treatment efficacy is to use a combination of multiple targeted therapies . However , the design of such combination therapies ( timing , sequence , etc . ) remains an open challenge . In this work we mathematically model the dynamics of CML response to combination therapy and analyze the impact of combination treatment schedules on treatment efficacy in patients with preexisting resistance . We then propose an optimization problem to find the best schedule of multiple therapies based on the evolution of CML according to our ordinary differential equation model . This resulting optimization problem is nontrivial due to the presence of ordinary different equation constraints and integer variables . Our model also incorporates drug toxicity constraints by tracking the dynamics of patient neutrophil counts in response to therapy . We determine optimal combination strategies that maximize time until treatment failure on hypothetical patients , using parameters estimated from clinical data in the literature .
Chronic Myeloid Leukemia ( CML ) is an acquired hematopoietic stem cell disorder leading to the over-proliferation of myeloid cells and an increase in cellular output from the bone marrow that is often associated with splenomegaly . The most common driving mutation in CML is a translocation between chromosomes 9 and 22 that produces a fusion gene known as BCR-ABL . The BCR-ABL protein promotes proliferation and inhibits apoptosis of myeloid progenitor cells and thereby drives expansion of this cell population . By targeting the BCR-ABL oncoprotein , imatinib ( brand name Gleevec ) is able to induce a complete cytogenetic remission in the majority of chronic phase CML patients . A minority of patients , however , either fail to respond or eventually develop resistance to treatment with imatinib [1] . It is thought that a primary driver of this resistance to imatinib is point mutations within the BCR-ABL gene . A recent study utilizing sensitive detection methods demonstrated that a small subset of these mutations may exist before the initiation of therapy in a significant fraction of patients , and that this status is correlated with eventual treatment failure [2] . Second generation agents such as dasatinib and nilotinib have been developed and each has shown efficacy against various common mutant forms of BCR-ABL . This leads to the observation that the various mutant forms of BCR-ABL result in CML that have unique dynamics under therapy , and that combinations of these inhibitors may be necessary to effectively control a rapidly evolving CML population . Patients with CML often die due to transformation of the disease into an acute form of leukemia known as blast crisis . It has been shown that blast crisis is due to the accumulation of additional mutations in CML progenitor cells [3] . The goal of this work is to leverage the differential responses of CML mutant strains to design novel sequential combination treatment schedules using dasatinib , imatinib and nilotinib that optimally control leukemic burden and delay treatment failure due to preexisting resistance . We develop and parametrize a mathematical model for the evolution of both wild-type ( WT ) CML and mutated ( resistant ) CML cells in the presence of each therapy . Then we formulate the problem as a discrete optimization problem in which a sequence of monthly treatment decisions is optimized to identify the temporal sequence of imatinib , dasatinib and nilotinib administration that minimizes the total CML cell population over a long time horizon . There has been a significant amount of work done in the past to mathematically model CML . For example , in [4] the authors developed a system of ordinary differential equations ( ODEs ) that model both the normal progression from stem cell to mature blood cells and abnormal progression of CML . A hierarchical system of differential equations was used to model the response of CML cells to imatinib therapy in [5]; this model fit the biphasic nature of decline in BCR-ABL positive cells during imatinib treatment . In [6] the authors investigated the number of different resistant strains present in a newly diagnosed chronic phase CML patient . An optimal control approach was utilized to optimize imatinib scheduling in [7] . Particularly relevant to our work is [8 , 9] where the authors investigated simultaneous continuous administration of dasatinib , nilotinib and imatinib; in particular , they explored the minimal number of drugs necessary to prevent drug resistance . In the current work , we focus on understanding the optimal administration schedule of multiple therapies to prevent resistance , and studying the impact of toxicity constraints on optimal scheduling . Since several of the available tyrosine kinase inhibitors ( TKI ) share similar toxicities ( in particular neutropenia , see e . g . , [10–12] ) combining them together can lead to elevated risk of adverse events . Thus we consider sequential combination therapies in which only one TKI may be administered at a time . Moreover , it has been shown that the risk of treatment failure and blast crisis are highest within the first 2 years from diagnosis [1] . Therefore it is possible that optimized , sequential single agent therapy may be sufficient to minimize the risk of treatment failure . Allowing only one treatment at a time leads to a complex , time-dependent discrete optimization problem . Another line of research closely related to the current work is the use of optimal control techniques in the design of optimal temporally continuous drug concentration profiles ( see , e . g . , review articles [13 , 14] and the textbook [15] ) . In this field the tools of optimal control such as the Pontryagin principle and the Euler-Lagrange equations are used to find drug concentration profiles that result in minimal tumor cell populations under toxicity constraints . Particularly relevant to the current work is [16] where the authors searched for optimal anti-HIV treatment strategies . They dealt with the similar problem of treating heterogeneous populations with multiple drugs . One major drawback of these works is the fact that it is nearly impossible to achieve a specific optimal continuous drug-concentration profile in patients , since drug concentration over time is a combined result of a treatment schedule ( e . g . sequence of discrete oral administrations ) and pharmacokinetic processes in the body including metabolism , elimination , etc . Thus the clinical utility of an optimal continuous drug concentration profile is limited . In contrast to these previous works , here we model the optimization problem as a more clinically realistic sequence of monthly treatment decisions . Imposition of this fixed discrete set of decision times leads to a challenging optimization problem . Such dynamical systems are referred to as ‘switched nonlinear systems’ in the control community [17] , and our problem additionally imposes fixed switching times . In this work we will leverage the system structure and tools from mixed-integer linear optimization [18] to solve this problem numerically , resulting in optimal therapy schedules that are easy to implement in practice .
We consider an ODE model of the differentiation hierarchy of hematopoietic cells , adapted from [5 , 19 , 20] . Stem cells ( SC ) on top of the hierarchy give rise to progenitor cells ( PC ) , which produce differentiated cells ( DC ) , which in turn produce terminally differentiated cells ( TC ) . This differentiation hierarchy applies to both normal and leukemic cells [21] . We consider in our model leukemic WT cells as well as preexisting BCR-ABL mutant cell types . We use type 1 , type 2 , and type i ( 3 ≤ i ≤ n ) cells to denote normal , leukemic WT , and ( n − 2 ) leukemic mutant cells; layer 1 , 2 , 3 , 4 cells to denote SC , PC , DC , and TC; and drug 0 , 1 , 2 , 3 to denote a drug holiday , nilotinib , dasatinib , and imatinib , respectively . Let xl , i ( t ) denote the abundance of type i cell at layer l and time t , and x ( t ) = ( xl , i ( t ) ) be the vector of all cell abundance at time t . If drug j ∈ {0 , 1 , 2 , 3} is taken from month m to month m + 1 , then the cell dynamics are modeled by the following set of ODEs . x ˙ ( t ) = f j ( x ( t ) ) , t ∈ [ m Δ t , ( m + 1 ) Δ t ] , x ( m Δ t ) = x m , for some function fj , where Δt = 30 days and xm is the cell abundance at the beginning of month m . The concrete form of function fj under drug j is described as follows . SC level x ˙ 1 , i = ( b 1 , i j ϕ i - d 1 , i j ) x 1 , i , i = 1 , … , n PC level x ˙ 2 , i = b 2 , i j x 1 , i - d 2 , i j x 2 , i , i = 1 , … , n DC level x ˙ 3 , i = b 3 , i j x 2 , i - d 3 , i j x 3 , i , i = 1 , … , n TC level x ˙ 4 , i = b 4 , i j x 3 , i - d 4 , i j x 4 , i , i = 1 , … , n . ( 1 ) See Fig 1 for an illustration of the differentiation hierarchy of hematopoietic cells , including neutrophils as part of the TC . Here we describe the function of each parameter of this model . For a detailed discussion of how these parameters were estimated from biological data , please see Section A of S1 Text . Type i stem cells divide at rate b 1 , i j per day under drug j . The production rates of type i progenitors , differentiated cells and terminally differentiated cells under drug j are b l , i j per day for l = 2 , 3 , 4 , respectively . The type i cell at layer l dies at rate d l , i j per day under drug j , for each i , l and j . The competition among normal and leukemic stem cells is modeled by the density dependence functions ϕi ( t ) , where ϕ i ( t ) = 1 / ( 1 + p i ∑ i = 1 n x 1 , i ( t ) ) for each i; these functions ensure that the normal and leukemic stem cell abundances remain the same once the system reaches a steady state . The parameter p1 ( resp . p2 ) is computed from the equilibrium abundance of normal ( resp . leukemic WT ) stem cells assuming only normal ( resp . leukemic WT ) cells are present , and we set pi = p2 for each i ≥ 3 . In particular , p 1 = ( b 1 , 1 0 / d 1 , 1 0 - 1 ) / K 1 and p 2 = ( b 1 , 2 0 / d 1 , 2 0 - 1 ) / K 2 , with K1 ( resp . K2 ) being the equilibrium abundance of normal ( resp . leukemic WT ) stem cells assuming only normal ( resp . leukemic WT ) cells are present .
Note the total leukemic cell abundance at day t is given by ∑l≥1∑i≥2 xl , i ( t ) . The OTP can be formulated as the following mixed-integer optimization problem with ODE constraints . min ∑ l ≥ 1 ∑ i ≥ 2 x l , i ( M Δ t ) ( 2 ) s . t . x ˙ ( t ) = ∑ j = 0 3 z m , j f j ( x ( t ) ) , t ∈ [ m Δ t , ( m + 1 ) Δ t ] , m = 0 , 1 , … , M - 1 , ( 3 ) ∑ j = 0 3 z m , j = 1 , m = 0 , 1 , … , M - 1 , ( 4 ) z m , j ∈ { 0 , 1 } , j = 0 , 1 , 2 , 3 , m = 0 , 1 , … , M - 1 , ( 5 ) x ( 0 ) = x 0 . ( 6 ) To summarize the previous display , in eq ( 2 ) we state that our objective is to minimize the leukemic cell population at the end of the treatment horizon . In eq ( 3 ) we stipulate that the cell dynamics are governed by the system of differential equations given by eq ( 1 ) . Together eqs ( 4 ) and ( 5 ) stipulate that during each time period we administer either one drug or no drug . The OTP problem is a mixed-integer nonlinear optimization problem , in which some constraints are specified by the solution to a nonlinear system of ODEs eq ( 3 ) . This optimization problem is beyond the ability of state-of-the-art optimization software . However , if we assume the TKI therapies do not affect the stem cell compartment , then it is possible to handle the ODE constraints numerically . This is because the non-linearities in the ODE model are only present in the stem cell compartment , and the remaining compartments are modeled by linear differential equations . Thus we are able to build a refined linear approximation to the ODE constraints ( see Section C of S1 Text ) , and recast the problem as a mixed-integer linear optimization problem , which can be solved efficiently for the size of our problem ( see Section B of S1 Text ) . It should be noted that our model does not explicitly consider the phenomena of TKI resistance acquired during therapy . Our model suggests a method for designing optimized treatment plans , beyond monotherapies , at the beginning of treatment . In addition , this optimization procedure can be re-run and modified during the course of treatment , with updated inputs from each patient’s response to the treatment—including the presence of acquired mutations . Below we summarize our notation for the ease of the reader .
We first utilize the model to demonstrate the dynamics of CML populations with preexisting BCR-ABL mutations under monotherapy with the standard therapies imatinib , dasatinib and nilotinib . Recall that the standard dosing regimens are 300mg twice daily for nilotinib , 100mg once daily for dasatinib , and 400mg once daily for imatinib [22] . The birth rate parameters for each leukemic mutant type under different TKIs ( b l , i j for j = 1 , 2 , 3 , l = 1 , 2 , 3 , 4 , and i ≥ 3 ) in the model are estimated using in vitro IC50 values reported in [23] for each drug . The initial cell populations at the start of therapy are derived by running the model starting from clonal expansion of a single leukemic cell in a healthy hematopoietic system at equilibrium [19] until CML detection ( when the total leukemic burden reaches approximately 1012 cells [24] ) . At this point the total cell burden is 2–3 times the normal cell burden in a healthy individual and thus the total leukemic cells make up approximately 77% of the total cell population; this is consistent with clinical reports [25] . Details on deriving the initial cell abundances at diagnosis are provided in Section A of S1 Text . In the first example we consider a patient harboring a low level of the BCR-ABL mutant F317L before the initiation of TKI therapy . According to the in vitro IC50 value reported in [23] , F317 is resistant to dasatinib , and moderately resistant to nilotinib and imatinib . The initial population conditions are given in Table 1 with the leukemic WT and F317L cells taking up 95% and 5% of the leukemic cells , respectively . We plot in Fig 2 the cell dynamics over 120 months for four treatment plans: ( 1 ) nilotinib monotherapy ( 2 ) dasatinib monotherapy , ( 3 ) imatinib monotherapy , ( 4 ) no therapy—control . We observe that as predicted , the disease burden responds well to imatinib and nilotinib; the percentage of cancerous cells after a 24 month treatment drops to 0 . 19% with nilotinib and 0 . 26% with imatinib , respectively . However , the F317L mutant population is fairly resistant to dasatinib; we observe that the percentage of cancerous cells after 24 months is 58 . 1% with dasatinib and 95 . 4% with no treatment . Over the 120 month period dasatinib treatment provides only modest improvement over the ‘no drug’ option in controlling the F317L population; however , dasatinib remains quite effective in controlling the WT leukemic population . It is interesting to note that overall , nilotinib is the most effective in controlling both the WT and F317L leukemic populations . However , nilotinib also negatively impacts the healthy cell population more severely than imatinib , which is slightly less effective in controlling the leukemic populations . This suggests that some trade-offs between these drugs exist , and these trade-offs may be exploited in designing combination therapies . In the next example we consider a patient with BCR-ABL mutant type M351T preexisting therapy . In contrast to the previous example , this commonly occurring mutant has been found to be partially sensitive in varying degrees to all three therapies . The initial conditions are given in Table 2 . Once again we have assumed that WT and M351T cells take up 95% and 5% of total leukemic cells , respectively . In Fig 3 the cell dynamics over 120 months for the four standard treatment plans are plotted: ( 1 ) nilotinib monotherapy ( 2 ) dasatinib monotherapy , ( 3 ) imatinib monotherapy , ( 4 ) no therapy—control . Since the M351T mutant is responsive to each drug in contrast to the previous example , the percentage of cancerous cells after a 24 month treatment drops to 0 . 18% with nilotinib , 0 . 18% with dasatinib , and 0 . 25% with imatinib , respectively . Without treatment , the percentage of cancerous cells after 24 months is 95 . 4% . Here , we observe that although nilotinib is more effective than dasatinib in controlling the total mutant M351T burden , the effect is reversed in the progenitor population . Higher levels of stem and progenitor populations will lead to faster rebound during treatment breaks , suggesting another trade-off to consider in the combination setting . We next solve the discrete optimization problem to identify sequential combination therapies utilizing imatinib , dasatinib and nilotinib to optimally treat CML patients with preexisting BCR-ABL mutations . We consider schedules in which a monthly treatment decision is made between one of four choices: imatinib , dasatinib , nilotinib , and drug holiday . During months in which one of the three drugs is administered , the dosing regimen is fixed at 300mg twice daily for nilotinib , 100mg once daily for dasatinib , and 400mg once daily for imatinib . In the following we optimize over feasible treatment decision sequences that result in a minimal leukemic cell burden after 3 years . Each treatment plan is completely characterized by a temporal sequence of drugs over a long time horizon . The extent to which TKIs affect leukemic stem cells is currently unknown . Several studies have demonstrated that these cells may in fact be resistant to TKIs , see e . g . [27] and references therein . Based on these results we have assumed throughout this work that TKIs do not affect the leukemic stem cell population . Given that there is uncertainty regarding this issue , in this subsection we investigate how our optimal schedules will perform if we assume TKIs impact the leukemic stem cells . In Tables 9 and 10 , we look at the leukemic cell burden after a 36-month treatment with our optimal therapy and the best monotherapy for preexisting M351T and F317L , respectively , assuming that each TKI reduces the production rate of leukemic stem cells by a certain level . In both tables we observe that the optimized schedule outperforms the monotherapy even if leukemic stem cells are susceptible to TKI . These result imply that the optimal treatment schedules we derive here outperform a traditional monotherapy regardless of the impact of TKIs on leukemic stem cells .
In this work we have considered the problem of finding optimal treatment schedules for the administration of a variety of TKIs for treating chronic phase CML . We modeled the evolution of wild-type and mutant leukemic cell populations with a system of ordinary differential equations . We then formulated an optimization problem to find the sequence of TKIs that lead to a minimal cancerous cell population at the end of a fixed time horizon of 36 months . The 36-month therapeutic horizon is clinically meaningful since it appears that the risk of therapeutic failure and disease progression to blast crisis is highest within the first two years from diagnosis [1] . At first glance the optimization problem studied in this work ( OTP ) is quite challenging . It is a mixed-integer nonlinear optimization problem , where the nonlinear constraints are specified by the solution to a nonlinear system of differential equations . However , one factor mitigating the complexity of the problem is the assumption that the TKIs do not effect the stem cell compartment . This has the effect of making the evolution of the stem cell compartment independent of the TKI schedule chosen . In addition , the remaining layers in the cellular hierarchy are modeled by linear differential equations . We can thus numerically solve the differential equation governing the stem cell layer , and treat this function as an inhomogeneous forcing term in the linear differential equation governing the progenitor cells . This allows us to approximate the nonlinear constraints specified by the differential equations by linear constraints with high accuracy . Then the OTP problem can be approximated by a mixed-integer linear optimization problem , which we are able to solve with state-of-the-art optimization software CPLEX [28] within one hour . Importance of minimizing progenitor cell population . We first aimed to minimize leukemic cell burden at 36 months after initiation of therapy , starting with an initial leukemic population of wild-type CML cells and either M351T ( sensitive to all three therapies ) or F317L ( resistant to dasatinib ) mutant leukemic cells . For both starting mutant populations , we observed that the optimal schedule involves initiating therapy with dasatinib and later switching to nilotinib , although the timing of the switch differed . To further understand this result , we noted that within this parameter regime , dasatinib is the most effective of the three TKIs at controlling leukemic progenitor cells , while nilotinib is the most effective at controlling the differentiated cells , which comprise most of the total leukemic burden . Thus , we note that controlling the leukemic progenitor cell population is important in long-term treatment outcome . This is further supported by the observation that blast crisis emerges due the acquisition of additional mutations in CML progenitor cells ( not stem cells ) [3] . Our approach suggests that using combination TKI therapies may be a viable method of controlling this population . Our modeling suggests that it is best to reduce the progenitor cells early and then reduce the differentiated cells towards the end of the treatment planning horizon . An early reduction in progenitor cells pays off in later stages of the treatment planning horizon , since a small progenitor cell population results in a lower growth rate for differentiated cells which leads to a greater response to subsequent TKI therapy . Effects of toxicity constraints . We also imposed a toxicity constraint on therapy optimization procedure by mandating that patient ANC levels stay above a given threshold that reduces the risk of infections . We observed that incorporating this toxicity constraint does impact the structure of the optimal schedules significantly , resulting in mandated treatment breaks as well as switching some months to imatinib therapy , which has a lower toxicity effect . We also noted that the choice of treatment breaks occurring also in one-month intervals may result in dangerous rebound of leukemic burden to levels close to pre-treatment , suggesting that shorter breaks to combat toxicity may be recommended . Although the model we have used for describing the dynamics of the ANC levels is simple , our findings demonstrate that incorporating a mechanistically modeled toxicity constraint into optimization of therapy scheduling is both feasible and important in determining optimal scheduling . Multiple preexisting mutant types . While some previous studies have suggested that the majority of CML patients are diagnosed with 0 or 1 preexisting BCR-ABL mutations , some patients do harbor multiple mutants at the start of therapy [2 , 6] . Thus we also investigated the impact of having 2 mutant types present ( M351T and F317 , or E255K and F317L ) at the start of therapy , on optimal schedules . We observed the number and specific combination of preexisting mutants present can significantly impact the optimization results . This points to the importance of determining which BCR-ABL mutations preexist in patients at diagnosis , before treatment planning is done . Throughout this work we have observed that the structure of the optimal therapy depends heavily on model parameters , e . g . , cellular growth rates and ANC decay rates . It is likely that each individual patients will have unique model parameters , and therefore a unique best schedule . An exciting application of this work would be the development of personalized optimal therapeutic schedules . Determination of ( i ) the mutant types ( if any ) present in a patient’s leukemic cell population , ( ii ) growth kinetics of their leukemic cell populations , and ( iii ) patient ANC level responses under various TKIs , would enable our optimization framework to build treatment schedules in a patient-specific setting . At the start of treatment for acute lymphoblastic leukemia ( ALL ) , Quantitative RT-PCR or similar techniques are sometimes used to perform mutational analysis to identify the preexisting mutant types . Indeed , studies have demonstrated that BCR-ABL mutants are present at the time of diagnosis in many ALL patients , and as sequencing technologies improve , smaller and smaller subclones with resistant phenotypes will likely be discovered [29] . In addition , the ANC level can be tested weekly by a routine complete blood count and in principle , the growth kinetics of each leukemic cell type could be analyzed in the laboratory using standard techniques such as flow cytometry or quantitative imaging . However , very few laboratories currently have the ability to analyze the growth kinetics of each type of leukemic cell population in a reproducible way . We believe that the testing procedure needs to be standardized before it becomes helpful for the treatment of CML . If such a standardized growth kinetic analysis can be realized , during treatment a patient’s response to various TKIs could be monitored so that the impact of TKI on growth kinetics of leukemic cell population and ANC levels could be quantified . Then the optimization procedure could be re-run on the fly with updated patient parameters , providing dynamic feedback into each patient’s optimal therapy schedule .
|
Targeted therapy using imatinib , nilotinib or dasatinib has become standard treatment for chronicle myeloid leukemia . A minority of patients , however , fail to respond to treatment or relapse due to drug resistance . One primary driving factor of drug resistance are point mutations within the driving oncogene . Laboratory studies have shown that different leukemic mutants respond differently to different drugs , so a promising way to improve treatment efficacy is to combine multiple targeted therapies . We build a mathematical model to predict the dynamics of different leukemic mutants with imatinib , nilotinib and dasatinib , and employ optimization techniques to find the best treatment schedule of combining the three drugs sequentially . Our study shows that the optimally designed combination therapy is more effective at controlling the leukemic cell burden than any monotherapy under a wide range of scenarios . The structure of the optimal schedule depends heavily on the mutant types present , growth kinetics of leukemic cells and drug toxicity parameters . Our methodology is an important step towards the design of personalized optimal therapeutic schedules for chronicle myeloid leukemia .
|
[
"Abstract",
"Introduction",
"Model",
"of",
"CML",
"dynamics",
"Treatment",
"optimization",
"problem",
"Results",
"Discussion"
] |
[
"medicine",
"and",
"health",
"sciences",
"pathology",
"and",
"laboratory",
"medicine",
"cancer",
"treatment",
"cell",
"differentiation",
"toxicology",
"oncology",
"optimization",
"toxicity",
"developmental",
"biology",
"mathematics",
"pharmaceutics",
"stem",
"cells",
"animal",
"cells",
"stem",
"cell",
"therapy",
"differential",
"equations",
"clinical",
"genetics",
"cell",
"biology",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"physical",
"sciences",
"drug",
"therapy"
] |
2016
|
Optimized Treatment Schedules for Chronic Myeloid Leukemia
|
The dynamics of infectious diseases are greatly influenced by the movement of both susceptible and infected hosts . To accurately represent disease dynamics among a mobile host population , detailed movement models have been coupled with disease transmission models . However , a number of different host movement models have been proposed , each with their own set of assumptions and results that differ from the other models . Here , we compare two movement models coupled to the same disease transmission model using network analyses . This application of network analysis allows us to evaluate the fit and accuracy of the movement model in a multilevel modeling framework with more detail than established statistical modeling fitting methods . We used data that detailed mobile pastoralists’ movements as input for 100 stochastic simulations of a Spatio-Temporal Movement ( STM ) model and 100 stochastic simulations of an Individual Movement Model ( IMM ) . Both models represent dynamic movement and subsequent contacts . We generated networks in which nodes represent camps and edges represent the distance between camps . We simulated pathogen transmission over these networks and tested five network metrics–strength , betweenness centrality , three-step reach , density , and transitivity–to determine which could predict disease simulation outcomes and thereby be used to correlate model simulation results with disease transmission simulations . We found that strength , network density , and three-step reach of movement model results correlated with the final epidemic size of outbreak simulations . Betweenness centrality only weakly correlated for the IMM model . Transitivity only weakly correlated for the STM model and time-varying IMM model metrics . We conclude that movement models coupled with disease transmission models can affect disease transmission results and should be carefully considered and vetted when modeling pathogen spread in mobile host populations . Strength , network density , and three-step reach can be used to evaluate movement models before disease simulations to predict final outbreak sizes . These findings can contribute to the analysis of multilevel models across systems .
Host movement influences disease dynamics [1–3] . In situations where the population size is too small to sustain long-term chains of transmission , movement of infected hosts can spark local outbreaks and increase incidence counts . Alternatively , migration of susceptible hosts away from outbreak locations can reduce abundance or local density of the susceptible population , reduce contact rates , and decrease incidence counts or outbreak duration [1] . In order to quantify and predict pathogen transmission in a mobile host population , researchers couple mathematical and statistical models of disease dynamics with host movement models to more accurately represent local abundance of diseased hosts in a multilevel modeling framework [4–9] . The problem still remains , however , of how to best represent host movement since host behavior is complex and full movement trajectories are rarely observed . To account for heterogeneity in host movement , multiple models have been proposed . These models range from stochastic models of movement , such as the set of random walk models [10] , to more mechanistic models of movement [11] , including distance-based and gravity-based movement models [12 , 13] . Each movement model has its own set of assumptions and simulated movement trajectories can differ . When different movement models are coupled with disease transmission models [4–9] the movement model selected can impact results and conclusions relating to disease dynamics . It is important that the movement model used in a multilevel modeling framework with a disease transmission model accurately represent contacts that result in disease transmission . Statistical methods have been developed to resolve this specific problem and help find the most parsimonious movement model ( reviewed in [14] and [15] ) . One approach is to develop multilevel models that incorporate alternative hypothesized modes of movement and use statistical techniques to find which model fits the data best . For example , one multilevel model would include a movement model at level one and a disease transmission model at level two; another multilevel model would include a different movement model at level one and the same disease transmission model at level two . Using classical methods , one could find which multilevel model best fits the data using a goodness of fit method–e . g . , maximum likelihood–and infer that whichever movement model best represents movement patterns in the host population [15] . This approach works well when host location and disease incidence data are detailed enough to accurately select the most parsimonious model and when the outcome desired is the selection of the best-fit model at the population level . However , when model fit varies through time or by host , additional methods are needed for model evaluation and selection . Here , we extend traditional methods of model selection by developing a suite of network analysis tools for multilevel disease transmission models that incorporate host movement . Networks consist of nodes , which represent hosts , that are connected by edges , or contacts among nodes through which pathogens propagate . This network approach has three advantages . First , networks permit flexibility in representing heterogeneous behavior that can affect disease transmission and overcome homogeneous behavior assumed in some traditional models [16] . Second , networks permit flexibility in representing host-pathogen systems–as the edge can represent any relevant mode of transmission–and surmount the need to develop specific methodology for each pathogen transmission scenario [17–19] . Third , networks can be implemented in a way that displays and measures effects of the movement level in a multilevel model even if these effects vary through time or vary by host . Together , these advantages highlight the usefulness of network analysis for model evaluation and selection . To demonstrate our network analysis tools , we use livestock data from the Far North Region , Cameroon . This Region hosts many important livestock diseases–including five different serotypes of foot-and-mouth disease virus [20 , 21]–making disease quantification and prediction important here . Host movement shapes disease transmission , because 7% of the estimated 650 , 000 cattle in this area participate in seasonal transhumance as a livestock management practice [22] . During transhumance , herders move their camps , which consist of mixed-species livestock herds and households , over great distances ( >100 km ) . While pastoralists follow the same general patterns of movement [23] , detailed spatiotemporal characteristics vary between herds . Each pastoralist makes a decision about the location of the campsite and the dates of arrival and departure and considers multiple factors for each move [24] , suggesting that movement is a complex process . To account for the uncertainty regarding in pastoralist movements in the Far North Region , Cameroon , we developed two movement models that represent camp movement and resulting contacts as dynamic processes . The first model is a spatial-temporal mobility model ( STM ) that estimates the probability of pastoralist movements based on detailed movement survey data [25] . This model categorizes pastoralists into three groups based on the shape of their movement trajectories and assigns corresponding probability distributions to herd locations [26] . The second model is called Individual Movement Model ( IMM ) that uses a set of rules in an agent-based simulation of pastoralist movements . This model assigns a herd to a sequence of seasonal grazing areas with random movements within these seasonal grazing areas [27] . We couple these movement models with disease transmission models to develop a multilevel modeling framework to quantify incidence among livestock in the Far North Region . In this work , our objective is to determine which to determinine how effective each network analysis tool quantifies the effect of movement model selection and its impact on disease simulations , considering both temporal variation and the effect of host variation . First , we generate networks from stochastic simulations of both cattle movement models and analyze these networks . Second , we simulate and quantify disease transmission across networks stochastically generated from both cattle movement models . Finally , we determine which network connectivity metrics correlate with simulated disease incidence . By evaluating the accuracy of competing movement models and quantifying the impact of movement models on predicted outbreaks in this system with consideration of temporal variability and the effect of herd characteristics , we contribute information that could help control of infectious diseases among these mobile pastoralists . More broadly , we contribute a set of model evaluation tools that can be applied across systems quantified and analyzed with multilevel models .
Location data were obtained from 72 mobile camps , which comprise a near complete set of all Cameroonian mobile pastoralists who visited the Logone Floodplain on any date between August 16 , 2007 and August 15 , 2008 [25] . Each camp contained , on average , approximately 8 households , 30 people , and 600 cattle [24] . Data were collected in two phases . First , locations of pastoral households were observed two times–in February 2008 and August 2008 –and the global positioning system ( GPS ) coordinates of campsite locations were recorded . Second , locations of pastoral households were obtained by structured interviews in August and September of 2008 . These interview data represent campsite location at a finer temporal scale and inserted information about campsite location at dates that fell between first phase data collection events , thereby documenting the entire annual transhumance . Herders provided names of all locations in which they erected a campsite during the previous year and number of days spent at each location . Members of the research team visited each of the places that pastoralists listed in the transhumance interviews to obtain geographic coordinates . The full location data set consists of the GPS coordinates of the centroid of each named location for every date between August 16 , 2007 and August 15 , 2008 . The data set is publically available at MoveBank [28] . The data collection protocol was reviewed and approved by the Ohio State University Institutional Review Board / Human Research Protection Program ( Federal-wide Assurance #00006378 from the Office for Human Research Protections in the Department of Health and Human Services: protocol 2010B0004 ) . We obtained informed consent after explaining the protocol and potential risks . Location data are input for stochastic simulations of the STM and IMM models , both of which model camp movement and resulting contacts dynamically . The STM model simulates 71 mobile camps and the IMM model simulates 67 mobile camps ( Fig 1 ) . We ran 100 simulations of the STM model and 100 simulations of the IMM model [26 , 27] . These produced 200 sets of simulated locations from which we generated networks . To generate the networks , we first calculated symmetrical daily adjacency matrices in which we defined adjacency as all camps with simulated locations within a given distance k . If two camps , i and j , were separated by k km or a shorter distance , the {i , j} entry of the daily adjacency matrix would contain a one; if two camps , i and j were separated by more than k km , the {i , j} entry of the daily adjacency matrix would contain a zero . We defined adjacency by setting k at three different distances: 0 km , 5 km , and 10km . At 0 km , camps are have simulated locations at the same coordinates , and share the same simulated location by definition . In practice , a single coordinate was chosen to represent many camps located at the same general campsite . At 5 km , we consider camps that have overlapping cattle grazing zones to be connected , because the grazing radius of a camp in the Far North Region is approximately 5 km [29] . At 10 km , we consider camps that have adjacent cattle grazing zones to be connected , again based on the 5 km grazing radius . Daily adjacency matrices were calculated for each of the 366 days for which location data and model outputs were available , because 2008 was a leap year . In total , we calculated 3 daily adjacency matrices for each of the 200 sets of simulation locations generated from two movement models , which resulted in a total of 219 , 600 daily adjacency matrices for the 2007–2008 year . To simulate disease across networks , it is important that the adjacency matrices summarize a duration equal to the infectious period of the pathogen [30] . Since the infectious period of foot-and-mouth disease is approximately one week in cattle [31] , we summarized the daily adjacency matrices as weighted weekly adjacency matrices . To calculate the weighted weekly adjacency matrices , we summed daily matrices over seven days , resulting in 31 , 200 matrices . From these weekly matrices , we generated 31 , 200 weighted weekly networks . We simulated disease transmission across weekly networks using a susceptible , infected , recovered ( SIR ) framework in which each herd was categorized in a single disease state and transmission occurred between herds . Susceptible herds had not previously experienced disease , were immunologically naïve , and capable of catching the pathogen . Infected herds harbored the pathogen and were capable of transmitting the pathogen to susceptible herds . Recovered herds mounted convalescent immunity after experiencing infection and were removed from chains of transmission [32] . Transmission propagated through the networks when an infected herd was connected to a susceptible herd by an edge . We simulated various values of R0 , or the number of secondary cases produced by a primary infectious case in a completely susceptible population . We performed transmission simulations under three values for R0 ( 1 , 5 , 10 ) , looping through the following protocol until no additional nodes could be infected . This transmission propagation procedure keeps the number of secondary infections caused by each infectious node relatively constant and not greater than the value at which R0 is set . The nodes receiving the secondary infection are randomly chosen from the set of all possible nodes connected to the infectious node . We assume that cattle herds are infectious for 7 days , which is longer than the mean duration of infectiousness of an individual animal ( 4 . 5 days [33] ) , but implies that most of the herd gets infected within days . This duration was chosen considering the size of herds in the Far North Region , Cameroon , the proximity of animals in these mobile herds , and our assumption that the entire herd is susceptible . We performed multiple disease transmission simulations for each value of R0 for a maximum of 366 days . Disease was initiated at each node and on each day , with simulations lasting from the day of initiation to the end of the 366-day period . For each of the 31 , 200 weighted weekly adjacency matrices , we performed simulations at 3 different R0 values , with disease initiated on each of the 52 weeks , for a total of 4 , 867 , 200 disease transmission simulations . We used five different connectivity metrics to compare the structure of weekly networks: three node-level metrics averaged across the entire network and two network-level metrics [34] . The node-level metrics were strength , betweenness centrality , and 3-step reach . Strength is calculated for each node by summing the weights of all connected edges . Strength is a single metric that represents both the number of connections between the given node and all other nodes plus the duration of those connections . Betweenness centrality is calculated for each node by counting the number of shortest paths between any two other nodes on which the node lies; to do this , we used the betweenness command in R . This metric represents how many other nodes a given node plays a role in connecting; in terms of disease transmission , a camp with high degree centrality is crucial for propagating disease transmission . Three-step reach is calculated by counting the number of nodes within three edge lengths or three “steps” from any given node . This metric indicates how connected the entire network is; if three-step reach is high , an epidemic will propagate though a large part of the network in three time steps . The network-level metrics were network density and network transitivity . Network density is calculated as the sum of weighted adjacency matrices divided by the total number of edges possible in the network . This metric is a different way to calculate how connected the entire network is and also has implications for the speed and extent of infectious disease transmission . Network transitivity is calculated as a ratio where the numerator is the proportion of closed triads and the denominator is the proportion of open triads [34] . This metric measures the cliquishness of a network; infections initiated in a network with high transitivity will show pockets where disease transmitted and other pockets that the disease would not reach . To measure transitivity , we used the transitivity command in R . To compare the two movement models , we calculated the value for each of these 5 network metrics for each of the 31 , 200 weekly adjacency matrices . We quantified how the 5 metrics correlated with the final epidemic sizes of disease simulations using Pearson’s correlation coefficient and p-value implemented with the cor . test function in R . The metrics that correlated well with outbreak size can be used as tools for analysis of the movement level of a multilevel model . Because network characteristics might not immediately correlate with infectious disease spread , we quantified temporal variation in the movement model network metrics and disease simulation results . We tested lag periods of 0 weeks , 1 week , 2 weeks , 3 weeks , 4 weeks , and 5 weeks . For each lag period , we used Pearson’s correlation coefficient described to quantify the correlation between connectivity metrics and outbreak size . For each connectivity metric , we identify maximum correlation coefficient and report it , along with the time period that produced the maximum correlation coefficient . This method detects correlations that might have been obscured because they did not happen simulatelously with an increase in final epidemic size .
Here , we report results from analyses conducted when adjacency was defined at 5 km . Results from analyses conducted when adjacency was defined at 0 km and 10 km are reported in the supporting information . In order to quantify network connectivity with node-level metrics , we found the strength , betweenness centrality , and 3-step reach for each node in the weekly networks for 100 simulations of the IMM and 100 simulations of the STM and averaged these values across each network ( Fig 2 ) . Average strength in the IMM and STM were similar , especially at the start and end of the year; however , average strength showed three peaks in the IMM simulations that were not observed in the STM simulations ( Fig 2A ) . Average betweenness centrality showed small peaks in the IMM and STM simulations that differed in timing ( Fig 2B ) . Average 3-step reach was greater in the STM simulations than in the IMM simulations and showed greater peaks in the STM simulations ( Fig 2C ) . In this way , the connectivity differed between the two movement models . In order to quantify connectivity with network-level metrics , we found the density and transitivity in the weekly networks for 100 simulations of the IMM and 100 simulations of the STM ( Fig 3 ) . Density in the IMM and STM were similar , especially at the start and end of the year ( Fig 3A ) and mirrored the average strength results ( Fig 2A ) ; again , density values showed peaks in the IMM simulations that were not observed in the STM simulations ( Fig 3A ) . Transitivity values were high in networks drawn from simulations of both movement models but IMM and STM values differed through time; however , the IMM values showed more peaks ( Fig 3B ) . Similar to the individual connectivity metrics , the network-wide connectivity differed between the two movement models . In order to quantify disease transmission across all networks , we ran SIR simulations ( Eqs 1 and 2 ) across all networks . With R0 = 1 , most simulations showed small final epidemic sizes , similar to those described as stuttering chains of transmission in homogeneously mixing models , and many outbreaks failed to take off ( Fig 4A and 4B ) . With R0 = 5 , simulations showed slightly larger final epidemic sizes; still , though , many outbreaks failing to take off ( Fig 4B and 4E ) . Results with R0 = 10 were similar to results with R0 = 5 , indiciating transmission saturation of the network ( Fig 4C and 4F ) . Transmission simulated over the networks generated from IMM output tended to produce smaller outbreaks ( Fig 4A–4C ) than transmission simulated over the networks generated from STM output ( Fig 4D–4F ) and the STM output showed a bimodal distribution at higher values of R0 . Therefore , the differences in the two movement models resulted in differences in disease simulation results . In order to correlate connectivity metrics with disease transmission , we computed the correlation coefficient for each of the five network analysis metrics and the final epidemic size produced by transmission simulations . For the IMM , we found that 3-step reach consistently and positively correlated with final epidemic size across a range of R0 values ( Table 1 ) . Betweenness centrality also showed consistent positive correlations , although the magnitude of the association was smaller ( Table 1 ) . For the STM , we found a different pattern in associations . In this case , strength and density consistently and positively correlated with final epidemic size across a range of R0 values ( Table 2 ) . Transitivity also showed consistent positive correlations , although the magnitude of the association was smaller ( Table 2 ) . In this way , multiple metrics appear to correlate with disease simulation when averaged across nodes and across simulations . However , this averaging diminishes the advantages of network analyses , which excel in ability to capture individual behavior . To capture temporal variation in the movement model network metrics and disease simulation results , we correlated weekly connectivity metrics with weekly outbreak results between 0 and 5 weeks lagged . We found , in general , equal or higher correlation when considering time-variation . We found that lagged strength , lagged betweenness centrality , and lagged 3-step reach correlated with final outbreak size for the IMM; lagged strength and lagged 3-step reach correlated with final outbreak size for the STM ( Table 3 ) .
Transmission models are often matched with models of host movement seeking greater accuracy in rep resenting heterogeneity and other processes that drive disease spread [4–9] . Which host movement model is chosen to couple with a disease transmission model can affect incidence results . We found that network analysis metrics–strength , 3-step reach , network density–correlated with simulated transmission when network and disease dynamics were averaged across time . When network metrics show a large degree of temporal variation , lagged metrics provide even stronger correlations . However , when considering time-variation in both the movement model and disease transmission , betweenness centrality did not provide a correlation with final epidemic size . For these reasons , we can recommend mean strength , mean 3-step reach , and network density as analysis tools that can evaluate the results of a movement model and correlate with disease transmission in both time invariant and time-varying scenarios . The practical significance of betweenness centrality measures for networks has been noted for medicine , veterinary health , and public health . Because this metric quantifies the number of shortest paths on which a node lies [35] , it can be used to identify individuals to target for vaccination that would most effectively cease outbreak transmission [36] , especially in assortative networks and at the start of an outbreak [37] . These works suggest that betweenness centrality is a metric that captures the characteristic of network structure that affects disease spread . It is surprising , then , that in our study , betweenness centrality very weakly correlated only when averaged across the IMM network or with time-lagged values from the IMM simulations; there were no positive correlations with any of the STM simulations . This suggests that nodes that have high betweenness centrality might change over the course of the simulations in the STM simulations . High betweenness centrality might not be an intrinsic quality of a node in this system , but it might be a product of movements , seasonality , and network structure . Quantifying network density leads to a different set of practical applications . Interestingly , we found that this metric was time-varying . Based on the networks drawn from empirical data , we found that network density was low in August and September at the end of the rainy season . When herds begin to move to their dry season pastures , in October 2007 , herds cluster together and display the highest network density of the entire year , because network density steadily declines until the following August . Because herds are geographically clustered , this mustering before dispersal presents a time period in which information or vaccinations could be efficiently disseminated among herders enrolled in this study . We quantify this seasonality in contacts and simulate its effects on transmission of an infectious disease in another study [38] . Uncertainty is inherent in location data . To address uncertainty in location data , we considered herds to be adjacent at three different distances . Adjacency defined at 5 km or 10 km can be thought of as analogous to a 5 km or 10 km error in location data . Interestingly , we found similar quantities of infected herds from disease transmission models over networks with adjacency defined at 5 km and 10 km . This suggests that precise locations of each camp or each animal are not needed to make accurate conclusions about network characteristics and disease transmission , similar to what has previous findings that full location data are not needed to determine optimal control [39] . Disease dynamics can be inferred with reasonable accuracy with datasets at less resolution than perfect observation of all animals . Our transmission model has been formulated to represent generic pathogens that spread either by direct contact or by aerosol through a population using a simulation procedure that keeps R0 near the value set in simulations . Simulations parametrized as R0 = 1 represent the threshold value between epidemics occurring and not occurring; with the stochasticity inherent in our transmission algorithm , this results in some stuttering chains of transmission and other instances where the epidemic fails to take off . Because of the way transmission was propogated , the variation in simulation outcomes was relatively low and the number of stochastic fadeouts might have been lower than for other more random simulation procedures . For the other values of R0 simulated , these models would be especially applicable to livestock pathogens such as foot-and-mouth disease virus . Nevertheless , livestock diseases are spread by many other modes , including environmental or vector borne transmission . Conceptually , our model with R0 = 5 that assumes direct transmission will correspond to a model with a lower R0 that assumes environmental transmission , because our networks are defined by sharing geographic locations . It remains a limitation of our model that not all modes of transmission are represented and that these mobile camps might interact with livestock reared in sedentary farms or traded at markets that are not included in the system . Nevertheless , we have found that the choice of a movement model that is coupled with a disease transmission model can influence connectivity of the host population and disease transmission results . We have found that strength , network density , and 3-step reach can be used to analyze the results of a movement model simulation in order to correlate with disease transmission simulations . We advise quantitative analysis of the movement model , evaluating sensitivity to movement and location parameters , as a requisite to multilevel modeling of disease dynamics in a mobile host population . Overall , these results have two major implications for pathogen spread among mobile camps . First , they indicate that identity of centrally connected camps can change over time . This suggest that connectivity might not be an innate property of a camp but a property that emerges though some combination of movement , seasonality , and other factors . Second , they indicate that multiple movement models should be considered and vetted before using to make a single movement model in combination with an infectious disease transmission model for decisions about the management infectious disease transmission but exact camp locations are not needed . For this reason , models of infectious disease transmission could greatly inform preventative policies and decisions about infectious disease control .
|
Epidemics of infectious disease vary geographically and vary through time . A large part of this variation is caused by movement of individuals who are susceptible to the disease or infected with the disease . To study how movement affects epidemics , researchers often combine movement models with transmission models . However , multiple movement models have been proposed , and their effect on infectious disease model output is not well understood . Here , we combine two different movement models that we developed to represent mobile pastoralists in the Far North Region , Cameroon , with the same disease transmission model . We use network metrics to test how different movement models can affect the output of the disease transmission model . We found that three metrics could be applied to movement model output in order to predict epidemic model output . We conclude that movement models coupled with disease transmission models can affect disease transmission results and should be carefully considered and vetted when modeling epidemics .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"methods",
"Results",
"Discussion"
] |
[
"infectious",
"diseases",
"computer",
"and",
"information",
"sciences",
"medicine",
"and",
"health",
"sciences",
"veterinary",
"diseases",
"livestock",
"pathology",
"and",
"laboratory",
"medicine",
"simulation",
"and",
"modeling",
"agriculture",
"infectious",
"disease",
"epidemiology",
"epidemiology",
"network",
"analysis",
"disease",
"dynamics",
"pathogens",
"biology",
"and",
"life",
"sciences",
"veterinary",
"science",
"centrality",
"research",
"and",
"analysis",
"methods"
] |
2019
|
Network analyses to quantify effects of host movement in multilevel disease transmission models using foot and mouth disease in Cameroon as a case study
|
Spores of Bacillus anthracis , the causative agent of anthrax , are known to persist in the host lungs for prolonged periods of time , however the underlying mechanism is poorly understood . In this study , we demonstrated that BclA , a major surface protein of B . anthracis spores , mediated direct binding of complement factor H ( CFH ) to spores . The surface bound CFH retained its regulatory cofactor activity resulting in C3 degradation and inhibition of downstream complement activation . By comparing results from wild type C57BL/6 mice and complement deficient mice , we further showed that BclA significantly contributed to spore persistence in the mouse lungs and dampened antibody responses to spores in a complement C3-dependent manner . In addition , prior exposure to BclA deletion spores ( ΔbclA ) provided significant protection against lethal challenges by B . anthracis , whereas the isogenic parent spores did not , indicating that BclA may also impair protective immunity . These results describe for the first time an immune inhibition mechanism of B . anthracis mediated by BclA and CFH that promotes spore persistence in vivo . The findings also suggested an important role of complement in persistent infections and thus have broad implications .
Persistent colonization of the host by microbial pathogens can cause chronic infections , which are often difficult to treat with conventional antibiotics . It is recognized that persistent infection is a unique phase often involving specific virulence factors and pathogenic mechanisms [1] . Identifying and understanding these persistent mechanisms is key to developing new strategies to more effectively combat chronic infections . Bacillus anthracis is a spore forming , Gram-positive bacterium that causes anthrax . Infections are initiated by entry of spores into the host via the respiratory system , the gastrointestinal tract , or cuts/wounds in the skin . Among the three forms of anthrax infections , inhalational anthrax has the highest mortality rate . One of the characteristic features of inhalational anthrax is the ability of spores to persist in the host lungs for prolonged periods of time [2–7] . Viable spores can be recovered from the lungs of exposed animals including non-human primates weeks or even months after the initial exposure . In addition , incubation periods of up to 43 days have been observed in humans [6] . This led to the 60-day antibiotic regimen recommended by the Centers for Disease Control and Prevention for people with pulmonary exposure to B . anthracis spores [7] . The mechanism underlying B . anthracis spore persistence is poorly understood . Mechanisms used by other bacterial pathogens for persistent infections include biofilm formation [8–12] , residing in intracellular niches [13–15] , suppression of innate and adaptive immune responses [13 , 16–18] , and changes in bacterial physiology and metabolism that favor persistent colonization [19–21] . B . anthracis spores are metabolically inactive and resistant to microbicidal effectors present in vivo . It was originally thought that the dormancy and resilience of spores were responsible for their ability to persist in the host . However , in a mouse model for spore persistence , B . anthracis spores were found to be significantly better at persisting in the lungs than Bacillus subtilis spores , suggesting the existence of persistence-promoting mechanism ( s ) beyond spore dormancy and resilience [4] . B . anthracis spores were also observed to be distributed throughout the lungs as single spores with the majority being extracellularly located [4] , suggesting that biofilm formation or hiding in an intracellular niche is unlikely to be the major underlying mechanism . It is known that pulmonary exposure to B . anthracis spores does not elicit robust inflammatory immune responses in the lungs . Although the spore surface lacks typical pathogen-associated molecular patterns such as lipopolysaccharides , lipotechoic acid , and flagellin [22] , spores have been shown to be capable of activating Toll-like receptor 2 and MyD88-dependent signaling [23] , triggering inflammatory cytokine production [24 , 25] , and activating natural killer cells [26 , 27] . Therefore the subdued immune response is likely due to an active immune evasion/suppression mechanism rather than a passive inactivity of the spores . The anthrax toxins are known to inhibit host immune responses . However , spores of a B . anthracis strain devoid of the anthrax toxins persisted as well as the parent toxin-producing strain [4] . This speaks against the possibility that low levels of anthrax toxins produced by a small amount of germinated spores in vivo may inhibit the overall immune response in the lungs and contribute to spore persistence . These observations provide support for a spore-mediated mechanism of immune suppression that has yet to be identified . Bacillus collagen-like protein of anthracis ( BclA ) is the most abundant protein on the exosporium , the outermost layer of B . anthracis spores . It is the structural component of the hair-like nap on the exosporium [28] . Because of this spatial localization , BclA sits at the forefront with respect to interactions with host factors upon entry into the host . A number of studies have shown that BclA mediates spore uptake by macrophages and epithelial cells in both complement-dependent and–independent manners [29–33] . However despite its abundance , localization and interactions with host cells , the precise role of BclA in B . anthracis pathogenesis remains unclear . In animal models of acute anthrax infections BclA did not appear to contribute to virulence [29 , 34] . In this study , the ability of BclA to manipulate the complement system and its role in spore survival and persistence in vivo was investigated . We found that BclA mediated the recruitment of complement factor H ( CFH ) , the major inhibitor of the alternative pathway , to the spore surface where it facilitated C3 degradation; thereby inhibiting downstream complement activation . We further showed that BclA significantly promoted spore persistence in the mouse lungs and dampened antibody responses to spores in a complement-dependent manner . Finally we showed that BclA impaired protective immunity against lethal B . anthracis challenges . These findings have important implications in B . anthracis pathogenesis , bacterial manipulation of complement and persistent infections in general .
Spores of B . anthracis Sterne strain 7702 and the isogenic BclA deletion mutant ( ΔbclA ) were incubated with purified human CFH . Spore-CFH interaction was analyzed using flow cytometry ( Fig 1A ) , solid phase binding assays ( Fig 1B ) and spore pull down assays ( Fig 1C ) . In all three different assays , deletion of BclA led to significantly reduced CFH binding compared to 7702 spores . Complementation of the deletion with the full-length bclA gene ( ΔbclA/BclA ) restored CFH binding ( Fig 1A–1C ) . Surface expression of BclA in the complemented strain was confirmed by immunofluorescence microscopy and flow cytometry ( S1 Fig ) . We next investigated if BclA could mediate recruitment of CFH from human and mouse serum , and mouse bronchial alveolar lavage ( BAL ) fluids . In order to distinguish between direct CFH binding and indirect binding through C3 fragments deposited on the spore surface , the binding assays were performed using heat-treated serum and BAL fluids so that the complement system was inactivated while CFH remained functional [35] . 7702 spores were able to recruit more CFH from normal human serum ( NHS ) , mouse serum and mouse BAL fluids , compared to ΔbclA and B . subtilis spores , respectively ( S2 Fig ) . To further determine if BclA was sufficient to mediate CFH binding to spores , we expressed BclA on the surface of B . subtilis spores , which do not contain any BclA-encoding genes . Surface expression was verified by immunofluorescence microscopy and flow cytometry ( S1 Fig ) . We observed that expression of BclA significantly enhanced the binding of purified CFH and CFH in human serum , mouse serum and mouse BAL fluids to B . subtilis spores ( Fig 1A–1C and S2 Fig ) . BclA was further expressed as a His-tag recombinant protein ( rBclA ) . Results from ELISAs showed that rBclA bound to CFH in a concentration-dependent and saturable manner , with an apparent KD of 0 . 91±0 . 45 μM ( Fig 1D ) . Taken together , the results described above indicated that B . anthracis spore surface protein BclA mediated direct binding of human and mouse CFH to spores . One of the principal functions of CFH is to act as a co-factor for complement factor I ( CFI ) to cleave C3b to the inactive iC3b , which disrupts the formation of the alternative complement pathway ( ACP ) C3 convertase . We first investigated the effect of BclA-mediated CFH recruitment on C3b cleavage to iC3b on the spore surface using purified complement components C3b , CFI and CFH . The results showed that the iC3b/C3b ratio on ΔbclA spores was significantly lower than that on 7702 and ΔbclA/BclA spores ( Fig 2A and 2B ) . We further incubated the different spores with NHS for various length of time . The rate of iC3b accumulation on ΔbclA spores was significantly slower compared to that on 7702 and ΔbclA/BclA spores ( Fig 2C and 2D ) . These results indicated that BclA-mediated CFH recruitment significantly promoted the cleavage of C3b to iC3b on the spore surface . The increased cleavage of C3b to iC3b in the presence of BclA could potentially reduce the available C3b necessary for efficient C3 convertase formation , thereby reducing further C3 activation . We therefore determined if BclA-mediated CFH recruitment affected C3a production in NHS incubated with the different spores . The results showed that C3a concentration was significantly higher in samples incubated with ΔbclA spores compared to those incubated with 7702 or ΔbclA/BclA spores ( Fig 2E , no antibody ) , suggesting that C3 cleavage was inhibited in the presence of BclA-expressing spores . To further determine whether the inhibition was due to CFH , we tested the effect of a CFH functional blocking antibody ( OX24 ) [36] . Pre-treatment of NHS with OX24 increased the C3a concentration in samples incubated with 7702 or ΔbclA/BclA spores to a similar level as that seen in those with ΔbclA spores; whereas pre-treatment with the isotype control antibody ( mouse IgG1 ) showed a similar pattern as that seen in the no antibody control ( Fig 2E ) . Taken together , these results suggested that BclA-mediated CFH recruitment significantly reduced further activation of C3 . Cleavage of C3b to iC3b prevents the formation of C5 convertase complexes that cleave C5 to C5a and C5b and the downstream formation of the membrane attack complex . Therefore , we next investigated the effect of BclA-mediated CFH recruitment on downstream complement activation . We first performed an indirect complement hemolytic activity assay to measure terminal stage complement activation [37] . NHS was preincubated with 7702 , ΔbclA or ΔbclA/BclA spores and centrifuged . The supernatants were used as the source of complement for hemolysis assays using opsonized sheep erythrocytes ( EA-SRBC ) as the target . If BclA led to inhibition of downstream complement activation , 7702 or ΔbclA/BclA pre-incubated serum should contain more intact complement components than ΔbclA pre-incubated serum , and thus cause more hemolysis . We observed ~ 100% hemolytic killing of EA-SRBC in sera pre-incubated with 7702 and ΔbclA/BclA spores respectively , but only 20% in serum pre-incubated with ΔbclA spores ( p < 0 . 0001 ) ( Fig 3A ) . We further measured the level of C5a in serum incubated with the different spores as a direct method to evaluate downstream complement activation . The results showed that the level of C5a was significantly higher in samples incubated with ΔbclA spores compared to those incubated with 7702 or ΔbclA/BclA spores ( Fig 3B , no antibody ) . To determine if the inhibition was due to CFH , C5a assays were performed using OX24 or control antibody pre-treated serum . The results showed that pre-treatment with OX24 increased the C5a concentration in samples incubated with 7702 and ΔbclA/BclA spores , respectively , to a similar level as that seen in those incubated with ΔbclA spores ( Fig 3B ) . In contrast , mouse IgG1 had no effect on the level of C5a in any of the samples . These results indicated that CFH was responsible for the apparent effect of BclA on C5a . We next investigated the effect of BclA on C5a level in vivo . Mice were intranasally ( i . n . ) inoculated with the different spores . BAL fluids were then collected by lavaging the lungs with sterile PBS containing EDTA , which stops complement activation . We observed that C5a concentration in the BAL fluids from mice infected with 7702 or ΔbclA/BclA spores was significantly lower than that from mice infected with ΔbclA spores ( Fig 3C ) . Taken together , the results described above indicated that BclA-CFH interaction led to reduced C5 cleavage both in vitro and in vivo . We first investigated if BclA was important for spore persistence . C57BL/6 mice were i . n . inoculated with sub-lethal doses of spores . Total bacteria and spore load in the lungs at two and four weeks post inoculation was determined . At both time points , C57BL/6 mice inoculated with 7702 spores harbored significantly more total bacteria and spores in the lungs than those inoculated with ΔbclA spores ( Fig 4A and S5A Fig ) . Complementation of BclA in the ΔbclA background significantly increased the spore counts in the lungs at both time points . We tested the germination efficiency of 7702 , ΔbclA and ΔbclA/BclA spores in three different media: a chemically defined germination media , LB and 100% NHS ( S3 Fig ) . We did not observe any difference in the germination efficiency between the spores in any of the media . We also tested bacterial dissemination to distal organs such as the spleen and found no significant difference in bacterial burden in the spleen of mice inoculated with the different spores ( S4A Fig ) . Hematoxylin and Eosin ( H&E ) staining was performed on lung sections from C57BL/6 mice collected at two weeks post i . n . inoculation with either 7702 or ΔbclA spores . Minimum pathology was observed in sections from both groups ( S7 Fig ) , consistent with a previous report [4] . The alveolar and small airway epithelium appeared intact in both groups and lymphocyte infiltration was only occasionally observed . Overall , we did not see obvious differences in inflammatory responses in the lungs between the two groups . To determine if BclA alone was sufficient to promote spore persistence in the lungs , we further examined B . subtilis spores expressing BclA . The results showed that expression of BclA on the surface of B . subtilis spores significantly increased both total bacteria and spore burden in the lungs compared to the vector control ( Fig 4B ) . Together , these results suggested that BclA significantly promoted spore persistence in vivo . We next investigated the role of complement in spore persistence . CFH deficiency in mice caused uncontrolled complement activation resulting in C3 consumption [38] . Therefore , we compared spore persistence in C3-/- mice , as C3 is where the three complement pathways converge . We reasoned that if BclA-mediated inhibition of complement was responsible for the increased spore persistence in the lungs , we should see no difference in spore persistence between 7702 and ΔbclA in C3-/- mice . Indeed , no significant difference in either total bacteria or spore burden was observed in C3-/- mice between 7702 and ΔbclA-infected groups at 2 or 4 weeks post inoculation ( Fig 4C and S5B Fig ) , suggesting that BclA-mediated promotion of spore persistence was C3-dependent . It was previously reported that BclA bound complement component C1q . The binding leads to internalization of spores by epithelial cells through integrin α2β1 and opsonophagocytosis of spores by macrophages [30 , 32] . We examined spore persistence in C1q-deficient ( C1q-/- ) mice . The results from C1q-/- mice mirrored those from wild type C57BL/6 mice ( S6A Fig ) , suggesting that BclA-C1q interaction was not important for spore persistence in the mouse lungs . Taken together , the results suggested that spore persistence was promoted by BclA-mediated inhibition of complement activation . To test if BclA contributes to virulence in acute infections , C57BL/6 and C3-/- mice were injected with lethal doses of 7702 or ΔbclA spores by intraperitoneal injection ( i . p . ) . No significant difference in mouse survival was observed between 7702 and ΔbclA-infected groups in either mouse strains ( Fig 4D and 4E ) . We next compared the total bacteria and spore burden in the lungs and spleen of C57BL/6 and C3-/- mice 48 hours post inoculation . We did not observe any significant difference in total bacteria or spore load in the lungs or the spleen between mice challenged with 7702 and those with ΔbclA spores ( Fig 4F and 4G ) . These results indicate that BclA does not contribute to virulence in this lethal challenge model . This is consistent with results from previous studies using lethal infection models [29 , 34] The complement system not only shapes the innate immune responses , but also guides the adaptive immune responses [39–43] . We examined the effect of BclA on host antibody responses against spores in the persistence model . Anti-spore IgG antibodies in serum from infected mice were detected using ELISA . Both 7702 and ΔbclA spores elicited specific antibody responses in C57BL/6 ( Fig 5A ) , C3-/- ( Fig 5B ) and C1q-/- ( S6B Fig ) mice , respectively , compared to the saline control . However , C57BL/6 and C1q-/- mice exposed to 7702 spores had significantly lower antibody titers compared to those exposed to ΔbclA spores ( Fig 5A and S6B Fig ) , suggesting that BclA dampened antibody responses and that BclA-C1q interaction was not important in this process . The difference in anti-spore IgG titers between the 7702- and ΔbclA-infected groups was not detectable in C3-/- mice ( Fig 5B ) , suggesting that BclA dampened antibody responses against spores through downregulating C3 activation . Because of the significant difference in anti-spore antibody levels , we investigated if prior exposure to 7702 or ΔbclA spores triggered different protection against lethal B . anthracis challenges . In one set of experiment , C57BL/6 mice were i . n . inoculated with a sub-lethal dose of 7702 or ΔbclA spores and then challenged with a lethal dose of 7702 spores by intraperitoneal ( i . p . ) injection two weeks later . In another set , C57BL/6 mice were i . n . inoculated with sub-lethal doses of 7702 or ΔbclA spores at 0 and 2 weeks and then challenged with a lethal dose of 7702 spores by i . p . injection at four weeks . The results showed that for mice with one prior exposure ( Fig 6A ) , those pre-exposed to 7702 spores succumbed to lethal challenges within two days , similar to those pre-exposed to saline only , whereas those pre-exposed to ΔbclA spores had a significantly better survival rate ( p = 0 . 0308 vs . the saline control ) with a median survival time of 4 days . For mice with two prior exposures ( Fig 6B ) , the difference was even more pronounced ( p = 0 . 0002 vs . the control group , p = 0 . 0069 vs . 7702 pre-exposed group ) . Taken together , these results suggested that BclA impaired protective immunity against lethal B . anthracis infections .
In this study we discovered a novel function for the major B . anthracis spore surface protein BclA . We demonstrated that BclA mediated recruitment of CFH to spores , facilitated C3b degradation on the spore surface , inhibited further C3 activation , and reduced C5 cleavage both in vitro and in vivo . We further showed that BclA promoted spore persistence in the host lungs and inhibited antibody responses against spores in a C3-dependent manner . Furthermore , BclA impaired protective immunity against lethal B . anthracis challenges . These results describe for the first time a spore-mediated immune modulatory mechanism through inhibition of complement . The results also suggested an important role of complement in persistent infections , an aspect of pathogen-complement interaction that is poorly understood . The ability of BclA to mediate CFH binding was demonstrated by 1 ) ΔbclA spores bound significantly less CFH than the parent spores , and the defect was restored by complementing BclA , 2 ) BclA expressed on the surface of B . subtilis spores was sufficient to promote CFH binding , and 3 ) recombinant BclA protein bound to purified CFH in a concentration-dependent manner . We observed weaker CFH binding by ΔbclA and B . subtilis control spores . It is possible that there is another unknown low-affinity CFH binding protein on these spores or non-specific binding of CFH to spores . Our results also suggest that recognition of CFH by BclA is not human specific , i . e . , BclA can bind both human and murine CFH , unlike some other CFH binding proteins such as the CFH-binding protein ( fHbp ) of Neisseria meningitidis [44] and PspC of Streptococcus pneumoniae [45] . A group A streptococcal collagen-like protein ( Scl1 ) was reported to bind CFH via the C-terminal variable region of Scl1 [46] . While BclA is also a collagen-like protein , sequence comparison indicated no significant sequence similarities between the two proteins beyond the GXY triplet-repeating motif . BclA also did not show any significant sequence similarities to other reported microbial CFH binding proteins . Thus BclA is a novel CFH binding protein . BclA-bound CFH retained its co-factor activity , as shown by increased C3b degradation on the surface of parent and complemented spores compared with ΔbclA spores . BclA-CFH interaction inhibited further C3 activation , and decreased C5 activation as shown by C5a ELISA and hemolytic assays . The finding that C5 cleavage was also reduced in the mouse lungs in the presence of BclA further suggested that this effect was relevant in vivo . In addition , CFH functional blocking antibodies completely abolished the complement inhibitory activity of BclA , suggesting that the BclA-CFH interaction was responsible for this activity . It has been known for decades that B . anthracis spores were able to persist in the host lungs for prolonged periods of time . This capability was thought to be due to the dormancy and resilience of spores . The results from this study describe for the first time a specific persistence-promoting mechanism mediated by the spore surface protein BclA . The observation that the difference in spore load in the lungs between 7702 and ΔbclA-infected mice disappeared in C3-/- mice suggests BclA promotes spore persistence in the lungs by inhibiting complement activities . It was previously reported that BclA directly binds C1q and this interaction leads to activation of the classical complement pathway and opsonophagocytosis of spores by macrophages [30 , 32] . However , the results obtained from C1q-/- mice suggest that BclA-C1q interaction is not important for spore persistence or antibody response to spores . This suggests that in this model system , inhibition of the alternative pathway plays a dominant role in promoting spore persistence . Recently it was shown that binding of CFH to the PspC protein of S . pneumoniae promoted pneumococcal nasal colonization by CFH-mediated bacterial adherence to the epithelium [47] . In our case , CFH is present in C3-/- mice , suggesting that BclA-mediated promotion of spore persistence ultimately depends on C3 and works by inhibiting complement activities . Thus the persistence colonization mechanism described here is distinct from that of S . pneumoniae . The percentage of bacteria recovered from the lungs at 2 weeks post inoculation versus the initial inoculum was ~ 0 . 05% . This is in the same range as reported previously in Balb/c mice ( ~ 0 . 08% ) [4] . The role of BclA in pathogenesis has been controversial despite the fact that it is a dominant protein on the spore surface . Studies in lethal infection models did not show any contribution of BclA to virulence [29 , 34] . In the lethal spore challenge model here , our results also show no difference in either mouse survival or bacterial burden between mice challenged with 7702 and ΔbclA spores , consistent with previous studies . The findings here suggest that the primary role of BclA in vivo may be to promote the long term survival of spores through inhibition of complement activities . We observed that 7702 spores led to significantly lower anti-spore antibody levels compared to ΔbclA spores in C57BL/6 and C1q-/- mice . The reduced antibody response was not due to a lower bacterial burden of 7702 in vivo; on the contrary 7702 infected mice had a higher spore burden in the lungs and a similar burden in the spleen compared to ΔbclA-infected mice . The fact that there was no difference in antibody responses in C3-/- mice suggested that BclA-mediated inhibition of C3 and/or downstream complement activation was responsible for the reduced antibody response to spores . The complement system influences B cells , T cells and antigen-presenting cells , the major cell types in the adaptive immune system [48–57] . The interaction between BclA and CFH can potentially affect all these components of the adaptive immune system . It has also been reported that CFH binding led to impaired antibody responses against the corresponding CFH-binding protein [47 , 58–61] . Antibodies tend to recognize epitopes outside the CFH binding sites hence do not block CFH binding , or even enhance CFH binding . It would be interesting to investigate how BclA-CFH interaction affects antibody responses against B . anthracis spores . Finally we observed that pre-exposure to 7702 spores conferred virtually no protection against lethal challenges whereas pre-exposure to ΔbclA spores provided significant protection in our infection model . The finding that BclA not only inhibited antibody responses against spores but also impaired protective immunity against B . anthracis lethal challenges has important implications in anthrax vaccine development and in persistent infections in general . With respect to vaccine development , BclA has been pursued as a vaccine candidate together with protective antigen ( PA ) as a multicomponent anthrax vaccine . Vaccination with BclA either as a recombinant protein or as a DNA vaccine augmented the protective efficacy of PA [62–64] . However , vaccination with formalin killed spores showed that ΔbclA spores provided greater protection than BclA-producing spores [65] . Our findings here suggest that the latter observation may be due to the effect of BclA on complement . The fHbp of N . meningitidis was approved as a component in multicomponent vaccines against serogroup B meningococcus [66–68] . Recent studies found that fHbp mutant proteins defective in CFH binding were more immunogenic and elicited stronger protective antibody responses than wild type proteins [59 , 60 , 69] . This raised the possibility that perhaps BclA mutants defective in CFH binding may offer better protection against anthrax infections . Previous studies on the effect of pathogen manipulation of complement have been primarily focused on the more immediate effects of complement such as complement-mediated killing and opsonophagocytosis in the context of acute infections . For those bacteria that are susceptible to complement-mediated killing such as Gram-negative pathogens ( e . g . , N . meningitidis ) or spirochetes ( e . g . , Borrelia burgforderi ) , inhibition of complement activation by recruiting CFH or other mechanisms confers serum resistance to the bacteria and is important for bacterial survival and virulence in vivo [66 , 70–72] . For Gram-positive bacteria which are relatively resistant to serum killing due to their thick peptidoglycan cell wall , inhibition of complement activation can hinder phagocytosis and protects bacteria from phagocytic clearance . For example , Streptococcus pyogenes was found to inhibit phagocytosis by inactivating C3b in a strain-dependent mechanism [73] . Binding to CFH or C4-binding protein by S . pyogenes led to increased mortality in mouse models [74] . In contrast , the long-term effect of complement inhibition by pathogens has not been well studied in a systematic manner . The results presented here suggest that inhibition of complement by pathogens can play an important role in promoting persistent infections . In addition , because spores are resistant to lysis by complement and to phagocytic killing [30 , 75] , the role of complement observed in this study was likely due to the indirect activities of the complement effectors on the innate and adaptive immune system . This finding is particularly relevant to Gram-positive , encapsulated , or spore-forming pathogens , which tend to be relatively resistant to complement-mediated or phagocytic killing . With respect to how inhibition of complement promotes spore persistence , there may be multiple mechanisms involving both the innate and adaptive immune systems . Decreased production of C3a and C5a can affect cytokine production and the activation status of phagocytes . CFH binding to spores may not only dampen antibody responses but also affect the specific antibodies produced , as found in N . meningitidis . In addition , T cell and/or B cell functions can be affected [48–57] . Further studies to elucidate the detailed mechanism underlying the role of complement in persistent infections will be important . In conclusion , we characterized the first CFH-binding protein of B . anthracis and described for the first time a spore-mediated immune inhibition mechanism of B . anthracis . These results shed light on the role of BclA in vivo . In addition , our findings suggest that in addition to conferring resistance to complement-mediated killing and opsonophagocytosis , complement inhibition by pathogens have long-term consequences with respect to persistent infections and protective immunity . Considering a growing list of microbial pathogens capable of modulating complement activities [76–80] , our findings have broad implications .
Strains and plasmids used in this study are listed in S1 Table . Spores of B . anthracis and B . subtilis were prepared by culturing in a PA broth or on LB agar plates as described [4 , 30] . To inhibit spore germination , a germination inhibitor D-alanine ( 2 . 5 mM ) was included in solutions for assays involving spores . Normal human serum ( NHS ) , GVB0 buffer ( Gelatin Veronal Buffer without Ca2+ , Mg2+ , 0 . 1% gelatin , 5 mM Veronal , 145 mM NaCl , 0 . 025% NaN3 , pH 7 . 3 ) , VBS++ buffer ( 5 mM Veronal , 145 mM NaCl , 0 . 025% NaN3 , pH 7 . 3 , 0 . 15 mM CaCl2 and 0 . 5 mM MgCl2 ) , purified complement proteins and goat anti-human CFH , anti-human C1q and anti-human C3 antibodies , were from Complement Technology unless otherwise stated . Secondary antibodies were from Thermo Fisher Scientific unless otherwise stated . 2 , 2 , 2-Tribromoethanol ( Avertin ) , bovine serum albumin ( BSA ) , chicken ovalbumin ( OVA ) , D-alanine , and L-alanine were purchased from Sigma . Heat inactivation of complement was carried out at 56°C for 30 min . A DNA fragment containing the bclA gene and its upstream sequence ( ~ 1kb ) was cloned into an E . coli—B . anthracis shuttle vector pUTE583 [81] . The construct was then introduced into ΔbclA by electroporation as described previously [82] . To express BclA on the surface of B . subtilis spores , a DNA fragment encoding amino acid residues 39–400 of BclA was fused to the C-terminus of CgeA , a protein on the outermost surface of B . subtilis spores [83] . The first 38 amino acid residues were omitted because this region was reported to be proteolytically cleaved before anchorage of BclA onto the B . anthracis spore surface [84] . CgeA-BclA fusion was cloned into pDG1662 , which allows the ectopic integration at the non-essential amyE locus in the B . subtilis chromosome [85 , 86] . Surface expression was evaluated by staining spores with anti-BclA antibodies and fluorescently labeled secondary antibodies followed by immunofluorescence microscopy or flow cytometry analysis , as described in S1 Text . To detect CFH recruitment to the spore surface , ~ 5×107 spores were incubated at 37°C for 30 min in PBS containing 2 . 5 mM D-alanine and supplemented with one of the following: purified human CFH ( 10 μg/ml ) , bovine serum albumin ( 10 μg/ml ) , 10% ( v/v ) heat-inactivated NHS , 10% ( v/v ) heat-inactivated mouse serum ( from C57BL/6 ) , or heat-inactivated mouse BAL fluid ( from C57BL/6 ) . The spores were then washed three times with ice-cold PBS containing 2 . 5 mM D-alanine and resuspended in the same buffer . An aliquot of the spore suspension was used to titer the spores by dilution plating and the rest were frozen until ready for analysis . Equal amounts of spores were boiled in SDS-sample loading buffer and the supernatants were subjected to Western blot analysis using goat anti-human CFH ( 1:10000 ) or sheep anti-CFH antibody ( 1:2000 , Abcam ) followed by incubation with rabbit anti-goat antibody conjugated to horseradish peroxidase ( HRP ) ( 1:10000 , Invitrogen ) or HRP-conjugated rabbit anti-sheep IgG ( 1:10000 , Invitrogen ) for 1 hr . To detect iC3b deposition , ~ 5×107 spores were incubated in PBS buffer containing 500 μg/ml C3b , 100 μg/ml CFH , 4 μg/ml CFI , 0 . 1% BSA , 1 mM MgCl2 and 2 . 5 mM D-alanine at 37°C for 10 min . Spores were washed three times with ice-cold PBS containing 2 . 5 mM D-alanine . Equal amounts of spores were subject to Western blot analysis following the procedure described above . C3 fragments were detected using goat-anti human C3 ( 1:10000 ) and rabbit anti-goat HRP ( 1:10000 ) . Band intensities were quantified using Image J . For CFH binding , ~ 5×107 spores were incubated in buffer containing 2 . 5 mM D-alanine and supplemented with either 25 μg/ml purified human CFH or 10% heat inactivated NHS at 37°C for indicated length of time . Spores were then washed and fixed with 2% paraformaldehyde for 20 min at room temperature . Bound CFH was detected using goat anti-human CFH ( 1:400 , Santa Cruz ) followed by donkey anti-goat PE ( 1:400 , Santa Cruz ) . For iC3b deposition , spores were incubated in buffer containing 10% NHS and 2 . 5 mM D-alanine for indicated length of time . iC3b was detected using mouse monoclonal antibody to human iC3b ( neoantigen ) ( 1:400; Quidel ) and donkey anti-mouse 647 ( 1:400; SantaCruz ) . Samples were analyzed in a two laser Accuri C6 analytical flow cytometer using forward and side scatter parameters to gate on at least 20 , 000 spores . The red laser was used to measure the mean fluorescence intensity ( MFI ) of PE-labeled samples and data were analyzed using CFlow Plus ( Accuri Cytometers ) and graphed using GraphPad Prism 6 analysis software . Recombinant BclA ( rBclA ) was purified as described previously [32] . Microtiter 2HB plates were coated with 10 ug/ml purified CFH or ovalbumin ( OVA ) in HBS ( 20 mM HEPES and 50 mM NaCl , pH 7 . 4 ) overnight at 4°C . The wells were washed to remove unbound proteins by HBS with 0 . 05% Tween 20 ( HBST ) , blocked in HBST with 1% OVA for 1 hr at room temperature , and incubated with increased concentrations ( 0 . 01 , 0 . 1 , 1 , 5 , 15 and 30 μM ) of His-tagged rBclA in HBST for 2 hrs at room temperature . The wells were then washed three times with HBST and incubated with anti-His HRP ( 1:3000 , Alpha Diagnostic Intl . Inc . ) for 1 hr . Plates were developed with Sigmafast OPD and read at 450 nm . Apparent KD was determined by non-linear regression ( GraphPad Prism 6 ) . For spore binding , purified human CFH was immobilized onto wells of 96-well plates , blocked and incubated with ~ 1×107 biotin-labeled spores suspended in the blocking buffer supplemented with 2 . 5mM D-alanine for 30 min at 37°C followed by washing and incubation with streptavidin-conjugated to HRP . Approximately 5×107–1×108 spores were incubated in GVB0 buffer containing 20% NHS and 2 . 5 mM D-alanine at 37°C for 30–60 min . Complement activation was terminated by adding 50 mM EDTA . The samples were centrifuged to remove spores . C3a and C5a levels in the supernatants were determined using Human C3a ELISA kit ( BD OptEIA™ ) and Human Complement Component C5a DuoSet ( R&D ) , respectively . To determine the effect of CFH functional blocking antibody OX24 , 20% NHS in GVB0 buffer was pre-incubated with OX24 ( Pierce Antibody ) or isotype control mouse IgG1 ( Sigma ) at 240 nM or 480 nM final concentration at 37°C for 30 min . The reaction mix was then incubated with spores as described above . Spores were incubated in buffer containing 20% NHS and 2 . 5 mM D-alanine at 37°C for 60 min . After centrifugation , the supernatants were diluted ( 1:10 ) in VBS++ ( Ca2+ , Mg2+ ) and used as the source of complement in hemolytic assays with opsonized sheep erythrocytes ( 1×107 cells ) following the instructions of the supplier ( EA-SRBC , CompTech ) . %Lysis is calculated as OD540 ( test ) −OD540 ( Blank ) OD540 ( total lysis ) −OD540 ( Blank ) ×100 . C57BL/6 mice were i . n . inoculated with different spores ( ~1×108 spores/mouse ) and BAL fluid was collected 6 hours later by lavaging the lungs with 1ml cold sterile PBS containing 50 mM EDTA . The lavage fluids were centrifuged to remove cells and bacteria . C5a level in the supernatants was measured using the Mouse Complement Component C5a DuoSet ( R&D ) . All animal procedures were performed according to protocols approved by the Institutional Animal Care and Use Committee , Texas A&M Health Science Center ( TAMHSC ) . C57BL/6 ( originally purchased from the Jackson Laboratories ) , C1q-/- [87] and C3-/- [88] mice were maintained at the animal facility at TAMHSC . Mice were euthanized by i . p . injection with an overdose of 2 , 2 , 2-Tribromoethanol ( Avertin ) followed by terminal bleed . Intranasal inoculation was performed as previously described [4] . Briefly , 6–12 week old mice were anesthetized with Avertin ( 0 . 3 mg/g body weight ) and then inoculated with 20 μls of an indicated sub-lethal dose of spores . For lethal challenge experiments , mice were inoculated with a lethal dose of spores by i . p . injection . Mice were monitored for survival and other symptoms daily . Both male and female mice were used in the experiments in a sex-matched manner . Lungs and spleens were homogenized in 1 ml sterile ice-cold PBS containing 2 . 5 mM D-alanine , and either directly dilution plated to determine the total bacterial counts or heated at 68°C for 60 min and dilution plated to determine the spore counts . Lungs were also fixed for histological evaluation as described in S1 Text . Total exosporium proteins were extracted from 7702 spores as previously described [89] with slight modifications . Briefly , ~5×109 spores were resuspended in 200 μl of an extraction buffer ( 50 mM Tris-HCl , pH 7 . 4 , 8 M urea , and 2% ( v/v ) 2-mercaptoethanol ) , heated for 20 min at 90°C , and centrifuged at 13 , 000× g for 10 min . The supernatant was then treated with 20% ( v/v ) ice-cold trichloroacetic acid for 30 min on ice and centrifuged at 13 , 000×g at 4°C for 25 min . The pellet was washed once with 1 ml ice-cold acetone , centrifuged at 7000 r . p . m for 2 min , and dissolved in 100 μl of a solution of 200 mM Tris-HCl ( pH 7 . 4 ) and 0 . 1 M glycine . Blood was collected either from the saphenous vein , or by terminal bleed from the posterior vena cava , at two weeks after mice were inoculated with spores . Blood was allowed to clot at room temperature for 45 min before centrifugation at 4000 r . p . m . at 4°C for 10 min . Serum was either stored immediately at -80°C or at 4°C with 0 . 1% sodium azide . Extracted spore antigens were immobilized onto 96-well plates at 0 . 5 μg/well . The plates were washed twice with PBS containing 0 . 1% Tween-20 ( PBST ) , and blocked with PBST containing 3% BSA at 37°C for 1 hr . Serum samples were diluted ( 1:100 for serum from C57BL/6 and C1q-/- , and 1:2 for serum from C3-/- mice ) with PBS containing 3% BSA and incubated at 37°C for 1 hr . The wells were washed three times with PBST . Bound IgG was detected using goat anti-mouse IgG conjugated with HRP ( 1:2500 , Invitrogen ) . Spore germination was evaluated as described in S1 Text . Pairwise comparison was carried out using Student’s t test . Survival analysis was performed using the Log-rank test ( GraphPad Prism 6 ) . All animal experiments were performed in accordance to procedures approved by the Institutional Animal Care and Use Committee at Texas A&M Health Science Center ( IACUC# 2015-0361-IBT ) . The Texas A&M University Health Science Center—Institute of Biosciences and Technology is registered with the Office of Laboratory Animal Welfare per Assurance A4012-01 . It is guided by the PHS Policy on Human Care and Use of Laboratory Animals ( Policy ) , as well as all applicable provisions of the Animal Welfare Act . Mice were euthanized by intraperitoneal injection of overdosed Tribromethanol/Avertin followed by terminal bleed . Mice were anesthetized with Avertin before intranasal inoculation of spores . All efforts were made to minimize animal suffering .
|
We discovered an immune modulatory mechanism of Bacillus anthracis mediated by the spore surface protein BclA . We showed for the first time that BclA mediated the binding of complement factor H , a major negative regulator of complement , to the surface of spores . The binding led to the down-regulation of complement activities in vitro and in an animal model . Using mice deficient in complement components , we further showed that BclA promoted spore persistence in the mouse lungs and impaired antibody responses against spores in a complement-dependent manner . We further provided evidence suggesting a role of BclA in the development of protective immunity against lethal B . anthracis challenges . These findings draw attention to a previously understudied aspect of the complement system . They suggest that in addition to conferring resistance to complement-mediated killing and phagocytosis , complement inhibition by pathogens have long-term consequences with respect to persistent infections and development of protective immunity . Considering a growing list of microbial pathogens capable of modulating complement activities , our findings have broad implications .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods",
"and",
"Materials"
] |
[
"bacteriology",
"complement",
"system",
"medicine",
"and",
"health",
"sciences",
"immune",
"physiology",
"enzyme-linked",
"immunoassays",
"pathology",
"and",
"laboratory",
"medicine",
"pathogens",
"immunology",
"bacillus",
"microbiology",
"antibodies",
"immunologic",
"techniques",
"bacteria",
"bacterial",
"pathogens",
"antibody",
"response",
"research",
"and",
"analysis",
"methods",
"immune",
"system",
"proteins",
"complement",
"inhibitors",
"microbial",
"physiology",
"proteins",
"medical",
"microbiology",
"microbial",
"pathogens",
"immunoassays",
"bacterial",
"spores",
"bacillus",
"anthracis",
"immune",
"response",
"immune",
"system",
"complement",
"activation",
"biochemistry",
"bacterial",
"physiology",
"physiology",
"biology",
"and",
"life",
"sciences",
"organisms"
] |
2016
|
Bacillus anthracis Spore Surface Protein BclA Mediates Complement Factor H Binding to Spores and Promotes Spore Persistence
|
In mammals , the developmental path that links the primary behaviours observed during foetal stages to the full fledged behaviours observed in adults is still beyond our understanding . Often theories of motor control try to deal with the process of incremental learning in an abstract and modular way without establishing any correspondence with the mammalian developmental stages . In this paper , we propose a computational model that links three distinct behaviours which appear at three different stages of development . In order of appearance , these behaviours are: spontaneous motor activity ( SMA ) , reflexes , and coordinated behaviours , such as locomotion . The goal of our model is to address in silico four hypotheses that are currently hard to verify in vivo: First , the hypothesis that spinal reflex circuits can be self-organized from the sensor and motor activity induced by SMA . Second , the hypothesis that supraspinal systems can modulate reflex circuits to achieve coordinated behaviour . Third , the hypothesis that , since SMA is observed in an organism throughout its entire lifetime , it provides a mechanism suitable to maintain the reflex circuits aligned with the musculoskeletal system , and thus adapt to changes in body morphology . And fourth , the hypothesis that by changing the modulation of the reflex circuits over time , one can switch between different coordinated behaviours . Our model is tested in a simulated musculoskeletal leg actuated by six muscles arranged in a number of different ways . Hopping is used as a case study of coordinated behaviour . Our results show that reflex circuits can be self-organized from SMA , and that , once these circuits are in place , they can be modulated to achieve coordinated behaviour . In addition , our results show that our model can naturally adapt to different morphological changes and perform behavioural transitions .
In mammals , the developmental path that links the rudimentary behaviours observed during foetal stages to the full fledged behaviours observed in adults is still beyond our understanding [1] . We observe foetus generate spontaneous motor activity [2] , [3] , we observe newborns react reflexively to external stimulation , and later in life we observe adults skilfully strolling around . The period of time that goes from the first stage to the last can be longer ( e . g . in altricial species ) or shorter ( e . g . precocial species ) , but all mammals undergo this general developmental path . It is noteworthy that in this paper we restrict the scope of the term mammal to refer only to terrestrial mammals . It is commonly agreed that mammalian development is intrinsically incremental . Intuitively , this notion fits well with natural observations; simpler behaviours , like reflexes , tend to appear first , and only after these are in place , can one observe more elaborated and purposeful ones , like locomotion . Whether the presence of the former is required for the execution of the latter has been historically disputed [4]–[6] , but nowadays the contribution of reflexes to different coordinated behaviours is widely accepted [7]–[10] . When one looks at the circuitry of the most basic reflexes , one can appreciate their close relation to the underlying morphology of the musculoskeletal system . For example , the stretch reflex [11] p . 439–40 , which deals with muscle-length information , entails excitatory connections with synergist α-motoneurons and inhibitory connections with antagonist α-motoneurons ( see below ) . This symmetrical relation in the reflex circuitry mirrors a mechanical ( and geometrical ) relation in the musculoskeletal system: When one muscle stretches , its synergists are elongated while its antagonists are shortened . This is valid for muscle interactions at the legs as well as at the arms and torso , which could justify the invariant pattern of connectivity observed in different parts of the spinal cord . The symbiosis between reflex circuits and body morphology can as well be argued for other reflexes , like the spinal withdrawal reflex [12] , or the non-spinal vestibular , auditory , and pupillary reflexes . The proximity between reflexes and body morphology make the former ideal mechanisms to coordinate muscles at a local level . Sherrington , one of the first to recognize this relation , hypothesised that reflexes were more than stereotyped reactive responses; they were best seen as modular mechanisms that can be combined to achieve general motor coordination [5] , [13] . This hypothesis was further developed into what became known as the threshold control theory or TCT , first known as the equilibrium point ( EP ) hypothesis [14] , [15] . According to this theory , behaviour is the outcome of shifts in the equilibrium state between the organism and its environment ( see also Thelen's dynamic systems theory [16] ) . Equilibrium shifts can be caused voluntarily by the organism , or involuntarily by the environment [15] . At the voluntary level , the nervous system can modify the current EP by shifting the reference length value of different muscles; this induces activity in the muscle spindles and , through tunable reflex circuits , produces forces that bring the organism to a new EP . In this context , a particularly recent groundbreaking work has been that of Geyer and Herr [17] . They have shown that the motor coordination necessary to achieve stable walking in a musculoskeletal system can be brought about by tuning appropriately a number of reflexive feedback loops . However , to achieve this they have to manually establish and tune a large number of reflex circuits , which makes search for an appropriate set of parameters very difficult . A similar approach has been applied to a real-world robot , the RunBot , which can display smooth walking patterns through the coordination of reflex networks [18] . An alternative approach to the modularity of the motor system has been offered by the framework of muscle synergies [19]–[23] . This framework tries to resolve the problem of controlling a large number of degrees of freedom [24] by combining a small number of synergies ( or modules ) , which already incorporate basic muscle activation profiles . In general , “a module is a functional unit in the spinal cord that generates a specific motor output by imposing a specific pattern of muscle activations” [25] . Modules are combined by supraspinal systems to produce the muscle activations necessary to achieve a desired task . Similarly to reflexes , synergies also seem to strongly reflect the bodies mechanical constraints [21] . At the moment , this form of modularity is formulated in a rather abstract and mathematical sense , which neglects for the most part the neural circuits as well as the developmental processes necessary to implement them ( see for example [20] , [22] ) . From a theoretical perspective , we adopt the view that development is a key aspect to understand how the nervous system achieves coordinated behaviour [26]–[28] . Following this view , this paper proposes a computational model that links three behaviours which appear at different stages of development; in order of appearance these behaviours are: spontaneous motor activity ( SMA ) , reflexes , and coordinated behaviour ( see Figure 1 ) . The model proposed identifies the mechanisms according to which ( 1 ) SMA propels the self-organization of adaptive reflex circuits , and ( 2 ) reflex circuits are manipulated to achieve coordinated behaviour . The main motivation to build our model is to validate in silico four hypotheses that are currently very difficult to verify in vivo . First , we hypothesise that SMA induces sensory and motor responses which are sufficient to self-organize reflex circuits . This has been shown in vivo in the case of the spinal withdrawal reflex [29] , but has not yet been established for other reflexes . Second , we hypothesise that , once meaningful reflex circuitry is in place , it can be modulated to achieve coordinated behaviour . Third , we hypothesise that , since SMA is observed after birth , throughout the entire lifetime of an individual , it provides a mechanism to continuously adapt the reflex circuits to potential morphological changes ( e . g . due to injury or growth ) . And fourth , we hypothesise that we can achieve behavioural transitions ( i . e . switch between different behaviours ) by changing the modulation of the reflex gains over time . All the experiments have been carried out in a simulated musculoskeletal leg model . As a case study for coordinated behaviour we use vertical hopping . Hopping is a particularly convenient behaviour in the context of this paper . First , it requires the activity of several muscles to be coordinated over time . Second , it requires only a single leg , which bypasses the need to deal with the problem of inter-limb coordination at this stage ( see Discussion ) . Third , the behaviour is not limited to point-to-point movements , but it requires highly dynamical interactions between the leg and the environment ( in particular the ground ) . And forth , it includes motion patterns that can be periodic as well as aperiodic ( e . g . starting hopping from the ground ) . At the end of the paper , we also include an experiment in which we show how our model scales to achieve point-to-point behaviours . The reminder of this paper is organized as follows . The second section provides the different components of our developmental model . The third section describes our experiments and results . The fourth section discusses the results obtained .
The musculoskeletal system consists of a leg model comprising three rigid segments: pelvis , femur and tibia ( see Figure 3a ) . The model is implemented in MATLAB SimMechanics and visualized using the 3D Animation Toolbox ( also from MATLAB ) . The system is actuated primarily by six muscles , but in one of the experiments we use a four-muscle configuration ( see Results ) . The masses of the rigid segments are set to the lengths of the femur , and tibia are set to which is their approximate length in a human with [31] p . 302 . The hip and knee joints are simulated as revolute joints . An additional joint is added to the hip in order to restrict the movement of the pelvis to a vertical motion . This joint also prevents the rotation of the pelvis . We call this prismatic constraint , the hopping axis ( see Figure3a ) . It is worth mentioning that our model is intrinsically a 3D model , in the sense that every point in it ( e . g . the joint locations as well as the attachment points of the muscles ) is defined by three coordinates . However , in practice , given that the hip and knee joints are both hinge joints aligned along a single plane , the motions of all the rigid bodies are restricted to 2D movements . The muscle model used is a variation of the 3-element Hill-muscle model [32] , [33] , in which the tendon is simulated as a rigid element ( see Figure 3b ) . This simplification offers significantly higher computational speed at the expense of a relatively small error in accuracy [34] ( see Discussion ) . The model consists of a contractile element placed in parallel to a spring and damper systems . The total force , produced by a given muscle is given by: ( 1 ) where is the force produced by the contractile element , is the force produced by the parallel elastic element , and is the force produced by the parallel damping element . and are given by: ( 2 ) where is the spring constant of the parallel elastic element , is the muscle deformation given by the difference between the current length , and the resting length , of the muscle , is the damping constant of the muscle , and is the rate of change of the muscle length . In most of our experiments we simply set and but we also present results with biologically plausible parameters ( and ) as identified in [35] . The results obtained show that the model produces consistent results for different variations of these parameters ( see Results ) . The active force , is proportional to the activation of the muscle . The activation of each muscle is low-pass filtered to prevent large and instantaneous force variations in the muscle . The filter uses a time constant of and a passband gain of The asymmetric conditioning of the muscles ( i . e . the fact that muscles can only produce contractile , but not extension , forces ) was simulated by the following equation: ( 3 ) The ground contact model , which simulates the reaction forces produced when the end-effector is in contact with the ground , is computed as a spring-damper system: ( 4 ) where and are the spring and damping coefficients of the ground , respectively , is the position of the end-effector on the vertical axis , and is the location of the ground also on the vertical axis ( see also Figure 3a ) . By default these parameters are set to We have varied these parameters to the extent that the contact with the ground looks realistic and obtained similar results to those reported here ( see Results ) . The peripheral system is given by the sensor inputs from , and the motor outputs to , our leg model . On the motor side , we use the muscle activation which defines the force produced by the contractile element of the muscle . We denote the motor signal of muscle On the sensor side , we use approximations to the primary and secondary afferent fibres as observed in biological muscle spindles . In general , we use muscle velocity to approximate the response of primary fibres and muscle length to the approximate the response of secondary fibres . More specifically , the primary afferents are modelled as the rate of change in muscle length and the secondary afferents as the muscle deformation where provides a reference muscle length ( see Figure 3b ) . Note that parameter is different from which is part of the muscle model; whereas the latter is the mechanical resting length of the muscle the former provides a way to set the desired length of the muscle . The parameter is a simplification of fusimotor interactions between γ- and β-motoneurons , which act on the nuclear bag fibres to modulate the sensitivity of the spindle receptors . It can be seen as having a similar effect to that of in the TCT [15] . For each muscle , this reference value is obtained by manually setting the hopping posture of the leg and recording the muscle length . We have tested different leg postures and , as far the leg is kept in natural alignment for hopping ( i . e . with a slight flexion of the hip and knee ) , we could reproduce the results presented here . We denote the primary and secondary sensory afferents from a given muscle as and respectively . When referring to a general sensor input from muscle irrespective of its type , we simply use The model of spontaneous motor activity ( SMA ) , which is carried out during the passive stage , simulates the production of muscle twitches observed during early foetal development [2] , [3] , [36] as well as during sleep throughout the mammalian life span [37] , [38] . This type of activity causes the α-motoneurons to fire spontaneously , which in turn produces muscle contractions independently from sensory stimulation . The model of SMA used in this paper is intended to portray the process of myoclonic twitches observed during REM sleep , many of which “are dominated by a single muscle” [29] ( see Discussion ) . It is carried out by contracting all muscles in sequence; each muscle is contracted ten times allowing seconds between each twitch for the system to recover and stop oscillating . To mimick closer the environmental conditions in which SMA occurs , we disabled gravity during the reflex learning stage . The reason for this is two-fold . First , during sleep the body is typically displaced horizontally , and thus the influence of gravity on the sagital plane of the body is minimal . Second , in uterus the boyance provided by the uterine environment also reduces substancially the effects of gravity . Note thought that , as shown in [39] , this is not a requirement of the system . The sequence of muscle contractions during SMA is and Throughout the entire period of SMA ( i . e . the passive stage ) , the patterns of muscle contractions are fixed and do not change over time , i . e . they are carried out in a purely feedforward way and are thus unaffected by either sensor activity or reflex circuits . This is consistent with observations of SMA triggered in the context of REM sleep , during which reflex circuits seem to be inhibited [40] . For a twitch in muscle we set for seconds . A sequential activation is chosen for practical reasons; the assumptions and limitations of the model are discussed at length in the Discussion section . The self-organization of the reflexes is carried out by using the differential anti-Oja rule similarly to [29] , [30] ( see also [41] ) . This rule is a normalized version of the anti-Hebbian rule [42] , which in turn consists of the additive symmetric of the well-established Hebbian rule [43] ( see also [44] ) . Using the anti-Oja rule , the change in the reflex connection strength at time is given by: ( 5 ) where is the connection between the motor element of muscle and receptor at time is the motor activity of muscle at time , is the sensor activity of receptor at time and is the learning rate . In our system we set Once the reflex circuits are established , the motor signals generated in response to stimulation of the sensor receptors and are as follows: ( 6 ) where is the connection between sensor receptor and motor is the connection between sensor receptor and motor and gains and are global network parameters which modulate the overall strength of the reflex networks ( involving and afferents , respectively ) during the movement phase ( for gain modulation see [45]–[47] ) . The delay between sensor and motor activity is given by one simulator timestep ( ) which is consistent with that of short-latency reflexes like those presented here [48] , [49] . All the experiments described in this paper use a single set of gains to produce the coordinated behaviour ( and ) ; for simplicity we will simply denote these gains as and respectively . The only exception are the experiments carried out to address the hypothesis that behavioural transitions can be achieved by changing the reflex gains over time ( see Results ) . In these experiments , we apply a different set of reflex gains for different phases of the coordinated behaviour; for example , we apply a given set of gains during the stance phase and a different set during the flight phase . All the experiments follow a similar protocol in which reflex circuits are self-organized during the passive stage , and are then modulated in the the active stage . In the passive stage we produce 10 twitches in each muscle sequentially . The sequence of muscle activations is: RF , GM , IL , LB , VI , SB . Modifying this order produces only marginal changes in the reflex matrices obtained ( see also Discussion ) . At each simulation step , the reflex circuits are updated according to eq . 5 . The initial connection weights of the networks , and are set to zero . In the active stage , we set the gain parameters and we drop the leg from a height of and evaluate the hopping pattern obtained against the two criteria described below . It is not the goal of this paper to use feedback control of the hopping height , i . e . we do not have an explicit reference height that is intended to be achieved by the leg . Our goal is to show that once the reflexes are in place , the modulation of the reflex networks is sufficient to coordinate the muscle activity to achieve a stable hopping pattern . In our experiments , we tuned the reflex gains and manually . This is a relatively simple task which resembles that of tuning a PD-controller ( see Discussion ) . To measure the quality of the hopping behaviour we use two criteria that have to be met; we call these criteria hopping stability and conservation of the hopping height . The hopping stability , measures the average difference of the hopping height achieved in two consecutive hops , and it is given by: ( 7 ) where is the total number of hops and is the peak height achieved at hop is measured in millimetres per hop . A stable hopping pattern is given by a low . We set to be a stable hopping pattern , which indicates that on average the difference between two successive hops should be less than ( or equal to ) In addition , the hopping behaviour is also considered unstable if it achieves at least one hopping peak outside of the boundary The conservation of the hopping height , is given by the slope of the line fitted over the peaks of 100 successive hops , and it is measured in millimetres . In the long run , for the hopping height is increasing , for the hopping height is decreasing , and for the hopping height is kept at a stable value . We consider a system where to be able to conserve the hopping height .
We first address the hypothesis that meaningful reflex circuits can be self-organized from SMA . The experiment is carried out by triggering ten twitches in each muscle of the default leg model and by correlating the resulting sensor and motor activity during the passive stage . The reflex circuits obtained are shown in Figure 4 . These circuits were obtained using a twitch amplitude of which produces almost unoticeable muscle contractions . In Figure S1 we show the reflex circuits obtained with an activation of ( see Movie S1 ) . The figures show connectivity matrices in which each element represents a connection between a sensor and a motor elements . Unfilled circles represent excitatory connections , filled circles represent inhibitory connections; the size of the circle represents the magnitude of the connection . From a qualitative point of view the circuits generated are similar in the two sensor modalities . We observe excitatory connections between motor elements and homonymous type-Ia and type-II afferents ( e . g . and respectively ) . The same type of connectivity is obtained for synergist interactions ( e . g . and ) . This connectivity is consistent with the myotatic reflex circuitry . Antagonist interactions are mediated by inhibitory connections ( e . g . and ) . This circuitry is consistent with the reciprocal inhibition reflex . Together the myotatic and the reciprocal inhibition reflexes form the stretch reflex , which is one of the best known circuits in the mammalian spinal cord ( see Discussion ) . The similarity between the two reflex networks is not surprising since the general trend in both sensor inputs is very similar within a given twitch . From a quantitative point of view the differences observed are associated with the inherent difference between the two physical quantities that each receptor measures: positional information for the type-II sensors and velocity information for the type-Ia sensors . In addition , we observe that some of the connections obtained present very small magnitudes . This can be observed for example for the connection between ( or for the reciprocal connection between ) , the magnitude of which is too small to appear in the figure . The reason for this is that these connections mediate sensory and motor elements located at different joints . This can clearly be seen in the case of the Iliacus ( which actuates the hip ) ; when a twitch occurs in the Iliacus , it induces significantly less sensor activity in the muscles around knee ( e . g . or ) than in those around the hip ( e . g . or ) . This reduced sensory activity results in connections with smaller magnitudes . In Figure 5 , we show how the connections between the motor element of the Rectus Femoris , and the sensor afferents of all the muscles , evolve during the passive stage . As can be seen the weights converge to rather stable values after around time at which each muscle has twitched once . When modifying the parameters in eq . 5 our observations are consistent with those in the literature; lower values of ( ) lead to a slower convergence of the weights , and values significantly higher ( ) lead the weights to oscillate without any real convergence . We observe a similar convergence for the other muscles . These results indicate that spinal reflex circuits can be obtained from the self-organization of sensory and motor information induced by SMA . To address the hypothesis that coordinated behaviour can be achieved from the self-organized reflex circuits obtained in the previous section , we drop the leg from a height and search manually for a set of reflex gains ( and ) that can make the leg hop in a stable pattern ( see Models ) . Figure 6a shows the mean and five times the standard deviation of the main kinematic and dynamic variables collected after hops carried out with an appropriate set of gains ( see also Movie S1 . IV ) . These variables are the muscle forces , the hip and knee angles and the ground force ( see Figure S2 for similar results achieved using the reflex circuits obtained with a twitching amplitude of ) . All the variables have been aligned with respect to initial contact with the ground ( ) . As can be seen , the standard deviation of each parameter is relatively low , demonstrating that the hopping pattern is stable ( ) . We can also see that the hopping height is conserved as indicated by the low value of ( Figure 6c ) . To investigate the sensitivity of the results with respect to the ground model used and the weight of the leg we carried out two additional experiments . In the first , we test the system with three different ground models: ground model 1 ( and ) , ground model 2 ( ) and ground model 3 ( ) . In the second we test the system with a leg with the double of the weight of that of the default model ( in total ) . The results obtained are consistent with those presented for the default leg model; they can be observed in the supplementary materials , in Figures S3 , S4 . When analysing the muscle activity produced shortly after the touch down , one can observe a force increase in the extensor muscles and At this stage all these muscles undergo a sudden extension , which increases the activity of their corresponding and sensor receptors; this increased activity , in turn propels the reflex circuits to contract these muscles . A small force is also observed in the ( which is a flexor muscle ) ; this force is mainly caused by the stretch of the synergist When the leg leaves the ground , the situation is inverted: the flexors muscles ( and ) are now those undergoing a stretch as the leg over-extends beyond the desired posture . This causes the reflex circuits to increase the activity of these muscles , which brings the leg to its natural pose and prepares it for the next hop . Both before and after touch down the torques produced at a joint are distributed across the meaningful muscles ( see also [50] ) . This is what we mean by coordinated behaviour; the appropriate muscles are recruited in a timely way such that the final behaviour is attainable . To illustrate the importance of the gain parameter tuning process we show in Figure 6b the results obtained with parameters and smaller than those used in Figure 6a ( see also Movie S1 . II-III ) . With the new gains , the system is clearly unstable ( ) and it does not fulfil the conservation of the hopping height criterion ( ) as we observe a regular decrease in the hopping height ( Figure 6d ) . Moreover this is accompanied by a clear increase in the standard deviation in all the muscle ( and ground ) forces as well as in the hip and knee angles ( Figure 6b ) . A more in-depth analysis of the effect of each gain on the hopping stability is shown in Figure 7 . In this experiment we varied each of the gain parameters and observed the progression of the hopping height . For each plot we modified one of the gains while keeping the other fixed at the value that produced the stable hopping pattern shown in Figure 6b . In Figure 7a we varied and in Figure 7b we varied in each plot represents the value of the respective gain that achieved the stable hopping pattern . The results show that around the value is inversely proportional to the hopping height , i . e . the smaller the gain the higher the hopping height ( Figure 7a ) . This is because large values of reduce the duration of the stance phase , and prevent the leg to markedly change its posture . Consequently , this reduces the deformation in the extensor muscles ( after the contact with the ground ) and decreases the forces that would otherwise be produced by the reflex network modulated by In contrast , the higher the value of the gain the higher the hopping height achieved ( Figure 7b ) . This result is expected since by increasing we are increasing the forces produced in response to a given deformation , without altering the forces produced by the reflex network . Next , we addressed the hypothesis that coordinated behaviour can only occur once meaningful reflex circuits are in place , e . g . reflex circuits that reflect the interactions in the musculoskeletal system . In practice , it could happen that many different reflex circuits , if provided with the right gains , could lead to stable hopping , in a way reminiscent of reservoir computing , where a non-linear recurrent network with the connections weights selected from random distribution , can be shown to produce outputs that can be linearly combined to achieve a desired function [51] , [52] . We have tested more than thirty randomly generated reflex circuits and we have not managed to find appropriate gains to achieve any resemblance to a single hop . The resulting systems mostly fall onto the ground without being able to produce a single hop . These results show evidence that our framework establishes appropriate relations across the different muscles such that a stable hopping pattern can be obtained . To show that the hopping stability obtained is not restricted to the specific hopping height from which the leg is dropped we perform an experiment in which we dynamically change the ground position . In this experiment , while the leg is hopping we increase the ground height from to in intervals , and decrease it back to also in intervals . Our results are shown in Figure 8 . In spite of the varying ground position the leg keeps hopping , which shows the capability of the model to cope with external perturbations ( see also Movie S1 . V ) . To address the hypothesis that our system can naturally adapt to changes in the body morphology , we systematically modified the default musculoskeletal system ( see Figure 3a ) , and analysed how the developmental model adapts to the changes while maintaining the coordinated locomotor behaviour . We made three different types of changes . First , we removed the two bi-articular muscles ( and ) , leaving the leg only with four muscles . Second , we modified the attachment points of the such that the muscle is longer while maintaining the physical connection at the hip and knee . Third , we modify the such that the muscle has the same geometrical path as the i . e . the two muscles are placed in parallel and have the same mechanical effects at both the hip and the knee . In the adaptation experiments , we set the initial reflex circuits to be those identified for the default leg model ( see Figure 3a ) . The experiments start with a passive stage , in which the initial reflex circuits are updated according to the new musculoskeletal configuration; and then continue with an active stage , in which the updated reflex circuits are modulated to achieve a stable hopping pattern , as in the previous sections . With respect to the removal of bi-articular muscles , the reflex weights obtained are similar to those in Figure 4 , taking into account that any connections involving sensor or motor elements of the or are absent ( given that these muscles do not exist in this modified leg model ) . The hopping behaviour achieved with this system is shown in Figure 9a , d . In spite of the larger oscillations in the muscle forces and in the joint angles observed during the flight phase of the movement , the system manages to achieve a stable hopping pattern ( ) and to conserve the hopping height ( ) ( Figure 9d ) . The larger oscillations in the muscle forces and the joint angles achieved by this system , when compared to those in the default system , are consistent with mechanical observations described in [53] , [54] , where the stabilizing role of bi-articular muscles has been put forward . Relative to the changes in the attachment points of the we obtain qualitatively similar reflex connectivity as that in Figure 4 ( not shown ) ; we observe only minor quantitative differences in connections involving the sensor and motor elements of the which are due to changes in the attachment points of the modified system . The hopping behaviour obtained is shown in Figure 9b , e . The profiles of muscle forces and joint angles obtained are relatively similar to those observed with respect to the default leg model ( shown in Figure 9a , c ) . Our results confirm our intuition; given that the two mechanical systems are not fundamentally different the results obtained are very similar , both in terms of reflex circuitry and of behavioural coordination ( ) . The reflex circuits resulting from the modification of the geometrical path of the are shown in Figure 10 . When compared with the circuits obtained for the default configuration ( see Figure 4 ) , one can observe that all the connections with the afferents ( ) and motor elements ( ) have been drastically altered ( marked in red ) ; the only exceptions are the homonymous connections ( ) . In fact , these connections are now very similar to those of the which reflects the identical geometrical path followed by both muscles in the new leg configuration . In Figure 11 we show the progression of all the connections involving the motor element when switching from the default system , to the system with the misplaced , and back to the default system . When the transition to the system with the misplaced occurs , all the connections change signs with the exception of the homonymous connection . Similarly , when changing back to the default system , the connections converge to the values they initially had , presenting a clear case of reflex adaptation . Figure 9c , f , shows the hopping pattern obtained for the system with misplaced In the new configuration , although the system is almost capable of conserving the hopping height ( in Figure 9f ) , it cannot achieve a stable hopping pattern ( ) . The reason for this is that the modified system is rather unbalanced; it has three hip flexors ( and modified-LB ) acting against a single hip extensor ( ) , and three knee extensors ( and modified-LB ) acting against a single knee flexor ( ) . Because all the muscle activities are controlled using only two parameters ( through the reflex circuits ) it is difficult to find a set of parameters that can balance the torques at the joints both during the stance phase ( which requires mostly the activation of extensor muscles ) and the flight phase ( which requires mostly the activation of flexor muscles ) . Using the connectivity matrix obtained for the default system in the modified system , did not result in any behaviour resembling hopping , reinforcing the idea that the symbiosis between reflex matrix and body morphology is an essential element of the model . Furthermore , and consistent with our results for the 4-muscle arrangement , we observe larger oscillations in the muscle forces during the hopping cycle , which reinforces the idea that bi-articular muscles ( which in the new configuration are absent in the posterior part of the leg ) are an important stabilizing mechanism . In the experiments described so far , we observed that we could regulate the hopping behaviour using only a set of gains ( i . e . and ) . This was surprising to us since the two phases of hopping ( stance and flight phases ) have very different requirements in terms of muscle forces . During the stance phase hip flexors and knee extensors are required to pump energy into the system to overcome any energy losses resulting from the contact with the ground , whereas during the flight phase , hip extensors and knee flexors are required to bring the leg as close as possible to its initial posture , which prepares the leg for the next hop . This issue became clearly apparent in the last experiment of the previous section ( see Figure 9c , f ) , in which we had difficulties in balancing the two requirements using only two parameters . A more general way of using the reflex gains , is by exploiting the full potential of eq . 6 and set the gain parameters differently during the stance phase and the flight phase of the hopping behaviour . Indeed , there is plenty of evidence that this is the case with several mammalian coordinated behaviours [7] , [55]–[57] . Although such a system requires the control of more gains ( in our case , two gains over time instead of two fixed gains ) their identification is an easier task , because one can clearly partition the objectives of each phase of the movement . In the case of hopping , during the stance phase , we set the gains such that the leg can jump a certain height , without the constraint that the same parameters will be capable of bringing the leg to its desired posture during the flight phase . And conversely , during the flight phase we set the gains such that the leg can recover its posture , without the constraint that the same gains are able to make the leg jump during the stance phase . The experiments carried out here are similar to those in the previous section , the only difference being that instead of using a single set of gains we use two different sets , one for each hopping phase . During the flight phase we set a pair of gains and such that the leg can jump to a height of and during the flight phase we set gains and such that the leg can recover its original posture . As in the previous experiments the gains were tuned manually . The two phases are differentiated by whether or not the leg is in contact with the ground ( ) ; in the biological system this information could be retrieved from tactile sensors at the end-effector as shown in [58] . We first demonstrate the results of the new strategy using biologically plausible muscle parameters ( and ) . When using this model with only two gains , we could achieve a hopping pattern that was close to stable , but that did not meet our tight stability criteria . However , with four gains we can clearly achieve a very stable hopping pattern ( see Figure S5 ) . Next , the importance of the gain tuning is examined through the case study , which was not fully successful in the previous section , i . e . the one with the modified In Figure 12a , c one can observe the hopping pattern obtained in this system using one set of parameters for each movement phase . Using the new strategy , one can still observe oscillations during the hopping cycle due to the lack of a bi-articular muscle on the posterior part of the leg ( Figure 12a ) . However , when analysing the hopping height achieved we can observe that it is now stable across different hops ( and this contrasts with Figure 9f in which the hopping height changed considerably from one hop to the next To show the generality of our results we show an additional modification of the mechanical system , in which the is placed parallel to the ( see also Movie S1 . VI ) . Similarly to the modification of the we could not obtain a stable hopping pattern for this system with a single set of gains . Our results are shown in Figure 12b , d ( see also Movie S1 . VI ) . With the exception of the muscle oscillations , which are considerably reduced in this system , we obtained very similar results when compared to the system with the modified ( ) . We then tested the modified systems used in Figure 12 , with the reflex circuits learned for the default system . In these experiments the reflex networks do not fully capture the interactions of the musculoskeletal system , in particular the relations with the and the In both modified systems we could only make the leg hop a few times , but could not find a set of gains , and that achieved a stable hopping pattern . Critically , during the flight phase the hip and knee angles diverged progressively from those dictated by the desired posture , until the leg reached a posture which prevents it from jumping . These results reinforce the importance of the coupling between reflex circuits and body morphology , and indicate that , at least in our model , appropriate reflex circuits might be necessary to achieve coordinated behaviour . Another important aspect of the dynamic tuning of gains can be found in the transitions between different behavioural patterns . To test the hypothesis that behavioural transitions can be obtained by changing the modulation gains over time we carried out one additional experiment . In this experiment we show the switching from a standing behaviour , which keeps the leg standing on the ground , to the dynamic hopping behaviour described in the previous sections . We start the experiment by identifying three pairs of gains: one pair that allowed the leg to stand on the ground holding its own weight , another pair that caused the leg to fall down ( basically , ) , and a final pair of gains that produced the hopping behaviour ( here , we use the same gains identified in the context of hopping with the default leg model in Figure 6a , c ) . We then set these gains sequentially; first we set the standing gains , then we set the falling gains for causing the leg to start falling , and finally we set the hopping gains . Our results are shown in Figure 13 ( see also Movie S1 . VII ) . As can be seen , the change in the gain parameters is sufficient for the system to achieve stable hopping starting from a standing position . When investigating the sensitivity of the system with respect to the falling time , we observed that using the same gains we could still hop when setting the falling time to but not when it was set to However , if we change the hopping gains , we can still obtain a stable hopping pattern for a falling time of Overall the results of this section show that our model can be generalized to achieve behavioural transitions by changing the modulation gains over time . The experiments described so far have shown how the system can achieve a stable rhythmic hopping behaviour by modulating the gain parameters and Modifying these parameters changes the way in which the system responds to external perturbations imposed on a given desired posture . In this experiment we would only like to demonstrate how our developmental model can exploit the learned reflex circuits to perform point-to-point trajectories in a way consistent with the TCT [15] ( see also [59] , [60] for a related model ) . The experiment is not intended to be a systematic analysis of the performance of the model on this type of tasks , as this will be a matter of future work , but simply to allow the reader to have a broader interpretation of our developmental model . The experiment uses the reflexes learnt for the default leg model ( see Figure 4 ) . Prior to the experiment , we manually position the leg in three different postures . For each posture , we record the muscle lengths of all the muscles , such that we obtain three sets of muscle lengths and Subsequently , we assign sequentially each of the recorded sets of muscle lengths to the set of desired lengths of all the muscles . This assignment produces a change in the active resting length of the muscle , and induces sensory activity in the secondary afferent fibers . This activity , when propagated through the reflex matrix ( see eq . 6 ) , leads to a change in the resting position of the leg . The three sets of muscle lengths are applied with the leg in the same initial posture , which is the same as that used in the hopping experiments ( see black lines in Figure 14 ) . The and are set manually to minimize the oscillations of the end effector during a given trajectory . The resulting trajectories are shown in Figure 14 ( see also Movie S1 . VIII ) . The figure shows that by using different sets of muscle lengths we can shift the equilibrium posture of the leg , and achieve different end-effector positions . This experiment shows that our model has a natural capability to generate point-to-point movements in a way consistent with TCT ( see Introduction ) .
The mammalian circuitry mediating primary ( type-Ia ) and secondary ( type-II ) spindle afferents , and α-motoneurons is shown in Figure 15 . Relative to the primary Ia afferent fibres [61] , [62] , the α-motoneurons of a given muscle receive direct excitatory connections both from afferents of homonymous ( ) as well as of synergistic muscles ( ) . The same muscle receives inhibitory connections from afferents of antagonist muscles through inhibitory interneurons ( ) . From a functional point of view the connectivity of the secondary afferent is similar to that of the primary afferents , but in the former all the connections seem to be mediated via an additional excitatory interneuron [61] . This circuitry forms the stretch reflex , the purpose of which seems to be to counteract undesired stretches produced by an external load in a given muscle [11] p . 439–40 . When a load causes a muscle to stretch , it activates both the primary and secondary afferents of that muscle , which in turn recruit the α-motoneurons of the homonymous as well as the synergistic muscles [62]–[64]pp . 63-6 . The activation of both afferents inhibits the antagonist muscles and prevents them from counteracting the movement initiated by the agonist muscles [62]pp . 197–200 . The reflexes we obtained with our developmental model are functionally similar to those observed in the mammalian spinal cord ( see Figure 4 ) . We obtain excitatory homonymous ( e . g . and respectively ) and synergist connections ( e . g . and ) , while inhibitory connections are obtained between antagonist muscles ( e . g . and ) . Elsewhere , we have shown that other spinal reflexes can consistently be self-organized using an framework similar to that used here [1] , [30] , [65] for a mechanical system comprising a pair of agonist-antagonist muscles . We have shown the self-organization of the reverse myotatic reflex , which mediates afferent inputs from the Golgi-tendon organs to α-motoneurons , as well as the self-organizion of the withdrawal reflex , which mediates information from cutaneous afferents to α-motoneurons ( in a way similar to [29] ) . We have also obtained successfully the reflex circuitry pertaining the type-Ia afferents and motoneurons . In this context , the work of Petersson and colleagues [29] has been a major source of inspiration for our reflex learning framework . They have observed in vivo that by manipulating the sensory information induced by SMA in the context of REM sleep , they could modify the motor responses pertaining the withdrawal reflex . When taken together with our results , this observation raises the question of whether the general organization of spinal reflexes is determined ( or at least maintained ) by experience-dependent processes induced by SMA , or whether the impact of SMA is restricted to modify only the circuitry pertaining the withdrawal reflex . When compared with our previous work , this paper presents the following additional contributions . First , it includes the self-organization of reflexes involving type-II afferents , which were absent from previous investigations . Second , it shows that our reflex-learning framework scales to a more complicated ( and non-trivial ) mechanical system , involving three rigid bodies , two joints and the interactions between six muscles ( some of which bi-articular ) . Third , it shows how the acquired reflexes can be manipulated to orchestrate the activity of the different muscles and thus produce meaningful coordinated behaviour ( rhythmic and non-rhythmic ) . Fourth , it shows that the entire framework can cope naturally with morphological changes . One essential component of our developmental model is that of SMA [1]–[3] , [36]–[38] . This process consists in the spontaneous activation of α-motoneurons , which produce muscle contractions independently of sensory stimulation . SMA has been widely investigated in the chick and the mouse . In this respect chicks often serve as a good model of higher vertebrates because their spinal organization is similar to that of mammals [66] . ( Although we should be extremely careful when extrapolating the results from chicks which are a precocial species , to mammals , which can be altricial . ) In chicks and rodents , the patterns of SMA undergo strong changes throughout early development , both pre-natally [66]–[69] and post-natally during REM sleep [39] . In the chick , rhythmic bursts of spontaneous activity in α-motoneurons can be recorded by embryonic day 3 ( E3 ) , and before motoneuron axons have reached the extrafusal fibres and produce muscle contractions . The frequency of these patterns has been shown to be essential for guiding the axons to reach their appropriate targets [70] , [71] . Later , by E7 , when the α-motoneurons have already established connectivity with muscle tissue , spontaneous muscle contractions seem to occur often alternating the activation of antagonist muscles [67] but without any clear and distinguishable pattern . This is usually termed as type-I motility [72] . Early behavioural analysis has hypothesised an underlying random process [73] ( see also [74] ) , but electromyography ( EMG ) analysis has shown evidence to the contrary [67] . At this time , muscle afferents start making connections with motoneuron dendrites [75]; such connections seem to be rather coarse and not organized according to any synergist , or antagonist interactions . With embryonic maturation , clear patterns of SMA seem to progressively emerge . By E9 one can distinguish co-activation of flexor and extensor muscles as well as alternation between antagonist muscles [67] . By E18 stable rhythmic patterns showing the coordination between agonist and antagonist muscles within one limb can be observed [68] , [72] . Interlimb coordination seems to be present by E20 [68] . Once the animal is born SMA does not stop; it can still be observed during REM sleep throughout the lifetime of the animal . Like its embryonic counterpart , the patterns of SMA during sleep also seem to undergo a developmental process [39] . Whether all the patterns of SMA observed throughout development involve the same mechanisms seems unlikely . For example , over development brainstem mechanisms seem to be increasingly important for producing twitches [1] , which is very unlikely to occur at early embryonic stages ( see also [76]–[78] for mechanisms of twitching during sleep ) . The spinal cord is not the only place where spontaneous neural activity can be observed; but like spontaneous activity produced in other sites of the brain [79]–[82] , SMA has been shown to contribute to a number of developmental processes such as network formation [69] , synapse elimination [83]–[85] as well as regulation of synaptic strength [86] . At an initial stage of our project , we hypothesised that SMA could drive the actual formation of reflex circuits , which when fully established , could influence back the process of SMA , and produce more structured patterns of activity such as those observed by E9 in the chick . Although this hypothesis is still theoretically possible , the observation of flexor and extensor synergies as early as E9 in the chick , leaves a very short time frame for experience to drive the formation of reflex circuits ( basically from E7–E9 ) . Arguably , the alternation between flexor and extensors observed at E7 is more likely to be produced by some kind of early CPG-like mechanism . In this context , spontaneous motor activity after birth seems to present more favourable conditions for the kind of experience-dependent reflex learning ( or tuning ) that we carry out here . First , the muscle twitches occur against a background of muscle atonia . Second , although some muscle synergies can be observed [39] , “many of these twitches are dominated by a single muscle” [29] . Third , these twitches have already been shown to modulate the spinal circuitry relative to the withdrawal reflex [29] , [87] . Our current hypothesis is that coarse networks are initially established from specific patterns of spontaneous motor activity [69] which are then tuned to the specificities of the particular musculoskeletal system by experience-driven processes ( see also [88] ) . In this paper , we are not in a position to determine the exact role of SMA in forming , maintaining , or fine-tuning spinal reflex circuits . In this , and other papers , we have only pointed out that the correlation between the sensory and motor signals induced by SMA might contain sufficient information to self-organize a number of spinal reflexes in different sensory modalities . The extent to which such information is actually used in the mammalian system can only be determined empirically . Our reflex-learning model takes similar assumptions to those used in Petersson et al [29] to model the tuning of the withdrawal reflex . The only difference is that instead of obtaining stereotypical ipsilateral and contralateral interactions between α-motoneurons and tactile sensors , our model extracts stereotypical homonymous , synergist and antagonist interactions between α-motoneurons and muscle afferents ( in this case spindle afferents ) . In both models , the order of twitching is not a fundamental factor ( see [30] where this order is taken randomly from a uniform distribution ) , which means that they can handle alternation of antagonists but also alternation of synergist and distal muscles . This can be deduced from the fact that the connections with a given motoneuron are only updated when otherwise the they remain unchanged ( see eq . 5 ) . Our model , like that of Petersson et al [29] , should also be able to deal with some degree of muscle synergies being probabilistically activated simultaneously , independently of whether these synergies consist of co-contraction of antagonists , synergists , flexor or extensors . However , we might obtain some inappropriate reflex connectivity if such synergies are consistently activated , e . g . if two muscles are consistently recruited together . Synergies that are consistently recruited impose a strong bias on the musculoskeletal interactions , which in turn influence the reflex circuitry . The extent to which each synergy can be active for the appropriate reflex circuitry to emerge requires further investigation . In addition , our SMA patterns are static , i . e . they are influenced only by the activation of the motor elements and by no other factor . This means that reflex circuits do not contribute to the forces present in the system during the passive stage; this is consistent with the observation that spinal reflexes are inhibited during REM sleep [40] . An implication of the SMA process , is that since it provides millions of muscle contractions over the entire lifespan of a mammalian creature it has a great potential to be the driving force underlying the regulation of sensorimotor circuits as we have shown here [1] . In our model this provides one possible explanation for the role of sensory feedback in spinal adaptation [89] . Using our framework , we have shown that modifications in the musculoskeletal system lead the spinal circuits to re-structure ( see Figure 10 ) providing a clear case for motor adaptation . Addressing this hypothesis in vivo will be a great challenge that might have a strong impact , not only on the motor control community , but also in the sleep research community . To confirm this hypothesis in vivo we will need to carry out experiments in which we analyse how changes in the musculoskeletal system affect known spinal reflex circuits such as the stretch reflex or the withdrawal reflex . The mammalian neural system is plastic and highly redundant , and it is expected that many mechanisms are recruited to achieve complicated behaviours such as hopping , or point-to-point trajectories . In this paper we have simplified to a great extent the overall circuitry of the system . For example , the reflex circuits were modelled using direct connections between sensor receptors and α-motoneurons without any mediating interneurons ( e . g . Renshaw cells ) . The supraspinal control was modelled by setting the resting lengths ( ) for each muscle , and setting the two reflex gains ( and ) . This type of motor control , can be interpreted in two ways . First , it can be interpreted from the point of view of fusimotor control , which assumes that the supraspinal centres are capable of controlling the activity of static and dynamic γ-motoneurons , and thus modulating the sensitivity of type-II and type-Ia spindle afferents . This type of control is theoretically and experimentally grounded [15] , [90] , [91] ( see also [92] ) , although the underlying neural circuits have not yet been completely established . And second , it can be interpreted from the point of view of pre-synaptic inhibition [93] , [94] , in which the reflex synapses modulate the intensity of the reflex responses . Both mechanisms are biologically plausible , and potentially work together . However , whereas fusiform control provides a natural way to account for setting the desired muscle lengths ( ) through the activation of γ-motoneurons , pre-synaptic inhibition can only account for the gain modulation . In addition , the spinal cord contains many circuits other than those addressed in this paper . We neglected the effects of the reverse myotatic and the withdrawal reflexes ( which we were already able to develop with our previous framework [30] ) . This is because , from their connectivity , it was clear that they could only have a marginal role ( and probably even counterproductive ) during hopping . The interactions between these reflexes could easily be included in our model , but for the sake of simplicity we decided to leave them out . Behavioural data in cats [95] and humans [96] have shown that under certain conditions reflexes can be reversed to generate positive ( instead of negative ) feedback reflexes . This has been shown with respect to circuits comprising spindle afferents [7] , [97] as well Golgi-tendon afferents [95] , [96] , [98] . This reflex reversal is not achieved by reverting the nature ( excitatory or inhibitory ) of the established reflex circuits , but rather through oligosynatpic networks which coexist in parallel to these circuits [97] . It has been hypothesised that switching between the positive and the negative feedback circuits is carried out by modulating the reflex gains of the respective networks to favour the expression of ones or the others [97] . From a functional point of view , force ( as well as length ) positive feedback loops can be helpful in providing extra force to the system where other mechanisms might be unable to do so . For example , in the context of our paper , we are not imposing any limitations on the forces provided by the stretch reflex circuits , since we do not impose any upper limit on the reflex gains . In the mammalian system , this is a rather implausible assumption , and often the system might need to recruit force from additional sources other than those that can be provided by negative feedback loops ( as the stretch reflex ) alone . Positive feedback circuits are a good candidate mechanism to provide such forces . In our developmental framework , it would be possible ( in principle ) to obtain both types of feedback loops ( negative and positive ) in a consistent way , if in addition to the anti-Oja rule we allow for other networks to be defined by the standard version of the Oja-rule [41] . Critically , the balancing of the additional reflex networks could then be achieved through a gain modulation process similar to that described in this paper . At this stage of our research , we do not see a direct benefit of adding positive force reflex networks to our model , since we could achieve the desired behaviours without them . However , for a more realistic modelling of the spinal circuitry , such circuits would have to be taken into account . In mammals , there are a number of circuits that contribute to the observed patterns of coordinated behaviour . These range from reflex and central pattern generator ( CPG ) circuits in the spinal cord [99] , [100] , to more central circuits located in the brainstem [6] , cerebellum and cortical areas [101] , [102] . From a locomotor perspective , the roles of reflexes and CPGs have been historically disputed . On the one hand , the observation that deafferented mammals – including humans – can produce locomotor patterns seemed to undermine the contribution of reflexes during locomotion [103] , [104] . In these animals , the rhythmic motion was governed mainly by CPG networks located in the brainstem and spinal cord [6] . On the other hand , the observation that reflexes are modulated during the locomotor cycle , suggests that they are an intrinsic part of coordinated locomotor behaviour [55] , [56] . Currently , it is commonly agreed that reflexes are essential to achieve the full locomotor capabilities observed in mammals , which take into account adaptability to environmental perturbations as well as load bearing [8]–[10] . In our developmental model , we have deliberately left out any circuits pertaining CPGs . Although we are using a rythmic behaviour ( i . e . hopping ) as the main case study for coordinated behaviour , our framework is intended to have a broader interpretation , which is not restricted to periodic locomotor patterns . In general , our hypothesis is that the coordinated motor responses provided by reflexes go beyond a mechanism that deals only with individual perturbations imposed by the environment . Our hypothesis is that reflexes provide training signals ( either directly , or through the induced sensory stimulation ) that can be incorporated in the learning of new behaviours or in the modification of existing ones . This is not intended to dispute , nor undermine , the role of CPGs in producing locomotor behaviours; only to show that , in principle , reflexes also have an inherent capacity to coordinate behaviour . Such capacity can potentially be used to modify the activity of CPG circuits involved in locomotion , but also circuits pertaining other behaviours which are not necessarily connected with locomotion ( e . g . point-to-point trajectories ) . For this to be possible reflexes have to: 1 ) already provide some basic form of coordinated muscle activity , and 2 ) be modifiable to achieve some desired criterion . Although this is not a completely new idea ( e . g . [17] ) , it is one of the aspects we would like to highlight in our paper . In addition , we would like to note that to achieve rhythmic behaviours that require repeated shifts in the body posture , like walking or running , we would actually need to extend the model to incorporate a CPG-like mechanism . Such a mechanism would not only be responsible for setting the different leg postures , during the locomotor cycle , but it could also modulate the reflex circuits in time as we have shown in the context of behavioural transitions ( see also [105] ) . This is currently one of the challenges we are addressing with our model . Our muscle model consists of a spring and damper systems in parallel with a contractile unit . In this paper we do not model the elasticity provided by tendons . Whereas the effect of tendon elasticity would have a negligible effect during the passive stage ( i . e . during reflex learning ) , it could potentially influence the final hopping behaviour , since the latter entails much larger forces than the former . Tendon-elasticity has been shown to contribute to locomotor behaviours in several ways [106] , [107] . First , it provides passive energy storage and recovery , which allows for higher energy efficiency; second , it allows to recoil the muscle-tendon complex faster than the muscle alone . In this paper , we are mainly investigating the effectiveness of the learned reflex circuits in coordinating the overall muscle activity during the hopping behaviour , without any further considerations on the actual maximum muscle forces . In principle , incorporating muscle elasticity in our model should allow to reduce the overall muscle activity during hopping , and thus increase the energy efficiency ( a measure that falls out of the scope of this paper ) . However , we would not expect a significant difference on the capability of the reflexes to orchestrate the activity of the different muscles , provided that the gains and can be modified to accommodate the new dynamics of the system . This hypothesis will need to be addressed experimentally in future work . In our system , the reflex gains have been identified manually . This process of identification resembles that of tuning a general PD-Controller , i . e . a controller of the form where is the output of the controller , is the error between the current value of the variable being analysed and a given desired reference value , and and are constant parameters . The main difference here is that the variable that is being measured is not the same as the variable that is being used by the controller to produce motor activity . In our model , we use the hopping height to measure the quality of the coordinated behaviour , but the controller uses local information to produce the actual muscle activations ( i . e . muscle deformation and velocity ) . This is consistent with the idea that manipulating reflex circuits leads to different equilibrium points ( or trajectories ) without necessarily encoding them explicitly [15] . One aspect of the framework that needs to be verified is the generality of modulating reflex networks for the entire body , in contrast to single joints . We have recently equipped our leg with an upper torso as well as with the muscles required to balance it , and we have tested its capability to hop . Our preliminary results show that hopping and the balancing of the torso can be achieved simultaneously with a single pair of gains ( and ) , but we have not yet managed to achieve a stable hopping pattern with this system ( see also [108] for common muscle synergies during walking and balancing ) . We suspect that the reflex networks that govern the stabilization of the upper torso should be modulated independently from those that coordinate the leg muscles during the hopping , but so far we have not yet managed to frame such a separation in a developmental way . In our framework the muscle synergies are given by the reflex networks , which for a given pattern of sensory stimulation specify the activation profiles of all the different muscles . We believe that this is an important contribution since it realises the concept of muscle synergy at the neural level , and it shows how such synergies can emerge out of musculoskeletal interactions . In doing so we provide a specific and testable model which can be validated or falsified . Mathematically , our synergies can be compared with those in [108] , which consist of a linear combination of muscle activations; the only difference being that the temporal pattern of the synergy is given by the sensory input rather than being fixed . It must be noted that this pattern of sensory input cannot be just any pattern . Artificially induced afferent activity does not necessarily lead to motor coordination from reflex circuits . This has been shown , for example by Klein et al [109] in rats; when stimulating the fifth lumbar dorsal roots they observed periodic patterns of motor activity which were closer to CPG-like mechanisms rather than reflex responses . This observation contrasts with more localized and natural afferent stimulation . For example , the stimulation of spindle afferents of a given muscle produces a well known reflex-like response that resembles the stretch reflex ( this is usually termed the H-reflex ) . This type of stimulation simulates activity of the spindles when a stretch is imposed on the muscle , and thus produces a more natural reflex-like response in the animal . In our model , when the leg touches the ground it also produces a consistent set of afferent inputs in which the extensor spindles are active but not flexor spindles . Such a natural and consistent pattern of sensory stimulation is essential for the induced reflex activity to be able to produce a coordinated motor response . The issue of sensory stimulation , brings us to one of the challenges faced by our model; without sensory inputs there is no muscle coordination ( or muscle synergies ) . This contrast heavily with natural observations in deafferented mammals , which can still carry out coordinated behaviour in the absence of sensory stimulation [21] . From a developmental prespective , this type of observation needs to be carefully contextualised . Let us take the example of the Vestibulo-Ocular Reflex ( VOR ) , which produces contractions in the muscles of the eyes to compensate for rotations in the head ( as sensed by the vestibular system ) . As the VOR can be triggered in the dark [110] and can be observed in subjects who developed blindness [111] , one might be tempted to argue that vision is not a relevant stimulus for this reflex . In fact , the circuitry of the reflex confirms this hypothesis , since it contains only connections mediating vestibular and oculomotor elements . However , studies have shown that subjects who are congenitally blind do not develop the VOR , which brings a new perspective onto the original hypothesis [111] . This example shows that observations in adults are often insufficient to infer whether a certain neural element ( in our case the sensory inputs ) is required or not for a particular behaviour . As in the VOR , spinal and supraspinal systems could be entrained by the continued exposure to the motor patterns produced in the context of a given behaviour ( say , walking ) , and could then be capable of reproducing it after in the absence of sensory stimulation . This hypothesis makes a case for the importance of development to understanding motor control ( see for example , [16] , [28] , [112] , [113] ) ; but , unfortunately we are not yet at a stage where we can address it with our model . In addition , our model allows us to make a case for dimensionality reduction . The dimensionality of our system can be measured in a number of ways . The system can be thought to have 6 dimensions ( the number of muscles to be actuated ) , 18 dimensions ( 6 motor outputs + 12 sensor inputs ) , or 72 dimensions ( i . e . the number of reflex arcs , 6 motor outputs 12 sensor inputs ) . In spite of this , we control only two variables , i . e . and which manipulate the reflex circuits at the network level , rather than control individual elements of the system ( like motor outputs , or single reflex arcs ) . This is only possible because the self-organized reflex circuits already encode meaningful musculoskeletal interactions , and equip the system with a mechanism that coordinates the muscle activity at a local level . Although we do not know exactly how the modulation mechanisms of the central nervous systems work , our model provides a candidate mechanism to realize them . We believe that our developmental model can also contribute to the domain of artificial systems and robotics . The recent increase in the complexity of artificial systems ( i . e . systems endowed with large number of sensor and motor elements [114]–[118] ) is propelling research to identify novel methods that can automatize the process of gathering meaningful information about the body morphology . In this context , SMA can be taken as an analogue to the impulses used in control theory for the purposes of system identification , where they are typically applied to a system in order to characterise ( and model ) its input-output relations [119] ( see also [29] ) . In addition , forms of SMA have also been used in real and simulated systems to structure sensorimotor information [120] , [121] . Like in our reflex circuits , the sensorimotor signals are coupled according to some form of correlation measure; often some measure taken from information theory [120] . However , the correlated signals are typically exploited for some form of perception [121] , or prediction [122] , unlike our reflex circuits which are exploited to coordinate behaviour directly . In this respect , we believe that to make our developmental model fully applicable in an artificial system , we would need to automate the process of learning the reflex gains , which is currently carried out manually . This is out of the scope of the this paper , but it is a feature we are planning to address in future work . Such a learning mechanism could have the benefit of removing the human out of the control loop , and allow the artificial system to automatically acquire information about its body ( through the learning of reflexes ) and exploit it to produce coordinated behaviour . Our developmental model can also contribute to the field of locomotion of artificial systems . In this context , one of the most widely used models is the Spring Loaded Inverted Pendulum ( SLIP ) . This model consists of a point-mass system connected to the ground via a spring . In its original formulation the model assumes the conservation of energy [123] ( i . e . it assumes no energy losses due to contacts with the ground ) but such an assumption has been relaxed in subsequent extensions [124] . The SLIP model has been used ( and extended ) to analyse energy efficiency and behavioural stability during running [123] , walking [125] , hopping [126] as well as during behavioural transitions [127] . In this context , the model of Shafarbi and colleagues [126] provides a good comparison with our model as it addresses directly hopping . Their model , which they called XTSLIP , extends the SLIP model by including an upper torso and a leg ( both with mass ) . The leg consists of a spring and a damper system , which produce forces when perturbed during the contact with the ground . The control of the hopping height is done by changing the resting length of the spring during the stance phase . Some of the premisses of our model resonate with those in the XTSLIP . First , in both models hopping emerges out of the perturbations induced by the ground to the equilibrium position of the system . Second , in our model the muscle activations produced in the context of the network ( which deals with information relative to muscle deformation ) can be seen as an analogue of the mechanical spring in the XTSLIP , the stiffness of which can be controlled by the gain The main difference is that the equilibrium position in our system is maintained via a set of negative feedback loops , and in the XTSLIP it is abstracted in a purely mechanical system . In the XTSLIP , energy losses are dealt with by controlling the resting length of the spring during the stance phase . This modifies the equilibrium point of the system and allows energy to be pumped into it . In our model , the set of desired muscle lengths ( ) can be interpreted as the resting length of the spring in the XTSLIP . We have shown that by modifying these desired lengths we can change the equilibrium point of our system and obtain different leg postures ( see 14 ) . However , during hopping the set of desired muscle lengths is kept constant . In contrast , our strategy to pump energy during hopping consists in modulating the network which deals with information relative to muscle velocity ( i . e . the rate of change in the muscle lengths ) . This strategy can be compared with that of McGeer in his extension to the SLIP model [124] .
|
Mammals display a fascinating behavioural proficiency , which is a remarkable feature given the number of muscles that need to be continuously coordinated . Understanding the processes that give rise to this level of performance has been the main focus of many researchers during the last century , but until now a comprehensive model is yet to be established . From a theoretical point of view , we believe that a key element to understanding mammalian behaviour lies in processes that occur during early development . During this stage , a great deal of information is laid down in neural circuits about the structure of the musculoskeletal system . We believe that this information is essential to achieve coordinated behaviour . In this paper , we propose a model that links three behaviours that appear at different stages of mammalian development . In order of appearance these behaviours are: spontaneous motor activity , spinal reflexes , and coordinated behaviours . Our model shows that the sensor and motor activity induced by spontaneous motor activity is sufficient to self-organize spinal reflexes . These circuits enclose core knowledge about the muscle interactions occurring in the musculoskeletal system . In addition , our model shows that coordinated behaviour can then be achieved simply by modulating the reflex circuits .
|
[
"Abstract",
"Introduction",
"Models",
"Results",
"Discussion"
] |
[
"electroporation",
"neural",
"networks",
"nervous",
"system",
"neuroscience",
"synaptic",
"plasticity",
"computational",
"neuroscience",
"research",
"and",
"analysis",
"methods",
"circuit",
"models",
"specimen",
"preparation",
"and",
"treatment",
"spinal",
"cord",
"developmental",
"neuroscience",
"biophysics",
"neural",
"circuit",
"formation",
"mechanical",
"treatment",
"of",
"specimens",
"cellular",
"neuroscience",
"specimen",
"disruption",
"neuroanatomy",
"anatomy",
"neural",
"homeostasis",
"biology",
"and",
"life",
"sciences",
"sensory",
"systems",
"computational",
"biology",
"motor",
"system",
"biophysical",
"simulations"
] |
2014
|
From Spontaneous Motor Activity to Coordinated Behaviour: A Developmental Model
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.